Introduction

Urbanisation is one of the main worldwide threats to biodiversity, which is increasing at a disproportionate rate compared to other land use changes globally1,2. Future projections suggest that by 2050, 68% of the world’s population will be living in urban areas3. Significant land use changes are thus expected to accommodate the expanding urban population, with substantial pressures on Earth’s ecosystems4,5 and alterations of ecosystem characteristics and functions6,7. Overall negative effects on biodiversity are predicted8,9,10,11, thus adequate urban planning strategies are needed to accommodate both biodiversity requirements and growing human populations across the globe without increasing environmental injustice2. Even though the urban environment generally shows a reduced taxonomic diversity of animal communities in comparison to more natural habitats12, different levels of biodiversity can be found within the heterogeneous urban matrix. Whilst areas with a high cover of impervious surface (e.g. city centre or highly urbanised areas) generally host fewer species, green spaces of cities are characterised by a higher biodiversity13,14,15, with different degrees of species richness mainly dictated by vegetation composition and habitat management16. Green spaces such as urban parks can also act as refugia for rare and threatened species17,18.

Birds show high levels of environmental adaptability19 and behavioural flexibility20, thus making them one of the most successful taxonomic groups in urban environments21,22,23. Bird populations in highly urbanised areas typically consist of few common species (i.e. urban exploiters, hereafter winners), whilst specialised species (i.e. urban avoiders, hereafter losers) are generally filtered out by urbanisation, remaining only in green spaces or rural habitats outside the city12,24,25,26. These dynamics might lead to biotic homogenisation in highly urbanised areas whereby more widespread generalist species become dominant27,28. This process is mainly driven by filtering of species traits29 so that winners possess the necessary traits to exploit resources and ecological conditions that losers cannot30. Winners tend to exhibit generalist habitat selection and foraging behaviour31,32, usually being characterised by larger size, omnivorous diet, longer life-span and non-migratory behaviour, as well as other ecological and life-history traits that allow them to thrive in the urban environment29,33,34,35. On the contrary, losers possess traits of specialist species such as a narrow foraging niche, and they tend to be long-distance migrants or ground nesters31,32, traits that are negatively affected by urban landscape development (especially vegetation characteristics and lack of extensive untouched natural areas), and which ultimately influence species responses to urbanisation36,37,38. Thus, urban landscape features and the expansion of cities play critical roles in shaping urban bird communities.

While urban bird ecology has been widely studied, there are still key gaps in our understanding of how seasonal changes influence species responses to urbanisation. Most urban ornithological studies have focused on the impact of urbanisation during the breeding period37,39,40,41, with few taking into account bird communities year-round with consideration of seasonal effects, although there has been a slight increase in the number of studies in recent years31,35,42. Research on dynamics involving urban bird communities in other periods of the year is uncommon43,44, even though in temperate climates, winter is likely to be a critical period since trophic resources typically diminish. Moreover, the role of species traits in mediating these seasonal responses has been under-investigated, particularly in relation to both breeding and non-breeding dynamics. Addressing these gaps is crucial for a more comprehensive understanding of urban biodiversity and its conservation implications. Nonetheless, the provision of supplementary food is common in some urban areas, whether deliberate (such as through bird feeders, in particular areas of the world) or unintentional (e.g. garbage or areas where discarded food accumulates), which might reduce the risk of starvation and enhance the survival chances of birds45,46. Another benefit of urban living in winter is the urban heat island effect, where urban areas experience higher air temperatures compared to their surrounding areas, which might help birds reduce thermoregulatory costs47 and lower the propensity of urban bird species to migrate48.

