Lockdown effects on fear revealed direct and indirect effects of human presence on perceived predation risk

Abstract

The Covid-19 lockdown reduced drastically human presence outdoors, providing an uncontrolled experiment for disentangling direct and indirect effects of human presence on animal fearfulness. We measured 18,494 flight initiation distances (FIDs, the distance at which individual animals fly away when approached by a human) from 1333 populations of 202 bird species taken in four European cities both before, during and after the lockdown. FIDs decreased during lockdown in rural habitats but increased in urban habitats, especially for singing birds. Height above ground increases during lockdown in non-singing birds only, and birds adjusted horizontal tolerance to approach according to height outside lockdown, in rural habitats and while not singing. Responses showed lagged effects after lockdown in urban but not in rural habitats. Differential responses to lockdown among habitats and between signing and non-singing birds were consistent with relaxation of direct disturbance effects on birds in rural habitats during lockdown, as well as with increased indirect fear effects mediated by predator release in cities. FIDs seemed to measure the balance of direct and indirect effects of humans on predations risk and food needs rather than direct effects of humans on fear.

Graphical abstract

1. Introduction

Our planet is undergoing an accelerated process of humanization (Anthropocene), due to the increasing urbanization, the expansion of intensive land uses, and recreational use of natural ecosystems (Ellis et al. 20102). This human component of global change interacts with components such as climate change and pollution in complex ways (Matesanz et al. 2009; Díaz et al., 2022). The effects of Anthropocene on local diversity is generally assumed to be negative (Sol et al. 2014; Ibañez-Alamo et al. 2017; Johnson et al. 2017; Morelli et al. 2021a), although recent work demonstrates that responses may vary considerably both geographically and for particular species or species groups (Atwell et al., 2012; Díaz et al., 2013; Møller and Díaz, 2018a, Møller and Díaz, 2018b; Morelli et al., 2020, Morelli et al., 2021b; Ordóñez-Delgado et al. 2022).

Animal fitness depends on a continuous balance between foraging, reproductive activities, and the associated risk of being depredated (Stephens and Krebs, 1986). Human presence and activity produce direct and indirect effects on both sides of this balance. For instance, wild predators are less abundant in human impacted habitats due to fear effects or to direct persecution (Díaz et al., 2013; Gotanda, 2020; Hentati-Sundberg et al., 2021; Torre et al., 2022), whereas access to food can be higher for the individuals and species that best tolerate human presence (Møller, 2015; Møller et al., 2015).

Behavior is a mechanism by which animals may flexibly respond to rapid environmental change. These responses effectively translate ecological effects into interactions with hetero- and conspecifics, and the acquisition of resources for survival and reproduction (Huey et al., 2003). The so-called flight initiation distance (FID), the distance at which an animal flees when approached by a human observer under standardized conditions (Blumstein, 2006 ), is a robust estimate of predator avoidance behavior, that involves a trade-off between avoidance of danger and acquisition of basic needs (Cooper and Blumstein, 2015). FID increases under favorable climatic conditions that decrease caloric needs (Díaz et al., 2021) and with increasing abundance of natural and feral predators (Hediger, 1934; St Clair et al., 2010; Gotanda, 2020; Hentati-Sundberg et al., 2021).

Responses to human presence are however variable. FIDs are longer in urban than in nearby rural habitats, which implies negative relationships between human density and FID (Sol et al., 2013; Samia et al., 2015), but both positive and negative relationships within cities have been documented (Mikula, 2014; Morelli et al., 2018; Carlen et al., 2021). Shorter FID in urban habitats is attributed to either habituation to human presence or to natural selection for fearless individuals under conditions of high human disturbance (Samia et al., 2015; van Dongen et al., 2015; Johnson and Munshi-South, 2017; Carrete et al., 2016), emphasizing the importance of direct effects of human presence on FIDs. However, indirect effects mediated by lower predation pressure and higher food availability in cities should also be involved (Díaz et al., 2013; Samia et al., 2015; Møller, 2015; Møller et al., 2015). Hence, FIDs will not only reflect direct effects of human disturbance, but whether direct effects of humans and indirect effects through predators and food give rise to an integrated response. The 2020 lockdown established to control the Covid-19 pandemic was an excellent uncontrolled experiment for disentangling these direct and indirect effects (Rutz et al., 2020).

