SEARCH

SEARCH BY CITATION

Keywords:

  • black-capped chickadee;
  • cities;
  • Dendroica petechia;
  • landscape connectivity;
  • most forested route;
  • Poecile atricapillus;
  • roads;
  • translocations;
  • urban development;
  • yellow warbler

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

1. Urbanization represents a major threat to biodiversity world-wide because it causes permanent degradation and fragmentation of biologically rich natural communities. This is particularly acute in coastal plains and river valleys, where cities are typically located. Although movement is essential to the persistence of populations in fragmented landscapes, little is known about how development and transportation corridors affect the movements of wildlife in the urban context.

2. We conducted a series of translocation experiments within the urban landscape of Calgary, Alberta, Canada, to assess the permeability of selected landscape elements for two species of forest songbirds with contrasting adaptabilities to urban development and migratory behaviours: the black-capped chickadee Poecile atricapillus, an urban-adaptable year-round resident, and the yellow warbler Dendroica petechia, an urban-sensitive Neotropical migrant.

3. Birds were caught in riparian habitats and translocated either within the riparian corridor of origin or across the urban matrix. Riparian treatments included continuous forest, one or several transportation bridges and a major river. In the urban matrix, birds were translocated across a single major road, well- or poorly-treed developed areas, or multiple gaps.

4. Using Cox regression we found that the presence of gaps in forest cover explained more variation in return time than the amount of forest cover for both species. Multiple gaps, in particular, resulted in significantly longer return times compared with continuous forest. Chickadees exhibited longer return times when translocated across linear gaps associated with bridges or roads. In contrast, yellow warbler movements appeared to be more constrained by urban development.

5.Synthesis and applications. Our results suggest that improving the permeability of urban landscapes for songbirds can be achieved by preserving connectivity along riparian corridors and other major swaths of natural vegetation while minimizing gaps in vegetation throughout the urban matrix. Our study also demonstrated a cumulative effect of multiple barriers, species-specific response thresholds to canopy cover and gap width, and an important effect of distance-to-territory on movement behaviour. Finally, we demonstrated the utility of ‘most forested route’ as a new, animal-based approach for quantifying the permeability of heterogeneous landscapes.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Habitat loss and fragmentation caused by urban development have been associated with declines or local extinctions of many native taxa including mammals (Hansen et al. 2005), birds (Crooks, Suarez & Bolger 2004) and arthropods (Holway & Suarez 2006). Most studies, however, have focussed on patterns of abundance and distribution of species. There has been comparatively little experimental work aimed at understanding the mechanisms underlying these patterns (Shochat et al. 2006). The concept of functional landscape connectivity, defined as ‘the degree to which the landscape impedes or facilitates movement among resource patches’ (Taylor et al. 1993), explicitly recognizes the behavioural link between process and pattern (Bélisle 2005; Chetkiewicz, St. Clair & Boyce 2006). Increasingly, empirical studies show that connectivity can play a role in explaining patterns of species occupancy (e.g. Betts, Forbes & Diamond 2007; Martensen, Pimentel & Metzger 2008) and genetic diversity (Lindsay et al. 2008; Ortega et al. 2008) in fragmented landscapes. A major challenge in assessing connectivity is that movement is influenced by an individual’s state and motivation. These in turn dictate the choice of destination and the amount of risk or energetic cost an individual is willing to incur (Bélisle 2005).

Experimental approaches have been developed to standardize both destination and motivation, allowing ecologists to quantify the permeability of specific landscape features. For example, researchers have used taped recordings of avian songs or calls to lure birds across small-scale landscape features to assess their willingness to cross these. Such experiments have revealed that the movements of forest birds are often constrained by the presence of gaps associated with anthropogenic features (Desrochers & Hannon 1997; Develey & Stouffer 2001) or natural openings in forest cover (St. Clair 2003). At landscape scales, translocation experiments have been used to evaluate the effects of multiple gaps (Bélisle & St. Clair 2001) and variations in the amount and configuration of forest cover (Bélisle, Desrochers & Fortin 2001; Gobeil & Villard 2002; Gillies & St. Clair 2008) on bird movements. Landscape-scale factors clearly affect movement but these effects often vary among species, possibly due to differences in navigational and flying ability, habitat preference or vulnerability to predation.

To date, experimental evaluations of functional landscape connectivity have been largely restricted to landscapes fragmented by agriculture or forestry, which are typically depicted as islands of suitable habitat within a matrix of non-habitat. The binary landscape model is unrealistic, however, in urban and other heterogeneous landscapes where different land cover types represent varying levels of suitability and permeability (Bender & Fahrig 2005; Chetkiewicz, St. Clair & Boyce 2006). Although there is a need to understand how human infrastructure and land use affect bird movements in urban settings, we know of no studies that have investigated the ability of birds to move within the urban context.

