Urbanisation-induced changes in the morphology of birds from a tropical city

ABSTRACT Urbanisation is accompanied by major environmental changes that impact the structure and functioning of communities and ecosystems, bringing new selective regimes for animal species and for eco-evolutionary dynamics. We aimed to evaluate whether urban intensification promotes ecomorphological changes in birds from a large city in Central Brazil. Analyses were performed on a set of 1314 individuals of 35 species, captured along a gradient of urban intensification. We found significant morphological changes associated with urban intensification by evaluating ten ecomorphological traits and body mass of the species assemblage. Beak length showed the most dramatic changes, and was significantly shorter as a function of urban intensification, mainly in individuals of insectivorous and omnivorous species. These results reinforce the notion that environmental changes caused by human activities in dense urban environments promote new selective pressures in resident bird species.


Introduction
Urban sprawl suppresses native vegetation and turns natural surfaces into impervious surfaces such as roads, houses, and buildings (Mckinney 2002).Urban areas are characterised by increased temperature, noise, pollution, and artificial light, promoting changes in habitat, food resources, and disease ecology for resident animal species (Grimm et al. 2008;Alberti 2015;Des Roches et al. 2021).Despite the little ecological knowledge we have for urban areas compared to natural areas (Chace and Walsh 2006), evidence is accumulating that urbanisation impacts biodiversity (McKinney 2002;McKinney 2006;Leveau et al. 2015;Ibáñez-Álamo et al. 2017;Johnson and Munshi-South 2017;Knapp et al. 2021), promoting a reduction in species richness (La Sorte et al. 2018;Sun et al. 2022;Yang et al. 2022).
In general, urban bird species are smaller, less territorial, have greater dispersal ability, broader dietary and habitat niches, have larger clutch sizes, greater longevity, and lower elevational limits (see review in Neate-Clegg et al. 2023).However, it is worth noting that much of the knowledge on the impact of urbanisation on wildlife comes from work conducted in the global north, with few studies conducted in the tropics (Capilla-Lasheras et al. 2022;Deviche et al. 2023;Neate-Clegg et al. 2023).Understanding the impacts of urban expansion and intensification on tropical biota is highly relevant, considering that most biodiversity hotspots are found in these regions (Mittermeier et al. 2011).
Urbanisation has been accompanied by substantial environmental changes that impact the structure and functioning of communities and ecosystems, bringing new selective regimes for animal populations and consequences for eco-evolutionary dynamics that are still poorly understood (Diamond 1986;Alberti 2015;Johnson and Munshi-South 2017;Alberti et al. 2020).For example, there is a negative relationship between body condition and urbanisation intensity (Liker et al. 2008;Meillère et al. 2015;Caizergues et al. 2018Caizergues et al. , 2021;;Jiménez-Peñuela et al. 2019;Avilla et al. 2021).Also, birds living near urban areas have more micronuclei (small nuclear bodies found close to cell nucleus), indicating a direct impact of pollution on animal health (Baesse et al. 2019).Also, urban animals (e.g.birds and frogs) change the structure of their songs (Slabbekoorn et al. 2007;Halfwerk and Slabbekoorn 2009;Bermúdez-Cuamatzin et al. 2011;Mendoza and Arce-Plata 2013;Halfwerk et al. 2019) and may have higher attrition of their telomeres, potentially directly influencing the dynamics of populations in these locations (Dorado-Correa et al. 2018).Therefore, despite some exceptions (Suárez-Rodríguez et al. 2013;Sandakova et al. 2018), it has become increasingly evident that urban expansion not only excludes or displaces local animal species through habitat changes but also promotes niche impacts and new selective pressures on remaining populations.
Changes in the diversity, quality, and availability of food resources are usually some of the main factors intrinsic to urbanisation (Marzluff 2001;Chamberlain et al. 2009), and can act as selective pressures on traits linked to resource acquisition, such as beak size in birds (Lamichhaney et al. 2015).For example, the availability and predictability of exotic plants and human waste (garbage) in urban environments may increase and keep food supplies constant throughout the year (Shochat 2004).Thus, species living in cities may benefit from exploiting new resources (human waste, or seeds in feeders) (e.g.Carvalho and Marini 2007;Plummer et al. 2019), even influencing their health and body condition.Urban House Finches (Haemorhous mexicanus) evolved morphological changes in their beaks (featuring stronger beaks due to the availability of harder seeds) when compared to their natural conspecifics (Badyaev et al. 2008).Beak morphological changes have also been observed in New Zealand Fantail (Rhipidura fuliginosa) populations.Fantails that inhabited old urban environments showed thicker beaks (shorter and wider) than their rural conspecifics (Amiot et al. 2021).
In this study, we aimed to investigate the effect of urban intensification on the ecomorphological traits of birds from a large city in Central Brazil.We hypothesised that: (H1) the morphology of individual bird species will vary along the urban intensification gradient.However, since each species may respond differently to changes in urban environments, given their foraging and resource acquisition strategies, we further hypothesised that (H2) the morphological changes responding to an urban gradient may vary across species.