Flying animals (birds, bats and most insects) have the opportunity to take advantage of ‘escape in the third dimension’ by adjusting fleeing behavior and post or perch height according to predation risk and display needs (Møller, 2010). FIDs have in fact two components: the horizontal distance at which animals flee, and the height above ground at which they were located when flushed (Møller, 2010). Different activities may modulate both predation risk and height selection. Singing birds tend to select higher posts to amplify sound transmission, exposing them more to aerial predators, which are more abundant in rural habitats, but less to ground predators, more abundant in cities (Møller, 2011). Escape from the ground should be costlier than escape from higher perches (Hedenström and Alerstam, 1992). Comparative analyses provided evidence for natural selection on song post height and adjustments in the vertical and horizontal components of escape behavior in relation to variation in predation risk associated with human presence. Birds having higher song posts and fleeing higher in the vegetation tend to have higher survival rates and lower senescence rates, a trend that is facilitated in urban habitats due to human effects on both aerial and ground-dwelling predators (Møller, 2010).

Differential responses to lockdown are thus expected according to habitat, height above ground and singing behavior if indirect effects of humans mediate effects of human presence on FIDs. These differences are not expected in a scenario of direct effects of human disturbance, that would lead to either increased, decreased or invariant FIDs across habitats, bird locations and singing behaviors (Díaz et al., 2013; Samia et al., 2015; Mikula et al., 2022). Specifically, we expected a) differential lockdown effects in urban as compared to nearby rural habitats, with stronger changes in the latter due to lower sensitivity of urban bird populations to changes in environmental conditions (Díaz et al., 2021); b) lack of changes in height above ground of birds when flushed if lockdown decreased direct human disturbance effects, and increased height in the urban an decreased height in rural habitats if effects were mediated by predator release (ground-dwelling predators are prevalent in urban habitats whereas aerial predators prevail in rural habitats; Møller, 2011; Díaz et al., 2022); c) stronger responses to lockdown by non-singing (i.e., foraging or resting) birds than by singing birds, due to constraints imposed by selection of optimal singing heights (Møller, 2011); d) differential adjustments of vertical and horizontal components of FID if predator release is at stake, by increasing the latter according to the former (i.e. longer horizontal FIDs when individuals perch higher) according to lockdown, habitat and singing behavior (Møller, 2010; Díaz et al., 2021); no adjustments, or invariant adjustments according to habitat and singing behavior, are expected under an scenario of direct disturbance effects; and e) delayed effects of lockdown on FIDs to the year after lockdown, as compared with pre-lockdown conditions, if disturbance, predation or food conditions during lockdown had lasting effects of bird behavior (Møller et al., 2013; Gotanda, 2020).

2. Materials and methods

Fieldwork was carried out in and around the cities of Brønderslev (Denmark), Paris (France), Madrid and Toledo (Spain; Supplementary file S1). We estimated flight initiation distance (FID, the distance at which individual birds flee when approached by a human under standardized conditions; Blumstein, 2006) in paired urban and nearby rural sites located 1–20 km away (Díaz et al., 2013). Briefly, trained observers (APM in Denmark and France and MD in Spain) approached individual birds previously detected at least 20 m away by eye or with binoculars during random walks across study areas during the breeding seasons of 2018, 2020 and 2021. We recorded species, age, sex when possible, and whether the birds were singing or not when detected. We avoided sampling young birds or individuals engaged in nest building or offspring provisioning. Only one individual per species and sex was recorded at each specific site to avoid sampling the same individual more than once. We estimated horizontal distance at fleeing by pacing or by means of a Nikon Forestry 550 hypsometer. We estimated height above the ground when the bird was detected by eye or measured it with the hypsometer to the nearest 1 m. FIDs were calculated as the square root of the sum of the squared horizontal distance and the squared height above ground (Samia et al., 2015). Permission by the local authorities during the 2020 lockdown allowed sampling even at restricted sites (i.e. urban parks). We were thus able to sample during (2020) and after (2021) the Covid-19 lockdown the same areas we sampled before lockdown (2018) in previous studies (Díaz et al., 2013, Díaz et al., 2021, Díaz et al., 2022). Paired comparisons according to study sites (see below) control for among-sites (cities and surrounding rural areas) variation in other factors potentially influencing FIDs, such as vegetation structure, human population density or the level of governmental restrictions (Díaz et al., 2022; Mikula et al., 2022; Morelli et al., 2022).