Here, we report on a set of experiments in which we used translocations to assess the permeability of broad-scale features of the urban landscape to the movements of two species of forest songbirds of similar size but with contrasting migratory behaviours and sensitivities to urban development: the black-capped chickadee Poecile atricapillus (Linnaeus 1766; hereafter chickadee), a nonmigratory omnivorous species commonly found within the urban matrix and a regular visitor to bird feeders, and the yellow warbler Dendroica petechia (Linnaeus 1766; hereafter warbler), an insectivorous migratory species that is generally restricted to riparian habitats in the study area. We asked: (i) How do specific elements of the urban landscape affect the movements of translocated birds? (ii) What characteristics of the landscape explain the most variation in return time? and (iii) What strategies can be used to improve the permeability of the urban landscape for songbirds? We were particularly interested in the role of gaps and canopy cover as determinants of movement within a heterogeneous landscape. We predicted that gaps in forest cover would constrain movements, resulting in longer return times, and that multiple gaps would have a stronger barrier effect than single gaps. We also hypothesized that threshold levels in canopy cover would be more important than mean canopy cover conditions to moving birds.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Study area

Our study took place within the city limits of Calgary, Alberta (51°05′N, 114°05′ W), a city of 1·2 million residents located east of the Rocky Mountains in western Canada. The city is bisected by the Bow River and its main tributary, the Elbow River. Riparian corridors are dominated by native balsam poplar Populus balsamifera forests. The intervening urban matrix comprises a mix of residential, commercial, industrial and recreational areas and managed parks. These areas support a forest dominated by a mix of native and non-native species of spruce (Picea spp.), poplar (Populus spp.), birch (Betula spp.) and fruit-bearing trees (e.g. Prunus and Malus spp.). Local canopy cover, measured at the scale of a 50-m radius circle, ranges from 2% to 36% in developed areas to 61–81% in forested natural areas (M. Tremblay, unpublished data). The landscape is heavily fragmented by a busy and rapidly expanding transportation network.

Field experiments

From 5 May to 14 July 2006 and 8 May to 11 July 2007, we performed translocation experiments to assess the relative permeability of various features and land cover types of the urban landscape to the movements of our two study species. Both species generally produce a single brood annually and weigh c. 10 g. Singing, and presumably territorial, males were lured to a mist net using a taped recording of a territorial song and decoys. Each bird was fitted with a US Fish and Wildlife Service metal leg band and two or three coloured celluloid bands to allow for individual identification from a distance. Once banded, birds were placed in cotton bags and transported by bicycle or automobile to the relocation point. During transport, bags were suspended by their drawstrings to ensure proper ventilation and reduce vibrations or impacts with hard surfaces. To further reduce stress, we ensured birds were not exposed to excessive noise or temperature extremes. Mean time in captivity was 53·7 min (±19·6 SD). All handling protocols were approved by the University of Alberta Biosciences Animal Policy & Welfare Committee (Protocol # 465601).

Birds were typically caught in near-natural riparian habitats and moved across selected landscape features either within the riparian corridor of origin or across the urban matrix (Fig. 1). Riparian treatments included: (i) continuous forest; (ii) one or several transportation bridges (which have been shown to inhibit movements along riparian corridors Tremblay & St. Clair 2009); and (iii) the Bow River. Urban matrix treatments included: (i) single road; (ii) well-treed developed areas; (iii) poorly treed developed areas; (iv) multiple linear gaps (e.g. multiple roads or road + Bow River); and (v) multiple gaps, both linear and nonlinear. Roads included in our study consisted of high-volume thoroughfares with at least four lanes and an unforested right-of-way averaging 184 m wide. The mean width of the Bow River was 123 m. We timed translocations to correspond to the nest building, egg incubation, nestling and early fledging stages of the breeding season for each of our study species (early May to mid-June for chickadees; early June to mid-July for warblers). To avoid possible biases associated with the breeding phenology, we spread treatments roughly evenly during the course of each field season. All birds were released in sites offering good tree cover to ensure they had access to shelter and forage before undertaking their return journey, regardless of treatment. Mean translocation distance was 1113 m (±421 SD; see Table S1, Supporting information for specifics).

image

Figure 1.  Examples of translocation trials conducted within riparian corridors (solid lines) and across the urban matrix (dashed lines). Specific treatments illustrated here include: (1) continuous forest; (2) transportation bridge; (3) multiple gaps; (4) residential; and (5) multiple linear gaps (road + river).