Study area
The study was conducted in Brasília, Distrito Federal, Brazil (15°47' Lat S 47°56′ Long W).The city is included in the Cerrado biome, which contains a mosaic of vegetation formations, ranging from open grasslands and savannas to forests (Eiten 1972;Ribeiro and Walter 2008).Planned and built in the late 1950s, Brasília is currently the 3rd largest city in Brazil, 5,760,784 km 2 in size and with an estimated population of over 3 million people (IBGE 2022).After decades of intense occupation, the inhabited area of Brasilia forms an urban gradient organised in a mosaic ranging from neighbourhoods with houses surrounded by gardens and green spaces to those with a predominance of built areas, with greenspaces restricted to flowerbeds between streets and squares between buildings and houses (Figure 1).

Urbanisation index -UI
We built an urbanisation index (UI) condensing six environmental variables directly linked to the development of cities: 'time of urbanisation', 'proximity to urban areas', 'proximity to natural areas', 'proximity to highways', 'population count', and 'artificial light at night (ALAN)' (for data source and manipulation details see Table S1).To create the index, we followed three steps: 1) In possession of the spatial data in raster format, we made a 50 km buffer from the boundaries of Brasilia and cut the rasters for our study area (as we sampled locations close to the boundaries of the studied area, the buffer created was essential to avoid those adjacent urban areas); 2) We condensed the six spatial variables into two axes, with the use of a Principal Component Analysis -PCA (Jollife and Cadima 2016).The two axes retained 67% of the total variation (axis 1 = 46.6%,axis 2 = 20.4%).Both axes represented positive values linked to higher urbanisation intensity, with the variables 'time of urbanisation', 'artificial light at night (ALAN)' and 'population count' being more important in the first axis and the variables 'proximity to urban areas' and 'proximity to highways' being more important in the second axis; and 3) We used the two PCA axes to generate a raster surface that represents our urbanisation index (Figure 1, for detailed access of the acquisition and processing of the rasters and PCA axis values, see Table S1, S2).

Avian sampling
Between April and September 2022, we sampled 115 sites at least 500 metres apart to ensure more sampling independence.The sampling sites covered much of the city's built-up coverage, with some sites located in the peri-urban area (Figure 1).We used mist nets (36 mm mesh, 12 m long, 3 m high, 5 shelves) to capture birds.The number of nets per point varied from 3 to 10 (average = 6 nets) due to the spatial characteristics of the urban matrix of each sampling site.Netting occurred between 0600 and 1700 (mainly until 1200) totalling an effort of 4383 net-h (Table S3).Birds were marked with numbered metallic or coloured bands to avoid undetected recaptures (see the detailed description of the capture sites and the adaptations we made during sampling in Santos et al.In press).

Ecomorphological traits
For each captured individual, we measured body mass (precision pesola scales 0.1 g and 0.5 g) and 10 morphological traits (precision digital pachymeter 0.1 mm, Figure 2): beak length from tip to skull along the culmen ('beak length_Culmen'); beak length to the nares ('beak length_nares'); beak width at the nares ('beak width_nares'); beak width in the mouth ('Gape width'); beak depth at the nares ('beak depth'); 'tarsus length'; 'tarsus width'; 'body size'; and 'tail length'.Also, based on measurements collected from the wing ('wing length' and 'first secondary length'), we calculated the 'hand-wing index (HWI)', which is a metric related to flight efficiency, dispersal ability, migration, and territoriality (Kipp 1959;Lockwood et al. 1998;Sheard et al. 2020;Tobias et al. 2022) (for a detailed outline of measurements associated with the HWI, see Pigot et al. 2020;Tobias et al. 2022).Based on all collected measurements, we used a total of 11 variables in our analyses.To control for the effect of variation in measurements between researchers, all measurements were performed by the same researcher (EGS).
To remove the effect of age on morphology, we excluded all juveniles from the analyses.We included in the analyses species that contained at least five sampled individuals and were present along the entire gradient of urban intensification.After all filters, our analyses included a total of 1314 individuals of native species (81.76% of captures), distributed among 35 species, 18 families, and 7 orders (see list of captures by species in Table S4).Considering the four levels of intensification that we can classify from the sampling points, the number of birds captured along the intensification gradient was balanced (see Figure S1).The species set we analysed was quite variable, from small 6 g hummingbirds (Chionomesa fimbriata), which feed primarily on nectar and have specialised flight abilities, to 300 g pigeons (Patagioenas picazuro), a generalist granivore that usually consumes human food waste.Furthermore, our analysis included representatives of the various feeding guilds, from specialist feeding species such as the Rufous Hornero (Furnarius rufus), to more generalist feeding species such as the Chalkbrowed Mockingbird (Mimus saturninus) (Table S4).