Effects of lockdown, habitat, singing behavior and their interactions on FIDs and their horizontal and vertical (perch or song post height) components were tested using generalized least square regression (PGLS) models on log₁₀-transformed means for all individuals of the same species sampled in each habitat of each city each study year (Díaz et al., 2013). We estimated the phylogenetic scaling parameter lambda (λ), that varies between 0 (phylogenetic independence) and 1 (complete phylogenetic trait conservatism or dependence; Freckleton et al., 2002), to calculate the phylogenetically corrected model estimates. We obtained a consensus phylogeny using Mesquite software (Maddison and Maddison, 2011) on 100 trees extracted from the phylogeny published by Jetz et al. (2012; birdtree.org). Site effects were controlled for by including the different populations of the same species as polytomies with a constant small genetic distance of 0.0001 (Díaz et al., 2013) (Supplementary file S2), that ensure pairwise comparisons before, during and after the lockdown for populations of each bird species in each site × habitat location, and removed statistically site, habitat or year effects other than the ones related to the fixed factors (Mikula et al., 2018). The inverse of sample sizes was used to correct for sampling effort (Garamszegi and Møller, 2010). We used the R libraries ape, MASS and mvtnorm and the function pglm3.3.r (Supplementary file S3). We tested for adjustments between vertical and horizontal components of FID following the same modelling approaches, but using the log-transformed horizontal distance as the dependent variable, log-transformed height above ground as a covariate, and lockdown, habitat, singing behavior and their interactions with height as factors (Díaz et al., 2022). Effect sizes were computed as Pearson's product–moment correlation coefficients from t values (Lipsey and Wilson, 2001) and were judged small (r < 0.10), intermediate (r = 0.11–0.49) or large (r > 0.50) following (Cohen, 1988).

3. Results

We obtained 18,494 FID records from 1333 populations of 202 bird species both before the Covid-19 lockdown (2018, N = 3261 records, 319 populations), during lockdown (2020, N = 3688 and 345) and after (2021, N = 11,545 and 669; Appendix). Phylogenetic, site and sample size effects were significant in all models (Table 1, Table 2). Lockdown, habitat and singing behavior had large effect sizes on FIDs and their components after controlling for these effects (Table 1). Lockdown also interacted with habitat and singing behavior with large effect sizes, although there were no three-way interactions (Table 1). FID and its horizontal component decreased during lockdown, and increased after it in rural habitats, especially for non-singing birds (Fig. 1). Singing birds, and non-singing birds in urban habitats, showed the same trend although less markedly, and singing birds showed the opposite trend (increased FIDs during lockdown) in urban habitats. Mean post height of singing birds did not change in neither habitat and did not differ among habitats. Non-singing birds perched higher during lockdown, and higher in rural habitats; perch height remained higher before lockdown in urban habitats and decreased in rural habitats (Fig. 1).

Table 1.

Results of phylogenetically corrected analyses on the effects of lockdown, habitat, singing activity and their interactions on Flight Initiation Distances (FID) and their horizontal and vertical (height above ground) components. Model statistics, adjusted R², AICc, the λ parameter measuring degree of phylogenetic dependence and its associated tests against the null hypothesis of independence, and the W parameter controlling for sample size effects, are also shown. Boldface indicates p < 0.05. Effect sizes (r) were computed from F (F = √t) values following Lipsey and Wilson (2001).