We aimed to explore the seasonal dynamics of urban bird assemblages and species traits in different urban habitats by comparing species responses to urbanisation across seasons using a set of six Italian cities distributed along a wide latitudinal range as a model system. Furthermore, we subsequently focused on species-specific responses to identify urban-tolerance plasticity of winners and losers. Our study, by adopting a stratified sampling design based on land use and land cover, provides new insights into the intra-specific variation in response to urbanisation based on the season, and contributes to the few studies that adopted a multi-season approach, considering non-breeding season responses. Species traits, in interaction with time since urban colonisation, are known to affect population trajectories of urban breeding birds49. Similarly, they could affect the species’ ability to dwell in the cities in relation to the different phases of their life cycle, leading to seasonal patterns. Many studies investigating these dynamics have successfully utilised citizen science data43 or have used multiple observers across different cities50,51. Our study uses the same professional surveyors across all cities, thus ensuring consistent data quality, i.e. avoiding temporal biases, variability in detection rates and lack of training52,53,54, providing a complementary perspective to previous studies. No extensive survey on urban bird communities across cities and across seasons has ever been carried out in Italy. Furthermore, deliberate provision of bird food is rare in Italy and more in general in Southern Europe55,56, a factor that can have significant effects on bird responses to urbanisation in the non-breeding season50, thus adding relevance to our study.

We hypothesised that species composition would change between the two seasons along an urbanisation gradient, these changes being mainly dictated by species traits characterising ecological niche (i.e. ecological and life-history traits). In particular, we wanted to identify the traits of winners and losers in the urban environment and whether they changed according to season. We also wanted to test if responses to urbanisation are consistent across seasons in the same species or whether different intra-specific patterns can be observed. Given the predicted increase in urbanisation at the global scale, shedding light on these dynamics throughout the full annual cycle will help to improve our understanding of how the urban landscape shapes the dynamics of bird communities in different seasons. This will be critical for the sustainable management of urban environments and for the development of biodiversity-friendly strategies that take into account urban biodiversity.

Materials and methods

Study area and stratified study design

We sampled bird assemblages in six Italian cities covering a latitudinal gradient of approximately 550 km along the Italian peninsula (Fig. 1). A standardised sampling design was adopted to select sites along a green area size-fragmentation gradient by imposing a 1 × 1 km2 grid on each city57. A gradient of green space area and a gradient of fragmentation degree of individual green spaces, each divided into four strata, were merged to create a 4 × 4 design58. We subsequently selected 1-km squares for each of the 16 combinations of green space area and fragmentation randomly. In each 1-km square, we selected three points, ideally one in green spaces, one in highly impervious-covered areas (i.e. roads and residential or industrial buildings) and one on the border between the two habitats. The number of cells and the numbers of points varied from city to city as not every combination of the stratified design was always available, and there were sometimes access difficulties. Furthermore, points were not located in highly touristic locations where large numbers of pigeons and their potential deliberate feeding could have biased the sampling. We surveyed a total of 72 cells with 47 points in Turin (45°04′ N, 7°40′ E), 45 in Milan (45°27′ N, 9°11′ E), 34 in Florence (43°46′ N, 11°15′ E), 35 in Rome (41°53′ N, 12°28′ E), 32 in Naples (40°50′ N, 14°14′ E) and 27 in Campobasso (41°43′ N, 14°49′ E) for a total of 220 points. We extracted land-use data for each point within a 100-m radius buffer (see Bird surveys) and calculated the % cover of impervious surface areas (imperviousness) by aggregating urban area categories (e.g. residential areas, roads, industrial areas). We used the most recent land cover data available for each region, prioritising datasets with the highest level of detail and accuracy. Where possible, we relied on regional land use maps (e.g., Piemonte 2021 for Turin https://www.geoportale.piemonte.it, Lombardia 2018 for Milan https://www.dati.lombardia.it) and the Copernicus Urban Atlas (2018) https://land.copernicus.eu/local/urban-atlas for the other 4 cities. Additionally, we validated the land cover classifications by cross-referencing them with high-resolution satellite orthoimagery and subsequent field validation to ensure correspondence with actual conditions during the study period. In this way, we created an imperviousness (i.e. urbanisation) gradient that ranged from less impervious areas with high green space cover (e.g. urban parks, natural areas such as woodland/shrubland remnants in the suburbs) to highly impervious areas (e.g. historical city centres, densely populated areas, industrial areas). To assess if low impervious areas were also more likely to be green spaces, we conducted a correlation test between imperviousness % and green spaces % which resulted in a high correlation (Pearson’s r = −0.999, p-value < 0.001). See Table S1 for details on the land use categories and their aggregation for the imperviousness gradient.