Statistical analyses
To evaluate the effect of the urbanisation gradient on morphology, we performed a Permutation Multivariate Analysis of Variance (PERMANOVA), with 9999 permutations (Anderson 2017), using the vegan package (Oksanen et al. 2022).We used the morphological measurements as response variables and the urbanisation index as the predictor variable.We added the factor 'species' as a covariate in analysis to control the effect of variations among species (given the existing interspecific morphological differences).Then, to identify which variable contributed the most to the observed result, we used the morphometric variables in a Guided Regularized Random Forest analysis -GRRF (Breiman 2001), using the packages 'RRF' (Deng 2013) and 'randomForest' (Liaw and Wiener 2002), building 1000 trees, 1000 times per variable (12100 trees built).This procedure was performed initially to evaluate the set of birds (35 species, Table S4) and later to analyse intraspecific variation (10 species, see below).Before any analysis, we standardised the variables using the function 'decostand' from the vegan package (Oksanen et al. 2022), setting the mean to zero and calculating the variances.In the bird set evaluation (before running the random forest), we did this standardisation by species, as we intended to remove the species effect from the analysis.
For further investigation of the effect of the urbanisation on morphology, we selected the 10 most common native species in our sample: Rufous Hornero, Swallow-tailed Hummingbird (Eupetomena macroura), Chalk-browed Mockingbird, Ruddy Ground Dove (Columbina talpacoti), Rufous-bellied Thrush (Turdus rufiventris), Saffron Finch (Sicalis flaveola), Great Kiskadee (Pitangus sulphuratus), Creamybellied Thrush (Turdus amaurochalinus), Bananaquit (Coereba flaveola), and Pale-breasted Thrush (Turdus leucomelas) (Table S4).For each species, we assessed the association of urbanisation with morphology.Upon detecting significant associations, we analysed by Random Forest the variables that best responded to the urban intensification gradient.We then applied simple linear regressions to assess the significance and magnitude of the association of these variables along the urbanisation index.All analyses were performed using the R programme (R Core Team 2020), and model assumptions were visually evaluated.

Results
There was a significant association between the urban intensification gradient and the morphological traits of the bird assemblage (PERMANOVA, r 2 = 0.00026, F = 11.152,P < 0.001).The variable 'beak length_Culmen' presented the highest contribution to the tree construction in the Random Forest analysis, indicating it has the highest predictive power to explain the morphological variations of the species in our sample (Figure 3).

Discussion
We showed that urbanisation is associated with changes in morphological traits in birds, corroborating that the urbanisation process and its environmental consequences act as selective pressures capable of inducing phenotypic changes in some species.As expected, we did not observe changes for all species evaluated, but surprisingly most of the changes occurred in the beak suggesting an effect of urbanisation on food resource availability.Resource acquisition is a major driver of beak changes in birds, including patterns of evolutionary convergence of species exploiting similar resources in different areas of the world (Pigot et al. 2020;Tobias et al. 2020).Furthermore, only insectivorous and omnivorous species showed significant changes.This highlights that the change in arthropod availability may be a key factor driving morphological changes along the urban gradient.Even with the high mobility of most birds (Winkler et al. 2016), we could observe significant phenotypic changes within a continuous gradient resulting from a recent urbanisation process.This reinforces the warning that the inherent processes of urbanisation act as a driver capable of inducing significant ecomorphological changes in animal populations (Diamond 1986;Johnson and Munshi-South 2017).Thus, we believe that morphological changes in species due to urbanisation should be considered in conservation discussions, given the projected global urban expansion (Grimm et al. 2008;Seto et al. 2010).
Despite the growing interest of researchers in studying the effect of urban sprawl on birds (Yauk et al. 2000;Lepczyk et al. 2017;Marzluff 2017;La Sorte et al. 2018;Baesse et al. 2019), little research has been done assessing morphological changes.Among them, the beak traits are always at the core of discussions because of recurrent observations of changes in this attribute (Badyaev et al. 2008;Giraudeau et al. 2014;Amiot et al. 2021).Beakrelated morphological changes may not only affect the survival of individuals, since the beak is linked to feeding, but also communication and mating since beak morphology alters song.For example, shorter beaks (with consequent reduction of the vocal tract), can alter the frequency of birdsong, promoting changes in sexual selection (Nowicki 1987;Nelson et al. 2005;Badyaev et al. 2008).In fact, an association of beak morphology with song has already been observed in House Finch (Haemorhous mexicanus) (Badyaev et al. 2008;Giraudeau et al. 2014).