Download figure to PowerPoint

Birds were caught between sunrise and 12:00 hours. We checked for the return of each bird to its territory according to the following schedule: 1·5 and 3·0 h post-release, between 18:00 and 20:00 hours in the evening of the capture, and then once a day (approximately midday following morning translocations, for logistical reasons) for the 5 days following release. We limited our observation period to 5 days based on previous translocation studies in which most birds returned within this time period (Bélisle, Desrochers & Fortin 2001; Bélisle & St. Clair 2001; Gobeil & Villard 2002). During each check, one or two observers patrolled the area surrounding the capture site for 30 min, intermittently playing a taped territorial song while watching and listening for birds. We adjusted our search radius based on the observed mean territory radius for each species (100 m for chickadees; 50 m for warblers). Birds that did not display clear territorial behaviour (non-territorial birds would have little incentive to return to the capture site) or were unusually difficult to locate upon their return (thus precluding an accurate measurement of return time) were excluded from our analyses (see Appendix S1, Supporting information for details).

Landscape characterization

With the aid of ArcGIS® 9 (ESRI 2007) and Hawth’s Analysis Tools (SpatialEcology.com 2008), we used two approaches for describing habitat conditions as they might be perceived by translocated birds. First, we described the habitat surrounding each translocation by applying a buffer around each translocation axis equal to 0·25 times the length of this line (Fig. 2a). The result was an ellipsoid with a constant length to width ratio of three, which we believed represented a reasonable search area for a bird seeking to return to its territory (see also Bélisle, Desrochers & Fortin 2001; Gillies & St. Clair 2008). Secondly, we described habitat conditions along a pathway representing the ‘most forested route’ (MFR) between capture and release points (Fig. 2a). Given that forest birds prefer to move through forested detours rather than open areas (Desrochers & Hannon 1997; St. Clair et al. 1998; Hadley & Betts 2009), we felt that an MFR would represent a more realistic pathway between capture and release points than a straight line. MFRs, which are conceptually analogous to ‘least-cost’ paths used in cost-distance modelling (Adriaensen et al. 2003), were manually digitized from a high-resolution (0·5 m) orthorectified digital photo following three simple rules reflecting the assumptions that moving birds seek to minimize risk (i.e. exposure to gaps) and energetic costs (i.e. travel distance): (i) gaps were either circumvented or, in the case of continuous gaps, crossed using the shortest crossing route; (ii) the entire route had to be contained within the ellipsoid; and (iii) the pathway had to reflect continual movement towards the capture point (i.e. no backtracking). For each translocation, we then developed a series of variables describing conditions either within the ellipsoid or along the MFR (Table 1).

image

Figure 2.  Example of determination of forest cover variables for a translocation across a road and river. (a) Capture and release points were connected by the translocation axis (solid line) and the ‘most forested route’ (MFR) (dashed line). An ellipsoid was then created by drawing a buffer around the translocation axis (dotted line). (b) A canopy cover value was assigned to each land cover polygon within the ellipsoid “(darker shades of grey represent higher canopy cover values)”. These values were then used to determine ‘mean’ canopy cover, either within the ellipsoid or along the MFR. (c) Land cover polygons were then reclassed into ‘forested’ (dark grey) or ‘unforested’ (light grey), based on whether canopy cover level was above or below a given threshold value, and these were used to determine ‘binary’ forest cover, either within the ellipsoid or along the MFR.

Download figure to PowerPoint

Table 1.   Definition of landscape variables
CategoryVariableDescription
  1. *Area representing a buffer surrounding the translocation axis with a radius equal to 0·25 times the length of the translocation axis.

  2. †Most forested route between the capture and the release point.

  3. ‡Gaps defined as ‘open’ areas containing <5% canopy cover, based on visual assessment of digital orthophoto.

  4. §Weights proportional to distance from capture site.

  5. ¶As determined from permeability experiments, reflecting a nonlinear relationship between gap width and barrier effect.