Fig. 1
figure 1

Study design and details on sampling methodology. Location of the six cities surveyed in the Italian peninsula (A), details of the study design within the city of Rome, as an example (B) and point counts (black dots) and the 100 m buffer (white circle) with land use (C). Map designed in QGIS Software (datum and projection: EPSG:3003—Monte Mario/Italy zone 1).

Full size image

Bird surveys

For each point, we used standardised 10-min point counts with a 100 m fixed-radius59 which were carried out within the first five hours of the day twice during the breeding season in April–June 2023, and once during the following winter in December 2023–January 2024, for a total of 660-point visits. During the wintering season, we also carried out an extra 5-minute survey immediately after the standard point count period where the observer walked around within the fixed radius and noted any additional species located to increase species detection due to lower singing activity outside the breeding season. We carried out two surveys in the breeding period to account for seasonal variations in the bird community (i.e. some resident species breed in early spring, while others, especially migrant species, breed later). We removed species that were exclusively migratory and not breeding in the area based on the Italian checklist of breeding birds60 and expert opinion. For both seasons, we removed individuals flying over that were not explicitly using the surveyed habitat and non-terrestrial species (e.g. waterfowl, rallids, herons61) from the dataset.

Traits

We extracted data on 11 species-specific traits for the breeding and wintering seasons from the literature62,63,64,65. We only retrieved traits related to species ecology, resource use and life-history that could have a biologically meaningful effect on the adaptation to the urban environment. These were: foraging strata generalism, diet, migratory strategy, body mass, life span, fledgling period, nest type, degree of sociality during and outside the breeding period, number of broods per year and mean clutch size (see Table 1 for details). Prior to analyses, all numerical trait variables were scaled and tested for possible multicollinearity by applying a Spearman correlation test. All traits showed low correlation coefficients (rs < 0.6) and were thus retained in the analyses (see Fig. S1 and S2 for correlation plots and matrixes).

Table 1 Summary of bird traits with explanation of the variable, season for which the trait was used (B = breeding, W = wintering), justification for its inclusion and source.
Full size table

Statistical analyses

We first tested whether bird species occurrence probability changed along the urbanisation gradient and whether these patterns varied according to season by fitting Generalized Linear Mixed Models (GLMMs) using the lme4 R package66 specifying a binomial distribution, using presence/absence data for each species as the response variable and imperviousness, season and their interaction as explanatory variables. We tested for non-linearities in responses to imperviousness by incorporating the quadratic term and its interaction with season when visual inspection of scatterplots suggested the possible existence of non-linear associations. We then selected the best model between a model with and without the quadratic term, based on the lowest AIC. Point identity and city were included in each model as nested random factors. We classified winners based on significantly positive linear responses to urbanisation and losers based on significantly negative responses (p < 0.05). When species showed a non-significant response to urbanisation, or when a non-linear response was significant, we classified them as urban adapters. If the species was absent from a given city due to biogeographical reasons, we carried out the analyses on a subset of cities to avoid zero redundancy in data and model fit problems. GLMMs were carried out on 32 species which had a reasonable sample size and showed no problems of convergence (≥ 10% of the total sample size i.e. ≥ 22 occurrences in the 220 points).

We subsequently assessed whether bird species composition varied along the urban gradient in breeding and wintering periods, and how bird traits were related to urbanisation using the mvabund R package67. For the first analysis, for each season separately we carried out a multivariate Generalized Linear Model (MGLM) by specifying a negative binomial distribution with the manyglm function using the bird species abundance matrix as the response and the imperviousness gradient (as singular-column matrix) as the predictor. For the bird traits, we carried out a fourth-corner analysis using the traitglm function for each season, by adding a third matrix (i.e. trait data) to the previous analysis. In this way, it was possible to identify specific traits associated with winners and losers, based on the association degree of each trait to the urbanisation gradient67. For both analyses, we carried out chi-squared tests by bootstrapping with 999 repetitions and tested the significance of the imperviousness gradient on bird traits by adding a LASSO (Least Absolute Shrinkage and Selection Operator) penalty to improve model fit. LASSO is an optimisation algorithm that performs both variable selection and regularisation in order to enhance the prediction accuracy and interpretability of the statistical model and removes non-significant interactions67. We then created a heatmap for bird traits for each season to visually represent the results. Goodness-of-fit was assessed in the analyses by checking for patterns in model residuals by means of a variogram68. All analyses were performed using R software 4.3.169.