Ecomorphological changes in the bird assemblage
Even with such interspecific variability in morphology and life history, our results show that urban intensification promotes strong pressures on the bird community, even more so when we consider the relatively short time of urbanisation (~60 years) of the evaluated city.Although our results are consistent with other work that has demonstrated alteration in bird beaks because of the intensity of urbanisation (Badyaev et al. 2008;Evans et al. 2009;Amiot et al. 2021), we are unaware of any work that has explored the simultaneous change of several syntopic species, as studies typically focus on a single species.This approach allows us to see the big picture, which can better direct our understanding of the generality of the phenomenon and guide wildlifefriendly urban planning.

Ecomorphological changes in bird populations
Of the 10 species evaluated, we found no significant ecomorphological differences for six species.No species with nectarivorous, granivorous, or frugivorous feeding habits showed changes due to the urbanisation gradient, which may indicate that the changes in some resources do not exert significant pressure on some specific guilds.Thus, our results strongly indicate that urbanisation impacts birds differently depending on the type of food exploited in urban areas, supporting our second hypothesis.We found differences in beak sizes for four bird species along the urbanisation gradient, with shorter beaks in areas with higher urbanisation intensity (Figure 4a).Similar results of reduced beaks in more urbanised areas were reported for the Common Blackbird (Turdus merula) (Evans et al. 2009) and the New Zealand Fantail (Rhipidura fuliginosa) (Amiot et al. 2021).We also detected a significant reduction in beak width for Mimus saturninus, suggesting a more fragile beak, opposite to hypothesis of stronger beaks (Figure 4b, see Giraudeau et al. 2014).Given that only insectivorous and omnivorous birds showed a reduction in beak length, we believe that the main factor inducing the beak morphometric changes may come from the effect of urbanisation on the arthropod population.Indeed, urban areas alter fauna composition (Castro et al. 2019;Lövei et al. 2019;Santos et al. 2019;Tzortzakaki et al. 2019) and seem (a) We present the regressions with the variable 'beak length_Culmen', which was the variable indicated as significant for all four species that showed variation along the urbanisation gradient.(b) We present the regressions of the other morphometric variables indicated as relevant in the model, the morphometric variable being distinct in each of the four species evaluated.
to promote the increase in the size of some arthropods available for consumption such as moths, crickets, butterflies, and spiders (Lowe et al. 2014;Merckx et al. 2018), reinforcing our hypothesis.It is important to emphasise that the pattern of change was not absolute, as we had a species considered insectivorous (Turdus rufiventris) that did not show significant changes in the beak.However, despite being considered insectivorous in our approach, this species is considered omnivorous in other studies (Wilman et al. 2014), since half of its diet can be based on fruits.Thus, it is possible that in our region the species feeds more on fruits, not showing changes similar to those observed for insectivores.
We detected a significant reduction in tarsus width for Pitangus sulphuratus in areas with greater intensity of urbanisation (Figure 4b).Thick tarsus suggests greater strength and may be linked to greater exploration of distinct perches.Given that urban areas generally alter perches, simplifying the landscape, we believe that using new perches, such as signs, buildings, and poles, may be linked to the observed reduction in tarsus width.Indeed, Pitangus sulphuratus is a species that usually uses urban substrates ('light poles', 'signs', 'buildings', 'fences', 'asphalt', 'cobblestone', and 'cement') during its foraging (Sick 1997;Martins-Oliveira et al. 2012), reinforcing our interpretation.This change in tarsus width warrants further investigation, looking at the adaptability of animals with thinner tarsus in urban environments.
Another significant change we recorded was the reduction in Hand-Wing Index (HWI) for Turdus amaurochalinus in areas with higher intensities of urbanisation (Figure 4b).This result may indicate that there was a reduction in flight efficiency and movement ability of the species with higher intensity of urbanisation (Sheard et al. 2020;Tobias et al. 2022).Also, the reduction in HWI may be related to increased territoriality (Sheard et al. 2020), suggesting a more sedentary lifestyle in more urban environments.Thus, we have two hypotheses to explain this observed change: migration and territoriality.The hypothesis of an urban influence on the migration of this species should be considered with caution.However, due to the major modifications employed by urban expansion, it is possible that such an impact is likely and future studies may address this issue.In fact, of all the species evaluated in our study, only the Creamy-bellied Thrush is considered migratory or partially migratory (Sick 1997;Somenzari et al. 2018;Collar and de Juana 2020), which reinforces the hypothesis of a possible influence of urban areas on the species' commuting flight patterns.Another hypothesis is that the species is becoming more territorial in areas with greater intensity of urbanisation, with animals becoming more sedentary, exploiting resources more concentrated in plant fragments in urban areas.The urban environment provides new resources and reduces the richness of species competing for those resources (e.g.Huang et al. 2015;La Sorte et al. 2018).Thus, it is conceivable that animals that persist in these environments may be more sedentary and take advantage of resources concentrated in patches of vegetation surrounded by concrete, with constant resources throughout the year (Shochat 2004).The fact that only the Creamy-bellied Thrush showed a change in HWI indicates the greater plausibility of the first hypothesis raised, with urbanisation influencing the displacement of the species in the study area.However, we need more data to understand the role urbanisation plays in bird migrations, which may be on the radar of urban ecologists in future projects.