Source	df	FID			Horizontal distance			Height above ground
		F	p	Effect size	F	p	Effect size	F	p	Effect size
Lockdown	2, 1132	4.53	0.011	0.521	6.59	0.001	0.790	16.19	0.000	0.992
Habitat	1, 1132	207.72	0.000	1.000	179.11	0.000	1.000	31.27	0.000	0.999
Singing	1, 1132	3.67	0.056	0.372	15.25	0.000	0.990	257.52	0.000	1.000
Lockdown × habitat	2, 1132	8.13	0.000	0.891	5.08	0.006	0.609	5.20	0.006	0.626
Lockdown × singing	2, 1132	9.18	0.000	0.929	9.23	0.000	0.930	1.11	0.329	0.037
Lockdown × habitat × singing	2, 1132	0.10	0.909	0.000	0.58	0.563	0.010	2.79	0.062	0.225
Model	12, 1144	23.96	0.000		21.76	0.000		31.3796	0.000
Adjusted R2		0.181			0.167			0.226
AICc		−329.56			251.54			377.33
λ		0.759	χ212 = 314.91, p = 0.000		0.546	χ212 = 199.62, p = 0.000		0.597	χ212 = 356.43, p = 0.000
W		100			300			100

Table 2.

Results of phylogenetically corrected analyses on adjustments of the horizontal component of FID to the height above ground where the birds were posted or perched, and the effects of lockdown, habitat, singing activity and their interactions on these adjustments. Model statistics, adjusted R2, AICc, the λ parameter measuring degree of phylogenetic dependence and its associated tests against the null hypothesis of independence, and the W parameter controlling for sample size effects, are also shown. Boldface indicates p < 0.05. Effect sizes (r) were computed from F (F = √t) values following Lipsey and Wilson (2001).

Source	Horizontal distance
	df	F	P	Effect size
Log10(height above ground)	1, 1126	3.15	0.076	0.284
Height × lockdown	2, 1126	4.26	0.014	0.476
Height × habitat	1, 1126	6.26	0.012	0.760
Height × singing	1, 1126	14.60	0.000	0.988
Height × lockdown × habitat	2, 1126	1.20	0.302	0.043
Height × lockdown × singing	2, 1126	2.13	0.120	0.134
Model	18, 1144	16.90	0.000
Adjusted R2		0.191
AICc		255.93
λ		0.534	χ218 = 167.80, p = 0.000
W		500

Fig. 1.

Mean Flight Initiation Distance (m ± SE) and its components before, during and after the 2020 lockdown associated with the Covid-19 pandemic. Filled symbols, lines: rural habitats; open symbols, dashed lines: urban habitats. Right: singing birds; left: non-singing birds.

Overall, birds did not horizontally adjust flight distance significantly according to the height above ground where they were located or perched. Rather, such adjustments were modulated by lockdown, habitat and singing behavior with moderate to large effect sizes. There were no significant higher-order interactive effects (Table 2). Horizontal distances were negatively related to height both before and after lockdown (i.e., birds that perched higher allowed for closer horizontal approaches), but this effect disappeared during lockdown (Fig. 2, above). Negative relationships between components were steeper in rural than in urban habitats, and birds allowed for closer approaches in urban habitats for the same perch height (Fig. 2, middle). Finally, non-singing birds modulated horizontal flight distances whereas singing birds did not (Fig. 2, below).

Fig. 2.

Adjustment of horizontal distances according to height above ground and their variation according to lockdown, habitat and singing behavior. Top graph (lockdown); grey dots, dashed line: After; filled dots, continuous lines: During; and open dots, pointed line: Before. Middle graph (habitat); filled squares, continuous line: rural; open squares, pointed line: urban. Bottom graph (singing behavior); filled diamonds continuous line: singing; open diamonds, pointed line: not singing). Note log10 (+1) transformations of both components. Lines are linear trends.

