Mobbing call experiment suggests the enhancement of forest bird movement by tree cover in urban landscapes across seasons

Title Mobbing call experiment suggests the enhancement of forest bird movement by tree cover in urban landscapes across seasons Author(s) Shimazaki, Atsushi; Yamaura, Yuichi; Senzaki, Masayuki; Yabuhara, Yuki; Nakamura, Futoshi Citation Avian Conservation and Ecology, 12(1): Article 16 Issue Date 2017-06 Doc URL http://hdl.handle.net/2115/67006 Rights(URL) https://creativecommons.org/licenses/by-nc/4.0/ Type article File Information ACE-ECO-2017-1013.pdf


INTRODUCTION
Movement, or a change in the spatial location of an organism, is a fundamental ecological process that determines the distribution, population dynamics, and community structure of organisms (Clobert et al. 2001, Nathan et al. 2008).As fragmentation of habitats proceeds, movement of individuals between habitats becomes impeded (Bélisle et al. 2001, Creegan andOsborne 2005), and the risk of extinction of regional and local populations increases (Fahrig 1998, Ferraz et al. 2007).Rapid changes in land use cause habitat loss and degradation as well as habitat fragmentation (Haddad et al. 2015, Newbold et al. 2015).Thus, how the movement of organisms among habitats can be enhanced is a matter of great concern.
Urban areas, which are among the most intensely used land, are often characterized by fragmented landscapes consisting of various dispersal habitats and a variety of factors that limit the movement of organisms, such as residential areas and pavement (Hodgson et al. 2007, Tremblay and St. Clair 2009, 2011).Thus, connectivity between isolated breeding habitats is a function of distance between them and the configuration of the elements comprising urban landscapes (Adriaensen et al. 2003).It also has been suggested that the density of the tree canopy (Gillies et al. 2011), building areas (Hodgson et al. 2007), and electric wire density (Shimazaki et al. 2016) can affect bird movement in urban landscapes.In addition, bird movement behaviors (Krebs 1971, Ekman et al. 1981, Siffczyk et al. 2003) as well as urban environments vary among seasons in urban environments.
In this study, we conducted playback experiments using mobbing calls (Desrochers and Hannon 1997, Bélisle and Desrochers 2002, Clair 2003) in experimental plots with different amounts of three urban elements (tree cover, buildings, and electric wire) during three seasons (breeding, dispersal, and wintering) and estimated the effects of these elements and seasons on the probability of forest birds to cross 50-m distances embedded in the urban area.This can be an important step to predict large-scale bird movements in heterogeneous landscapes based on the empirically measured movement data (Haddad 1999, Shimazaki et al. 2016) because habitat quality cannot be a reliable indicator of movement resistances (Haddad and Tewksbury 2005, Keeley et al. 2016, Ziółkowska et al. 2016).

Mobbing call and its recording
Our focal species were Marsh Tit (Poecile palustris), Varied Tit (Sittiparus varius), Japanese Tit (P.minor), and Eurasian Nuthatch (Sitta europaea), which are resident forest generalist species (hereafter: focal species).We recorded the mobbing calls of focal species (5 Marsh Tits, 2 Varied Tits, a Japanese Tit, and 2 Eurasian Nuthatches) and two other species, a Great Spotted Woodpecker (Dendrocopos major) and a Eurasian Treecreeper (Certhia familiaris), for 30 sec using a perched stuffed Ural Owl (Strix uralensis), which is one of the dominant predators for our focal species in our study region, in mid-April 2013 at the Tomakomai Experimental Forest of Hokkaido University (Shimazaki et al. 2016).Although it is better to use different recordings in each playback experiment to capture potential variations of mobbing calls in different locations or seasons, because the recording included definitive types of mobbing calls by focal species, we used this single mobbing call recording in all playback experiments (see Appendix 1).