Results

In our surveys, we recorded a total of 18,594 individual birds and a total of 119 species, 93 during the breeding surveys and 79 during the winter period. The 5 most abundant species during the breeding season were feral pigeon Columba livia (1225 individuals), Italian sparrow Passer italiae (946 individuals), European starling Sturnus vulgaris (512 individuals) and Eurasian blackbird Turdus merula (456 individuals). In winter, the most abundant were pigeon (1327 individuals), European starling (464 individuals), Italian sparrow (400 individuals), hooded crow Corvus cornix (336 individuals) and European robin Erithacus rubecula (312 individuals).

Species-specific responses

Species-specific responses to the urbanisation gradient were generally stronger in the breeding season than in the winter (Figs. 2 and 3). Significant effects of imperviousness, both linear and non-linear, were reported for most species (n = 26, 81%), whilst fewer showed a significant seasonal change in occurrence (n = 17, 53%). Winners (Fig. 2A) were represented by species that are closely associated with human settlements and highly impervious areas such as pigeon, yellow-legged gull Larus michahellis and black redstart Phoenicurus ochruros (only in the breeding season). A higher number of species was categorised as losers (Fig. 2B) which were negatively affected by urbanisation; these were species mainly tied to forested and green spaces (e.g. short-toed treecreeper Certhia brachydactyla, Eurasian nuthatch Sitta europea, Eurasian green woodpecker Picus viridis). A large number of species emerged as urban adapters (Fig. 3), most of them showing non-linear responses, but there were a few with non-significant responses (i.e. European greenfinch Chloris chloris, Eurasian jackdaw Corvus monedula, Eurasian jay Garrulus glandarius, monk parakeet Myiopsitta monachus, common chiffchaff Phylloscopus collybita). See Table S2 for species-specific model outputs.

Fig. 2
figure 2

Regression plots from the species-specific GLMMs with occurrence probability in relation to the imperviousness gradient across seasons for winners (A) and losers (B). Point identity and city were added as nested random factors in the models. Winners were classified based on significantly positive linear responses to urbanisation and losers based on significantly negative responses. For black redstart (Phoenicurus ochruros) a significant (p < 0.05) interaction between imperviousness and season was highlighted. See Table S2 for results.

Full size image
Fig. 3
figure 3

Regression plots from the species-specific GLMMs with occurrence probability in relation to the imperviousness gradient across seasons for urban adapters. Point identity and city were added as nested random factors in the models. Urban adapters were classified based on non-significant or significantly non-linear responses to urbanisation. For blue tit (Cyanistes caeruleus), blackcap (Sylvia atricapilla), Sardinian warbler (Curruca melanocephala) and blackbird (Turdus merula) a significant (p < 0.05) interaction between imperviousness and season was highlighted. See Table S2 for results.

Full size image

Bird species composition and traits

There was significant variation in bird species composition and bird traits across the urban gradient, both during the breeding (composition: χ² = 8.95, p-value = 0.003; traits: χ² = 7.00, p-value = 0.027) and winter (composition: χ² = 7.71, p-value = 0.035; traits: χ² = 5.78, p-value = 0.030) seasons. Abundance increased with increasing urbanisation during the breeding season in species which breed colonially, have a greater number of broods per year, a longer life span and, to a lesser extent, closed arboreal nesting and a granivorous diet. Invertivores, ground- and close to ground-nesting species, short-distance migrants and non-colonial species were more abundant in areas of low impervious cover (i.e. higher cover of green spaces). During winter, species with a greater number of broods per year, solitary behaviour and the ability to use multiple foraging strata had higher abundances with increasing urbanisation. On the contrary, species with larger clutch sizes and short-distance migrant species were more abundant in green spaces (Fig. 4).