  6. MFR, most forested route.

DistancedistanceStraight line distance between capture and release sites (m)
Forest covermean_cancov_ellipMean percent canopy cover in ellipsoid*
mean_cancov_mfrMean percent canopy cover along MFR†
bin_fcov_ellip_XProportion of ellipsoid containing at least X% canopy cover
bin_fcov_mfr_XProportion of MFR containing at least X% canopy cover
Gap‡sum_gap_widthSum width of all gaps along MFR (m)
sum_dist_gap_widthSum of distance-weighted widths of gaps along MFR (m)§
gap_proportionSum length of all gaps/total length of MFR (no units)
max_gap_widthWidth of widest gap along MFR (m)
gap_noNumber of gaps along the MFR (no units)
Barriersum_barrierSum of all barrier values along MFR (no units)¶
sum_dist_barrierSum of all distance-weighted barrier values along MFR (no units)
Forest cover variables

Using existing digital data sets from the City of Calgary, we created a polygon land cover layer in which developed areas were classified according to land use (e.g. residential, commercial, golf course, etc.) and natural areas were classified according to habitat type (e.g. spruce forest, low shrub, grassland or water). From vegetation surveys conducted across the study area, we then determined percentage canopy cover for each land cover polygon (Fig. 2b; see Appendix S2 for details). Finally, we calculated mean canopy cover for each ellipsoid by weighting the canopy cover value of each polygon contained within it by its area. We used an analogous method to calculate mean canopy cover along each MFR.

To test whether birds responded to threshold levels of canopy cover in their movements, we reclassified polygons within ellipsoids as either ‘forested’ or ‘unforested’ depending on whether canopy cover was above or below a given threshold value. We then calculated the proportion of ‘forested’ area within ellipsoids (Fig. 2c). We did this using a suite of threshold values ranging from 2% to 60% canopy cover, which resulted in 15 binary forest cover variables for each ellipsoid. To determine the most relevant threshold for each of our species, we compared the Akaike’s Information Criterion (AIC) values of univariable models explaining return time and retained, for subsequent analyses, only those thresholds generating the lowest AIC. We repeated this process for MFRs.

Gap and barrier variables

We developed a number of variables describing the characteristics of gaps along each MFR (Table 1). Although most of these variables are self-explanatory, one exception is the sum of distance-weighted gaps, which accounted for the possibility that birds may be more risk-averse with increasing distance from their territories (Gillies 2008). Accordingly, we weighted the width of each gap by the distance, in kilometre, between the far edge of the gap and the capture site. In a final set of variables, we accounted for the barrier effect of linear gaps by applying barrier ratings derived from previous gap-crossing experiments from the same study area (Tremblay & St. Clair 2009). These experiments showed that the likelihood of birds crossing gaps did not vary linearly with gap width, but rather, dropped sharply as gap width exceeded 25–50 m (Table S2).

Statistical analyses

We used Cox regression performed in STATA® 10 (StataCorp. 2007) to identify factors affecting the return time of translocated birds. Cox regression (Cleves, Gould & Guttierrez 2004) is useful for analysing ‘time to event’ data (here, return time to territory) and also takes into account whether an event occurs within the observation period (here, return success after 5 days).

In the first set of analyses, we compared differences in return time among treatments. We built separate species-specific models in addition to a combined model containing terms for species, treatment, and the interaction between treatment and species. In a second analysis, we strove to identify the role of landscape factors in explaining the return time of each of our study species. Given the exploratory nature of many of our variables and to reduce the number of candidate models, we divided our variables into three groups (forest cover, gap, and barrier) and retained only the variable from each group that produced the lowest AIC value based on a comparison of preliminary models containing a term for distance plus each landscape variable. For each species, we then built a series of candidate models representing all possible combinations of these retained variables. We included a term for distance in all our models to account for differences in mean translocation distances across treatments. We checked for collinearity among variables, ensuring no model contained variables with a Pearson correlation coefficient >|0·6| (Table S3).

Candidate models were evaluated using an AIC-based information-theoretic approach following techniques described in Burnham & Anderson (2002) and Whittingham et al. (2005). We calculated AIC differences and weights to assess support for each model. We then summed weights to identify the smallest subset of candidate models for which we had a 95% confidence that the set contained the model that best approximated the true model. We determined the relative importance of individual explanatory variables by summing the weights of candidate models containing each variable. This sum represents the probability that a given variable is included in the best approximating model (i.e. selection probability). Finally, we used model averaging to determine a mean coefficient for each variable. We did this by calculating the mean of all coefficients for a given variable weighted by the AIC weight of corresponding models.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We performed 96 and 92 translocations on black-capped chickadees and yellow warblers respectively. Of these, eight chickadees and four warblers were eliminated from analyses because they did not meet our criteria of territoriality and detectability (Appendix S1), leaving 88 retained cases for each species.