Study area and experimental plots
We conducted experiments in Sapporo, Hokkaido, Japan (43°3′ N, 141°20′E) and six adjacent cities (Otaru, Ebetsu, Ishikari, Kitahirosima, Chitose, and Tomakomai; Fig. 1) to examine the effects of three elements included in the urban landscape on bird movement.Sapporo is the fourth largest city in Japan, with a population of 1.93 million.Sapporo and the six adjacent cities are home to 2.57 million people and form a large urban area in the middle of Hokkaido.
We established 121 50×50-m experimental plots in the study area (Fig. 1).The center of one side of each plot (called the starting point) was positioned at the edge of woodlands > 2 ha in area, and the center of the opposite side (goal point) was located at a tree.The plots were 50×50-m square so that the gap crossing probabilities of the birds could be clearly differentiated (Creegan andOsborne 2005, Tremblay andSt. Clair 2009).We tried to place the plots at least 400 m from each other (Tremblay and St. Clair 2009).We used Fundamental Geospatial Data (Geospatial Information Authority of Japan) to measure the ratio of building area (0-35.8%)within plots.We used color aerial photographs to manually identify individual tree canopies, and measured the ratio of tree cover (0-100%) with Quantum GIS ver.1.8.0.The photographs were provided by the Geospatial Information Authority of Japan (http://maps.gsi.go.jp/).We visually checked in the fields if buildings and/or tree cover were still present as they were in the past, because at least six years had passed since the data were generated and the photographs taken.We did not use a plot after tree cover and/or building area had been altered during the experimental periods as a consequence of tree felling or a building being demolished.We measured the total length of electric wire in each plot, and treated it as the wire density.The absolute values of Pearson's correlation coefficient of those variables were ≤ 0.55.We positioned a portable speaker (EUROPORT EPA40; Behringer, Willich, Germany) connected to a player (iPhone4S; Apple, Cupertino, California, USA or NW-E083, Sony, Tokyo, Japan) at the starting (S) and goal (G) points within 1 m from the ground and oriented the speakers to the woodland and S points, respectively.After positioning the speakers, we started playbacks at the S points and recorded the birds attracted within 10 m of the S points.We continued to play the call for 6 min unless 1 min had passed since the last new individual was attracted.Immediately after we stopped playing the call at the S points, the speaker at the G points was turned on to play the call.Then we recorded birds crossing the plots from the S to the G points.If multiple individuals of the same species formed a flock or pair, they were treated as one individual to avoid the problem of pseudoreplication.We ceased the experiment and did not collect the data when tits formed a flock ambiguously (e.g., large time lag between individuals of the same flock crossing the plot) or multispecific flocks formed.The call was stopped when 6 min had passed or all birds had crossed the plots from the S to the G points.We did not record individuals flying from anywhere other than the S to the G points.Playback volume was adjusted to 60 dB (= environmental quality standards for noise in Sapporo city) at 5 m from the speaker, and we confirmed that surveyor at the S points heard the playbacks from the G points (Tremblay and St. Clair 2009).We did not conduct a survey if mobbing calls played at the G point could not be heard at the S point (Tremblay and St. Clair 2009).

Statistical analysis
We estimated the effects of tree cover, building area, electric wire density, and season on the probabilities that focal species crossed the plots (crossing probability) using logistic regression analysis.We treated the number of individual birds that were attracted to the S point in one experiment as the number of trials, and then the number of individuals crossed the plots from the S to the G points among them as the number of successes, and the crossing probability as a success probability of the binomial distribution.We then examined the effects of the predictor variables (tree cover, building area, electric wire density, and season) on the crossing probability.We constructed models for all possible combinations of variables, ranked them using Akaike's information criterion (AIC), and considered significant variables in the best model as meaningful predictor variables.However, we averaged the models if there were multiple well-supported models whose delta AIC scores were < 2, and significant variables in the averaged model were considered meaningful predictor variables.We calculated support for each model as Akaike weight (w i ) and computed a single coefficient using the w i weighted average and the coefficient of each model.We assumed that all variables were included in every model, so the coefficient corresponding to a variable that was not selected was set to zero in some models.We used R (ver.3.2.0)and 'MuMIn' R package (ver.1.15.1) for the model selection and averaging.

RESULTS
We carried out the playback experiments 250 times and observed 618 individuals for the analysis (Table 1).The model selection result showed several models that were well supported, with delta AIC scores < 2 for all focal species (Japanese Tit: six models; Marsh Tit: four models; Varied Tit: five models; and Eurasian nuthatch: six models: Appendix 2).Crossing probability was estimated to be the highest in the breeding season among the three seasons for Varied Tit (Table 2).The building area and electric wire density were not associated with the crossing probability of any of the focal species, and the RVI (Relative Variable Importance) of these variables was mostly low.The crossing probabilities of three focal species other than Varied Tit were significantly positively associated with the tree cover (Fig. 2; Table 2).When the interaction between the tree cover and season was added as a variable in the averaged model for focal species, no significant interaction terms were detected (Appendix 3).

Fig. 2.
Relationship between the ratio of tree cover and crossing probability of four focal species.The continuous variable other than the tree cover were set to average and the season was set to dispersal season.