4. Discussion

Lockdown, habitat and singing behavior had clear-cut direct and interactive effects on FIDs and their horizontal and vertical components. Responses varied between urban and nearby rural habitats, and on whether birds were singing or engaged in other activities. Overall, direct effects of relaxed human disturbance seemed to predominate in rural habitats, where FIDs decreased by more than a half in non-singing birds to mean distances as short as in urban habitats.

Weaker decreases were found for singing rural birds. Contrarily, no significant FID change was found for non-singing birds in urban habitats, whereas singing birds increased rather than decreased FIDs. This change in post height was consistent with expectations from increased risk of predation from ground-dwelling predators (Møller, 2011), but not with direct disturbance effects. Mean height above ground of non-singing birds increased during lockdown in both urban and nearby rural habitats. This result was again consistent with indirect effects mediated by increased risk of predation from ground-dwelling predators such as feral cats (Díaz et al., 2022), but not with direct disturbance effects.

Singing birds posted higher than non-singing birds, and post height was not affected by either habitat or the lockdown, suggesting strong stabilizing selection for song post height selection and/or low behavioral flexibility for this trait (Møller, 2011). Our results disagreed with those of a recent similar study based on much lower sample sizes measured in four European and one Australian cities (Mikula et al., 2022), where no significant changes in FIDs during lockdown were found. Low power to analyze a heterogeneous data set may account for this contrasting result, as responses of urban birds to lockdown have proven to be generally clear-cut (Derryberry et al., 2020; Montgomery et al., 2021).

Birds that perched higher in the vegetation allowed for closer horizontal approaches than species perched lower (Møller, 2010; but see Blumstein et al., 2004, Díaz et al., 2022 ), a behavioral adjustment that tended to reduce the variability of FIDs towards a species- and site-specific mean value, associated to increased survival and reduced senescence (Møller, 2011). Singing behavior and the Covid-19 lockdown suppressed this adjustment, implying higher FID variability during lockdown and for singing birds. Adjustments were stronger in rural habitats, as expected by the lower variance in FID usually found in urban habitats; this lower variance has been interpreted as resulting from selection of individuals tolerant to human disturbance, under conditions of low predation risk (Samia et al., 2015). Suppression of this adjustment during lockdown and in singing birds may have allowed individuals to better adjust horizontal and height components of FID to varying levels of predation risk. Such adjustments were consistent with increasing risks from ground-dwelling predators, that may have been released by human lockdown.

Overall, our results point to differential mechanisms underlying lockdown effects on bird fearfulness. Reduced direct disturbance seemed to predominate in rural habitats, where non-singing birds would have benefited from reductions in human disturbance. In urban habitats, however, human absence effects seemed indirect, mediated by increasing risk of predation by ground-dwelling predators, most likely the colonies of free-roaming feral cats maintained by animal-loving citizens (Díaz et al., 2022). Lockdown effects lagged to the following year, especially in urban habitats and for singing birds, suggesting stronger selection effects in sedentary urban and/or territorial populations than in more transient rural populations (Møller et al., 2013). Selection by these predators on bird scape behavior may explain time-lagged responses to lockdown in urban habitats, that may compensate for effects of continued human presence on loss of antipredator behaviors (Geffroy et al., 2020). Overall, results were consistent with the idea that FIDs to approaching human observers estimated the balance, on bird fearfulness, of direct and indirect effects of humans on predations risk and food needs, rather than just direct disturbance effects.

The following are the supplementary data related to this article.

Supplementary Appendix

Mean, SE, minimum and maximum (m) Flight Initiation Distances and their horizontal and vertical components according to bird species, city, habitat (U: urban; R: rural surroundings), year (2018, Before the Covid-19 lockdown; 2020, During lockdown, and 2021, After it), and whether birds were singing or not when detected. N = sample size.

Supplementary tables

CRediT authorship contribution statement

Investigation, Data analyses, Writing – original draft: MD Conceptualization, Methodology, Writing – review & editing: MD, APM.

Declaration of competing interest

Authors declare that they have no competing interests.

Editor: Rafael Mateo Soria

Data availability

All data are available in the main text or the supplementary materials.