Effect of treatment and species on return time and success

In general, chickadees took longer to return to their territories than warblers. By the end of the second visit (3–3·5 h post-release), almost twice as many warblers (84%) had returned to their territories compared to chickadees (47%) and differences in return rate were evident across most treatment categories (Fig. 3a). By the end of the 5-day observation period, return rates remained lower for chickadees than for warblers in all but three treatment categories (Fig. 3b).

image

Figure 3.  Proportion of birds returned to territory after the second (a; 3–3·5 h post-release) and eighth (b; c. 125 h post-release) visit, by treatment and species. Numbers above bars represent sample sizes and are the same for both graphs. (Dev low = poorly treed developed area; Dev high = well-treed developed area).

Download figure to PowerPoint

In translocations within riparian corridors, chickadees that were moved along continuous forest returned relatively quickly to their territories (median return time = 2·1 h; = 12). By comparison, chickadees were significantly slower to return to their territories in the presence of bridges but, surprisingly, returned just as fast when crossing the Bow River (Table 2). In translocations across the urban matrix, return times were significantly longer in the presence of either a single road or multiple gaps of various types. In contrast, developed areas, when not combined with other features, did not affect the homing time of chickadees relative to continuous forest, regardless of tree cover.

Table 2.   Effect of treatment on return time and success for individual species, as determined from Cox regression (= 88 for each species)
VariableHazard ratio*95% Confidence intervalPN
  1. *Change in odds of return corresponding to a one-unit increase in the independent variable (>1 indicates positive effect; <1 negative effect).

  2. †likelihood-ratio X2 = 40·30, d.f. = 8, < 0·0001, log likelihood = −277·1

  3. ‡likelihood-ratio X2 = 24·14, d.f. = 8, = 0·0022, log likelihood = −306·8

  4. §Poorly treed developed areas.; ¶Well-treed developed areas.

Black-capped chickadees†
Distance (m)0·9990·9980·9990·001 
Treatment (ref = forest)    12
 Riparian
  Bridge(s)0·2270·0850·6070·0039
  River1·2950·4863·4500·6059
 Urban Matrix
  Multiple gaps – linear0·1680·0650·4340·00012
  Multiple gaps – all types0·2360·0860·6500·0057
  Single road0·3710·1470·9350·0358
  Developed low§0·6450·2721·5320·32119
  Developed high¶0·8110·3511·8720·62312
Yellow warblers‡
Distance (m)0·9990·9991·0000·025 
Treatment (ref = forest)    14
 Riparian
  River0·4910·1961·2320·13010
  Bridge(s)0·9150·3772·2200·8458
 Urban matrix
  Multiple gaps – linear0·2940·1330·6470·00216
  Developed high0·2250·0810·6270·0048
  Multiple gaps – all types0·3310·1390·7850·01210
  Developed low0·5220·2291·1900·12213
  Single road0·7450·3141·7660·5049

Like chickadees, warblers translocated across continuous forest were quick to return to their territories (median return time <1·5 h; = 14) and took significantly longer to do so when translocated across multiple gaps (Table 2). In contrast, warbler movements within riparian corridors were not inhibited by bridges but were marginally constrained by the Bow River relative to forest. Whereas chickadee movements were unaffected by urban development, warblers exhibited longer return times in developed areas and, surprisingly, this effect was stronger for well-treed than for poorly treed areas. Finally, warblers translocated across a single major road returned to their territory just as quickly as individuals translocated in continuous forest.

Our combined model showed that warblers were significantly more likely than chickadees to return to their territories at any given time across all treatments (hazard ratio = 3·33; = 0·003; Table S4). The same model also revealed a significant species by treatment interaction in which warblers were more likely than chickadees to cross bridges (hazard ratio = 4·012; = 0·037) but less likely to cross the Bow River (hazard ratio = 0·284, = 0·041) and well-treed developed areas (hazard ratio = 0·336; = 0·081).

Effect of landscape variables on return time

In comparing the explanatory power of our binary forest cover variables in univariable models, we found that for chickadees, canopy cover thresholds of 2% and 4% generated the lowest AIC value within MFRs and ellipsoids respectively (Fig. 4). In contrast, higher threshold levels of 20% and 35% within MFRs and ellipsoids, respectively, offered the best fit for warblers. Interestingly, the patterns displayed in Figure 4 were better defined and displayed greater interspecific contrast for MFRs than ellipsoids. Further model comparisons showed that binary forest cover variables generally explained more variation in return time than our variables representing mean canopy cover (see forest cover variables in Table 3). Based on the grouped comparisons in Table 3, we retained for our candidate models for chickadees terms representing binary forest cover along the MFR using a 2% threshold, total number of gaps and distance-weighted barriers. The latter two variables were correlated and thus models containing both terms were eliminated as candidates. For warblers, we retained terms for binary forest cover along the MFR using a 20% threshold, the proportion of gaps and the sum of barriers.

image

Figure 4.  Comparison of the fit of univariable, Cox regression models describing the effect of different threshold values in canopy cover along most forested routes and within ellipsoids on the return time of translocated birds.