Conclusion
We observed changes in the ecomorphology of some species of birds living in cities along a continuous gradient.The observed morphological changes were restricted to insectivorous and omnivorous species and were generally concentrated in the length of the beak, with birds living in urban areas presenting shorter beaks.We also found changes in tarsus width for Pitangus sulphuratus and a change in HWI for Turdus amaurochalinus, indicating different traits under selective pressure in urban areas, depending on the habits of each species evaluated.Cities, with their major modifications in various structural, physical, and biotic aspects, appear to be evolutionary cauldrons (Johnson and Munshi-South 2017;Winchell et al. 2022) and much still needs to be clarified regarding their impacts on biodiversity.We believe that understanding the impacts caused by urban sprawl on biodiversity is paramount and should be at the centre of future discussions.

Figure 1 .
Figure 1.Urbanisation index created in our study area, in Brasília, Brazil, to evaluate whether urban intensification promotes ecomorphological changes in birds.We present the map with the sampled sites (blue dots) and Google satellite images of three areas with distinct values of the urbanisation index, showing the large variation of vegetation cover between points with extreme and intermediate indices.The histogram represents the frequency of pixels (30 m × 30 m) in the study area.(a) area with low urbanisation value (15°56′53.64″Lat S, 47°56′9.16″Long W), (b) area with intermediate urbanisation levels (15°48′9.16″Lat S, 47°55′50.38″Long W), and (c) area with high urbanisation value (15°48′43.96″Lat S, 48° 6′48.54″Long W).The red circle represents the 500-metre buffer around the birds capture area.

Figure 3 .
Figure 3. Result of the Random forest analysis used to identify the morphological modifications in the bird assemblage due to urbanisation intensity.The incMSE indicates the percentage increase that the variable brings to the predictability of the generated model (trees), thus indicating a greater change due to urbanisation intensity.The values of each generated tree (grey dots, 1000 trees) and the average contribution of each variable (blue dot) are shown.The morphometric trait 'beak length_culmen' (red dot) was the variable indicated with the greatest addition to model predictability.

Figure 4 .
Figure 4. Result of the random forest analysis used to filter the morphometric variables of the four bird species that showed morphological variation due to urbanisation intensity.The incMSE indicates the percentage of increase that the variable brings to the predictability of the generated model (trees), thus indicating a greater change due to urbanisation intensity.The values of each generated tree (grey dots, 1000 trees) and the average contribution of each variable (blue dot) are shown.The red dots indicate the relevant variables in predicting the intensity of urbanisation.For all four species evaluated, 'beak length_culmen' is indicated as a relevant variable.The other relevant variables changed depending on the species evaluated, showing that they acted differently depending on the species evaluated.

Figure 5 .
Figure5.Linear regressions generated with the morphological variables that showed variation along the urbanisation gradient (indicated in the random forest analysis), to evaluate whether urban intensification promotes ecomorphological changes in birds.(a) We present the regressions with the variable 'beak length_Culmen', which was the variable indicated as significant for all four species that showed variation along the urbanisation gradient.(b) We present the regressions of the other morphometric variables indicated as relevant in the model, the morphometric variable being distinct in each of the four species evaluated.