Fig. 4
figure 4

Heatmaps of the relationships between bird traits and the imperviousness gradient for the breeding and wintering seasons derived from the fourth-corner analysis with a LASSO penalty. White cells represent non-significant associations whereas darker colours represent stronger associations, with blue representing positive correlations (i.e. traits associated with highly impervious areas) and red negative ones (i.e. traits associated with green spaces). ADN is the association degree during nesting whilst AOBS is the association degree outside the breeding season. See Table 1 for further details on traits.

Full size image

Discussion

Our hypotheses were largely supported by our results. We predicted that bird species composition would change along an urbanisation gradient across seasons and that these changes would be modulated by ecological and life-history traits, with winners and losers distinguishable in the urban environment. Our study confirmed that bird communities vary along the urban gradient in both the breeding and wintering seasons, and that species traits significantly influence these responses. However, rather than a clear-cut division into winners and losers, most species appeared to be urban adapters, typically occurring at intermediate levels of urbanisation. Moreover, we observed notable intra-specific differences between the breeding and wintering seasons, suggesting that individual species may shift their responses to urbanisation throughout the annual cycle. Together, these findings underscore the complexity of urban bird community dynamics and highlight the importance of accounting for seasonal variations..

We found that being a colonial species and having a higher number of broods per year were the traits most closely associated with winners31,70, whilst solitary breeders tended to be filtered out from highly urbanised areas. Colonialism and sociality may facilitate the exploitation of urban areas through improved detection of food sources and avoidance of predators34,71. Furthermore, a higher number of broods during an extended breeding season was a winning trait. Species that are able to exploit the trophic resources in urban areas and are favoured by higher temperatures are more likely to show extended breeding seasons (e.g. 3 broods for the Italian sparrow and up to 5 for the feral pigeon). This in turn is associated with an advancement in reproductive phenology of urban in comparison to rural birds72,73. Somewhat anomalously, in winter, species with large clutch sizes were less associated with urban areas. This result likely arose because many small passerines (with relatively large clutch sizes) are migrant species that leave the city in winter (see below), whereas species with small clutches (in particular pigeons) remain.

Species nesting close to or on the ground were negatively affected by urbanisation. The loss of large open habitats within cities, together with higher nest predation pressure caused by corvids, cats and dogs and disturbance by human activities, is the main cause of the poor survival of urban ground nesting species74,75. Nesting in vegetation above the ground (either in open or enclosed nests) is beneficial for urban-tolerant species47,74, with documented switches in nest sites in some species in newly colonised urban habitats76. We found a positive effect of closed arboreal nests (e.g. magpie, monk parakeet), which is possibly a preferred strategy in comparison to cavity nesting due to the lower availability of natural cavities (i.e. mostly in trees) in highly impervious areas77, although some species can adapt to nest in non-natural cavities in buildings (e.g. jackdaw, Italian sparrow, pigeon).

Food availability can change drastically along the urbanisation gradient and especially in different seasons. Insectivorous birds were generally classed as losers because they were associated with areas with a great cover of green spaces, likely due to the difficulties in finding adequate food resources where impervious cover is high. However, some species manage to adapt to these conditions and to exploit new food sources provided by human activities, e.g. garbage, bird feeders46. Flexibility in diet, and the ability to learn and adapt, which can be correlated with a bigger brain size and a longer lifespan in some bird species, are successful features in an urban context35,78,79 because individuals might respond promptly to new ecological challenges80. Indeed, species that were more generalist in their foraging strata (i.e. exploiting all available strata in the urban environment) were more abundant in urbanised areas in winter, suggesting greater flexibility in foraging habitat selection when trophic resources become scarcer. We also found that a granivorous diet was associated with more urbanised areas34,47,81 during the breeding season. Some studies argue that this might be due to benefits from supplementary feeding82,83. However, it would seem more likely that such effects might be more evident in winter50 when natural food resources are scarcer and when supplementary feeding is more widespread84 yet there was no effect of granivorous diet in the winter in our study. Furthermore, supplementary feeding in Italy (and southern Europe in general) is not as popular as in northern European countries or the USA55,56. The positive effect of granivorous diet in the breeding season may instead have been influenced by the abundance of seed-producing ornamental plants85,86 which are widespread in the city. It is still possible to find intact native herbaceous patches in Italian cities, widely used by granivorous species breeding in highly urbanised areas such as the European serin Serinus serinus. Also, many granivorous species, such as the Italian sparrow and the feral pigeon, have developed a broad trophic niche to exploit a wide range of artificial food, which is abundant in cities.