Download figure to PowerPoint

Table 3.   Grouped comparisons of preliminary models for selection of candidate variables. All models contained a term for distance to control for differences in translocation distance among treatments. See Table 1 for variable definitions
ModelLLd.f.AICΔ AICWeight
  1. *Variables retained for candidate models.

Black-capped chickadees
 Forest cover variables
  dist + bin_fcov_mfr_2*−289·512583·0150·0000·449
  dist + bin_fcov_ellip_4−289·832583·6550·6400·326
  dist + mean_cancov_ellip−290·522585·0362·0220·164
  dist + mean_cancov_mfr−291·512587·0154·0010·061
 Gap variables
  dist + gap_no*−287·112578·220·0000·882
  dist + gap_proportion−289·692583·385·1590·067
  dist + sum_gap_width−290·342584·686·4600·035
  dist + max_gap_width−291·712587·429·1980·009
  dist + sum_gap_width_dist−291·932587·879·6450·007
 Barrier variables
  dist + sum_barrier_dist*−284·072572·130·0000·680
  dist + sum_barrier−284·822573·631·5040·320
Yellow warblers
 Forest cover variables:
  dist + bin_fcov_mfr_20*−314·152632·300·0000·389
  dist + mean_cancov_mfr−314·362632·720·4160·226
  dist + bin_fcov_ellip_35−314·762633·511·2120·217
  dist + mean_cancov_ellip−314·952633·891·5900·168
 Gap variables
  dist + gap_proportion*−314·152632·300·0000·309
  dist + sum_gap_width_dist−314·362632·720·4160·251
  dist + sum_gap_width−314·762633·511·2120·168
  dist + max_gap_width−314·952633·891·5900·139
  dist + gap_no−314·992633·991·6890·133
 Barrier variables
  dist + sum_barrier*−313·992631·980·0000·627
  dist + sum_barrier_dist−314·512633·021·0420·373

Our 95% confidence set for chickadees contained two of the six candidate models considered. For this species, the sum of distance-weighted barriers and binary forest cover along the MFR had relatively high probabilities of being included in the best approximating model whereas the equivalent probability for the number of gaps was negligible (Table 4). Model-averaged coefficients indicate that the probability of return increased with binary forest cover along the MFR but decreased with distance-weighted barriers and the number of gaps.

Table 4.   Comparison of Cox regression models explaining the return time and success of black-capped chickadees and yellow warblers (= 88 for each species). Only models included in the 95% confidence set for each species are shown here. Numbers in the second to fifth columns represent parameter estimates (hazard ratios) for variables included in each model. See Table 1 for variable definitions
Variable/ModelDistanceBin_fcov_mfr_2Sum_barrier_distGap_noAICΔ AICAIC weightCumulative weight
Black-capped chickadees
1 (AIC best)0·9992·2040·442 570·210·0000·7050·705
20·999 0·430 572·131·9230·2700·975
Selection probability*N/A†0·7170·9750·024    
Mean coefficient*0·9992·1990·4390·686    
Variable/ModelDistanceBin_fcov_mfr_20Sum_ barrierGap_proportionAICΔ AICAIC weightCumulative weight
  1. *Based on complete set of candidate models.

  2. †Variable included a priori in all candidate models.

Yellow warblers
1 (AIC best)0·999 0·830 631·980·0000·1900·190
20·999  0·522632·300·3210·1620·351
30·9991·5400·849 632·440·4560·1510·502
40·9991·720  632·610·6330·1380·641
50·999   632·981·0000·1150·756
60·999 0·8760·671633·181·1980·1040·860
70·9991·379 0·629633·671·6870·0820·942
80·9991·4440·8640·853634·352·3670·0581·000
Selection probabilityN/A†0·4290·5030·406    
Mean coefficient0·9991·5540·8490·629    

For warblers, all eight candidate models considered were included in the confidence set with all models receiving a similar weight (Table 4). Consistent with this, our three candidate variables had a similar, and relatively low, probability of being included in the best approximating model. The return probability of warblers increased with increasing forest cover (albeit to a lesser degree than chickadees) and decreased with the presence of barriers and gap proportion.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Movement behaviour in relation to treatments

Multiple gaps, whether linear or otherwise, inhibited the movements of both our study species even though such features did not necessarily represent impediments to movements when occurring in isolation. Individual roads and rivers, for example, had little or no effect on warbler movements. However, in combination with other gaps, return times of warblers were negatively affected. At least one other translocation study has shown that multiple barriers, even fairly narrow ones, can constrain the movements of some birds (Bélisle & St. Clair 2001). These findings underscore the importance of considering the cumulative effects of barriers when assessing functional landscape connectivity.