DISCUSSION
The tree cover had a positive effect on the crossing probabilities of the focal species other than Varied Tit, suggesting that planting trees effectively would promote forest bird movement in urban landscapes (Fig. 2), fitting well with previous studies showing that forest birds use individual trees or small patches as stepping stones in agricultural landscapes (Gillies et al. 2011) and our previous study in the same urban area (Shimazaki et al. 2016).Although we did not formally identify the tree species in the study area, broad-leaved tree species dominated the trees in the urban area.We observed that individual birds used coniferous trees as the perches during the experiments (A.Shimazaki, personal observation).However, forest birds can have preferences in tree species (Holmes andRobinson 1981, Yoshii et al. 2015), and what tree species are planted may be important for the gap-crossing behavior as well as bird movements in urban area.For example, in Japan, Coal Tit (Periparus ater) and Goldcrest (Regulus regulus) are known as specialists of coniferous trees, and their abundance is higher in conifer plantations than in natural forests dominated by broad-leaved forests, though abundance of most species and therefore bird species richness are higher in natural forests (Yamaura et al. 2008(Yamaura et al. , 2009)).These suggest that planting broadleaved trees may be beneficial for many bird species.
The result showing that the building area was not significant in the models for any of the focal species indicates that building density has only minor effects on the crossing probabilities of forest birds.However, our previous study found that increasing building area facilitates movement of tits and nuthatches compared to open areas (Shimazaki et al. 2016), andHodgson et al. (2007) showed that omnivores and nectarivores are more likely to penetrate edges adjoining high-density housing than those adjacent to low-density housing.In our previous study, we did not measure the electric wire length as we did here, and in this study wire density was the second most important variable for Japanese Tit (Table 2).Thus, the effect of building on bird movement might be confounded with that of electric wire.It seems that Psittaciformes and Passeriformes, which were the focus of Hodgson et al. (2007) and are distributed in the southern hemisphere, had different behavioral traits from our focal species, so they responded differently to increased building density.
The crossing probabilities of Varied Tit during the dispersal and wintering seasons were significantly lower than those during the breeding season.Other species also showed the similar seasonal effects though their effects were not significant (Table 2).Because the focal species form and defend territories during the breeding season (Krebs 1971), the strong response to our mobbing playback may be due to territorial aggression of respondents toward perceived intruders.Therefore, the seasonal effect on gap crossing behavior may not be due entirely to a change in gap permeability as much as to a change in testosterone levels.This is where the scale of studies like ours must be extrapolated carefully to larger scale connectivity, given that playback studies in general only test within-territory movement behavior (e.g., Sieving et al. 1996).The interaction between the tree cover and season was not supported in any of these focal species, suggesting that the positive effect of tree cover on the movement of forest birds was consistent across seasons.
We conducted a playback experiment and inferred bird movements at a 50-m scale.Although these small-scale behavioral decisions can be easily measured as the basis of the large-scale predictions (Lima and Zollner 1996, Haddad 1999, Bélisle and Desrochers 2002), large-scale movements may be driven by different motivations and cues, and cannot be extrapolated from small-scale behavioral studies (Desrochers et al. 1999, Zollner 2000, Bélisle 2005, Abrahms et al. 2017).Therefore, our results should be used alongside those generated by other approaches including large-scale translocations (Bélisle et al. 2001, Castellón and Sieving 2006, Gillies and St. Clair 2008), observation of spontaneous movements of birds (Grubb andDoherty 1999, Lees andPeres 2009), and species distributions (Watling et al. 2011).
Responses to this article can be read online at: http://www.ace-eco.org/issues/responses.php/1013

Fig. 1 .
Fig. 1.Study area and experimental plots.We conducted the experiments in Sapporo and six adjacent cities. Dots mean experimental plots.

Fig. A1. 1 .
Fig. A1.1.Examples of sound spectrograms of the mobbing calls of the marsh tit.We draw spectrograms of the type I, II, III calls using the recording used in the playback

Table 1 .
Number of experiments and individuals attracted to start and goal points of each seasons and species.Number of experiments when focal species were successfully attracted to start point.‡ Number of individuals attracted to start point.§ Number of individuals attracted to goal point from start point.| Crossing probability obtained by dividing number of individuals at goal point by number of individuals at start point. †

Table 2 .
Results of model averaging for four species in three seasons.
† Effect of electric wire density.‡Effect of tree cover.§Effect of building area.|Effect of dispersal and wintering seasons compared with breeding season.||Relativevariable importance (single values were obtained for two terms of seasonal effects).
). †Effect of electric wire density.‡Effect of tree cover.§Effect of building area.|Effect of season (indicating that whether the terms were included or not).¶ Differences in AIC with the model having the smallest AIC.||Akaike weights showing the likelihood that the individual model is the best model.