Negative effects of non-native bird species on native urban bird communities are still not fully supported87, yet studies have shown potential competition for nest sites88,89 and effects on foraging behaviour of native avifauna90. There is also evidence that native bird communities are more affected by habitat loss and disturbance in human-modified environments than by competition with non-native species such as parakeets, which may have limited negative effects on native avifauna86. The two non-native species of parakeets we found in our study area showed different traits which resulted in different responses to urbanisation. The monk parakeet is a colonial breeder that nests in the canopy of highly urbanised areas and has a higher number of broods per year, and can thus be considered a winner. The rose-ringed parakeet on the other hand can be considered a loser as it is more tied to green spaces of cities, in particular large parks, and is a solitary breeder, nesting in cavities. This species does, however, show a strong link to human-modified habitats in its non-native range86, and it is probably less likely to occur in natural or near-natural habitats which were not considered in our study.

Interesting patterns emerged from species-specific models in relation to urbanisation, showing that species were more generalist in the non-breeding season. Regarding migratory status and phenology, some short-distance migrant species that used highly urbanised areas during the breeding period were no longer present during the winter, moving from highly urbanised areas where they nest to winter in areas where trophic resources are likely less limiting (e.g. finches like European goldfinch Carudelis carduelis, European greenfinch and European serin, common woodpigeon Columba palumbus and European starling). Some species did not breed in the most impervious areas of the city or were scarce breeders in at least some of our cities, but were present in winter (e.g. European robin, black redstart, common chiffchaff). Callaghan et al.43 and Sorte et al.91 highlighted a greater use of urban areas during the non-breeding season by migrants in cities (i.e. urban tolerance increases during the winter). Indeed, many bird species associated with green spaces (e.g. common firecrest Regulus ignicapilla, common chaffinch Fringilla coelebs, European robin, Eurasian blackbird) were classified as short-distance migrants and were negatively affected by urbanisation in our study.

This study has identified winners, losers and urban adapters in the context of an urbanisation gradient across six southern European cities during the breeding and non-breeding seasons, thus addressing the bias towards northern European cities, evident from the literature44 and by considering seasonal effects, which are rarely considered in urban gradient studies that are usually undertaken in the breeding season40,41. Many of the traits that we identified as being associated with either winners or losers were consistent with findings of other studies from different geographical locations (e.g. generalists33 and more social species43 were winners, while short-distance migrants75 and ground-nesters33,75 were losers). However, there were also some notable differences (e.g. no advantage for granivores in the winter33 or for species with a large clutch size31,35). Interestingly, urban adapters, while not classified strictly as winners, showed the ability to exploit urban environments. Their intermediate level of tolerance to urbanisation might allow them to utilise resources across a variety of urban gradients, adapting behaviours and foraging strategies that allow them to persist in cities. In this sense, urban adapters could be considered secondary winners as their flexible strategies provide them with resilience in increasingly human-dominated landscapes.

Species classified in our sample of cities may not be considered winners, losers or urban adapters in other regions. This could be due to methodological differences, although it is very likely that region-specific factors have an important influence. A particular aspect that is likely to be important is the relative scarcity of supplementary feeding in Italy compared to the majority of similar studies that show a marked bias towards northern Europe and North America where this practice is more widespread55,56. Studies in countries such as Italy therefore offer the opportunity to assess impacts of urbanisation without the likely profound influence of supplementary feeding which can extend to non-urban areas92. Broadening the seasonal and geographic scope of urbanisation impact research, ideally using standardised methods, would increase our understanding of the traits that promote winners and losers in different contexts. This could inform urban planning and management strategies to maintain high-levels of avian diversity in urban environments33 and urban biodiversity more generally.