Other treatment effects observed were species-specific. The movements of chickadees were constrained by roads and bridges, which is similar to what we found using recorded bird calls to lure chickadees across linear gaps (Tremblay & St. Clair 2009). Yellow warbler movements were unaffected by such structures when tested individually. The warblers’ greater ability to negotiate gaps is likely to be due in part to their presumably superior flying ability associated with their long-distance migratory behaviour. We also observed a marked difference in the ability of our study species to move through developed areas. This difference may stem from the propensity of animals to move through habitats that are also suitable for other functions such as feeding or breeding (Chetkiewicz, St. Clair & Boyce 2006). This may help to explain why areas of urban development were less permeable to yellow warblers, which are strongly associated with riparian habitats in western North America, compared to the more urban-adapted chickadees.

Effect of landscape variables on movement

Consistent with our initial hypothesis, binary forest cover variables generally explained more variation in return time than those describing mean canopy cover, suggesting forest birds may be more sensitive to the presence of a certain minimum level of canopy cover than to average canopy cover conditions encountered along a travel route. We found that threshold values of 2–4% canopy cover were the most relevant for chickadee movements, implying that even sparsely distributed trees may act as effective stepping stones for this species and perhaps explaining why chickadees had no difficulty travelling through developed areas with low canopy cover. Other studies have demonstrated the effectiveness of stepping stones in facilitating inter-patch movements by birds (Boscolo et al. 2008; Gillies & St. Clair 2008; Robertson & Radford 2009), frogs (Angelone & Holderegger 2009) and even insects (Baum et al. 2004; Dover & Settele 2009). In contrast to chickadees, warblers seemed most responsive to threshold levels of 20–40% canopy cover, suggesting that they require higher levels of forest cover for their movements.

Our research provided two lines of evidence suggesting it may be more useful to focus on putative movement paths rather than landscapes or sub-landscapes (e.g. Bélisle, Desrochers & Fortin 2001; Gobeil & Villard 2002; Gillies & St. Clair 2008) when studying animal movements in fragmented landscapes. First, we found more distinct patterns of model fit and sharper interspecific contrasts when comparing forest cover thresholds along MFRs compared to within ellipsoids. Secondly, MFR-based forest cover variables generally outperformed their ellipsoid-based counterparts in our models. Taken together, our findings relative to forest cover illustrate the importance of quantifying habitat conditions in ways that reflect how they are perceived by moving organisms, particularly in heterogeneous landscapes.

The best-fit barrier variable outperformed its gap counterpart in our chickadee models (as measured by selection probability) suggesting that, for this species, the barrier effect of gaps was not linear but increased sharply as a critical width was reached. There is broad support in the literature for threshold distances from the forest edge beyond which songbirds are reluctant to venture (e.g. Desrochers & Hannon 1997; St. Clair et al. 1998; Robertson & Radford 2009; Tremblay & St. Clair 2009), most probably because of perceived predation risk (Rodriguez, Andren & Jansson 2001). For warblers, the best-fit barrier variable performed only marginally better than the retained gap variable. This weak result may indicate an absence of a gap threshold for warblers. Alternatively, it may stem from poorly calibrated barrier variables for this species (the permeability experiments from which we derived our barrier ratings did not include warblers).

The willingness of some birds to cross a gap seems to be influenced by its location relative to a bird’s territory, as evidenced by the negative effect of the sum of distance-weighted barriers on chickadee movements. This indicates a more cautious response to gaps with increasing distance from the territory, which is consistent with one other study of avian movement behaviour (Gillies 2008). A distance effect may explain why chickadees translocated across the Bow River returned to their territories just as quickly as those translocated in continuous forest. This was unexpected given previous studies showing a strong barrier effect of rivers when trying to lure chickadees across them (St. Clair 2003; Tremblay & St. Clair 2009). An alternative explanation is that the river might have facilitated navigation by offering a clear direction of travel and an unobstructed view of a bird’s territory across open water. In contrast, the Bow River had a marginally negative effect on warbler movements despite its proximity to capture sites, association with favourable habitat, and possible role as a navigational aid.

Finally, an obvious question that arises from our work is whether the barrier effects demonstrated by our translocation experiments are applicable to dispersal movements, which are of primary importance to the persistence of populations in fragmented landscapes. The movements of dispersing juveniles in search of a new breeding territory are undoubtedly more exploratory and less directional than that of translocated adults seeking to return to an established territory containing most, if not all, of their annual reproductive investment. Thus, our study probably underestimated the barrier effects of our treatments on dispersing individuals. Moreover, although we did not identify any absolute barriers to movement, our results, coupled with that of previous studies of bird (Desrochers & Fortin 2000) and butterfly (Haddad 1999) movements relative to edges, suggest that gaps in forest probably act as deflectors of movement. This may result in an anisotropic flow of dispersing individuals in heavily fragmented landscapes.

Management implications and conclusions

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Our results suggest two complementary approaches for increasing landscape permeability for urban birds. To benefit urban-sensitive birds like yellow warblers, conservation efforts should focus on preserving habitat connectivity along riparian corridors and other broad swathes of natural vegetation where movements are most likely to occur. Transportation bridges and other interruptions within riparian corridors caused by urban development should be kept to a minimum to allow for unfettered movement along such corridors. Where bridges are inevitable, their barrier effect can be mitigated by flanking them with tall trees to reduce gap width and to provide safe passage for birds above moving traffic (see Tremblay & St. Clair 2009 for specifics).

A second approach, which targets urban-adaptable species like chickadees, consists of enhancing the permeability of the urban matrix itself through the incorporation of elements designed to minimize gaps in vegetation. For example, even sparsely distributed trees can facilitate bird movements through heavily developed areas. Particular attention should be paid to major transportation corridors because they create continuous gaps in forest cover that cannot be circumvented by moving birds. The barrier effect of such features can be at least partially mitigated through the strategic planting of trees along roadsides and on central reservations (medians) with an aim to reduce gap width (see Tremblay & St. Clair 2009). Similarly, the barrier effect of multiple gaps can potentially be reduced by the provision of treed areas between adjacent gaps, which can act as stepping stones. In all roadside plantings, tall trees should be favoured over shrubby vegetation to minimize the risk of collisions with vehicles. Finally, ‘green bridges’ (i.e. structures featuring vegetative cover) may facilitate movement across major transportation corridors.

In summary, our study showed that cities contain significant impediments to the movements of birds although responses to specific landscape elements vary among species due to differences in vagility or adaptability to urban conditions. Although our study focussed on songbirds, it yielded a number of novel insights of broad relevance to the study of animal movements in fragmented landscapes. In particular, it illustrated the importance of considering the cumulative effect of multiple barriers, threshold levels in canopy cover and gap width, and distance to territory on movement behaviour. We also described innovative ways of quantifying landscape conditions in heterogeneous landscapes as they might be perceived by moving organisms.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

This research was supported by a Discovery Grant from the Natural Science and Engineering Research Council (NSERC) and a Canada Foundation for Innovation New Opportunities Grant to CCSC and NSERC and Alberta Ingenuity postgraduate scholarships to MAT. Further funding for the project was provided by The Calgary Foundation (Chris Dunkley Natural Environment Retention Fund, Community Fund, Calgary Beautification Fund, Harvey Kowall Memorial Fund), Alberta Conservation Association, University of Alberta Bill Shostak Wildlife Award, Alberta Sport, Recreation, Parks and Wildlife Foundation, ECO Canada, Nature Calgary, Mountain Equipment Coop, Lamont Development Inc., and O2 Planning and Design Ltd. In-kind support was provided by the University of Alberta, the City of Calgary, and The Orthoshop, Ltd. We are grateful to M. Coombe, S. Crook, K. Hennig, K. Jonasson, E. Miranda, and D. Goldsmith for their invaluable assistance in the field. We also thank E. Bayne, S. Nielsen, and three anonymous reviewers for their insightful comments on previous drafts of the manuscript.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Management implications and conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Appendix S1. Explanation of cases excluded from analyses.

Appendix S2. Details of canopy cover determination.

Table S1. Summary statistics of landscape variables across treatment categories.

Table S2. Barrier rating associated with linear barriers.

Table S3. Correlations among independent variables.

Table S4. Effects of species and treatment on return time.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JPE_1978_sm_appendixs1.doc47KSupporting info item
JPE_1978_sm_appendixs2.doc23KSupporting info item
JPE_1978_sm_tables1.doc53KSupporting info item
JPE_1978_sm_tables2.doc27KSupporting info item
JPE_1978_sm_tables3.doc88KSupporting info item
JPE_1978_sm_tables4.doc50KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.