Taxonomic and functional homogenization of butterfly communities along an urban gradient

Urbanization is a major cause of global insect declines, yet some species can persist, and even thrive, in cities. Research on butterflies frequently report reduced diversity in urban habitats compared to rural ones, but less is known about whether urbanization favours butterflies with specific functional traits. Further, few studies have evaluated whether urbanization leads to the biotic homogenization of butterfly communities, despite being frequently reported for other taxa. Here, we investigate how butterfly community composition changes along an urban gradient by surveying 44 sites around Montréal, Quebec, Canada. We test the hypothesis that urban butterfly communities are homogenized at the taxonomic and functional levels. We found clear differences in the structure of urban versus rural butterfly communities with urban areas favouring a few, highly abundant, non‐native species. These shifts were defined by losses of native species richness and abundance combined with increases in non‐native species abundance. For most of the butterfly community, species with longer flight periods were more common in urban areas. Finally, urban communities were homogenized at the taxonomic and functional levels as demonstrated by reductions in beta diversity and variation in several key functional traits (wingspan, larval diet breadth, oviposition style) in urban habitats compared to rural ones. Overall, urban habitats in this region support less diverse butterfly communities that are homogenized. Since urban areas are growing worldwide, a better understanding of how biotic homogenization arises and what its consequences are will be key to guiding future conservation efforts.


INTRODUCTION
Urbanization is one of the major causes of global insect declines (Sánchez-Bayo & Wyckhuys, 2019).This extreme form of land use change reduces the total amount of habitat in an area and impacts the size and quality of remaining habitat patches through pollution, introductions of non-native species and the urban heat effect (Fenoglio et al., 2021).At the same time, it is being increasingly recognized that green spaces in cities have the potential to support diverse insect communities, despite reductions in habitat quality (e.g., vacant lots, stormwater ponds; Gardiner et al., 2013, Perron & Pick, 2020).For example, 50% of Bangledesh's butterfly biodiversity can be found in green spaces in the city of Dhaka (Chowdhury et al., 2021).With more than 50% of the global human population living in cities and further increases predicted to continue over the next several decades (United Nations, 2019), it is important that we understand how insects are influenced by urbanization so that we can better inform policy, management and the design of urban green spaces.
Butterflies, a well-studied group of insects, are often used as indicator species to represent the responses of other insects to environmental change (Thomas, 2005).Yet, research conducted over the last several decades reveals there is variation in how butterflies respond to urbanization.Most research on butterfly community structure shows declines in richness and abundance with increasing urbanization (reviewed in Ramírez-Restrepo & MacGregor-Fors, 2017).However, some studies have shown that butterfly communities can peak in diversity in areas of moderate urbanization (e.g., suburbs; Blair & Launer, 1997, Hogsden & Hutchinson, 2004) and that some native butterfly species can reach very high densities in cities (Ramírez-Restrepo & Halffter, 2013;Shapiro, 2002).These results are in line with the broader literature on terrestrial arthropods, which mainly shows declines in richness and abundance with increasing urbanization (Fenoglio et al., 2020;Vaz et al., 2023) and that only certain taxa show increases (e.g., hemipterans; Raupp et al., 2010).It remains unclear why certain butterflies can persist, and even thrive, in urban areas.
The variation in butterfly species' responses to urbanization is likely mediated, at least in part, by their functional attributes, or traits (Table 1).For instance, some studies have found that butterflies with traits associated with a stronger dispersal ability (e.g., larger wingspans; Sekar, 2012) are more common in urban areas (Kuussaari et al., 2021;Merckx et al., 2018;Merckx & Van Dyck, 2019).This could be because stronger dispersal helps organisms travel between habitat patches when connectivity is low (Cote et al., 2017), as is typically the case in cities.However, some studies have found support for the opposite trend (Lizée et al., 2011).Traits describing butterflies' diet breadth during the larval stage may also play a role.Some studies report that species whose larvae are diet generalists are more common in urban areas than specialists since they are more likely to be able to exploit common urban plant resources (e.g., non-native plants; Blair & Launer, 1997, Hogsden & Hutchinson, 2004, Niell et al., 2007, Lizée et al., 2011).However, other studies have found no relationship between larval diet and urbanization (Merckx & Van Dyck, 2019).Similar patterns have been found within other insect groups such as increases in generalist bees (Buchholz et al., 2020) or more mobile orthopterans, moths and beetles (Merckx et al., 2018;Piano et al., 2017) in urban areas.
While many studies have documented declines in butterfly diversity in response to increasing urbanization, surprisingly few have explicitly evaluated whether this also leads to biotic homogenization of butterfly communities (reviewed in Ramírez-Restrepo & MacGregor-Fors, 2017).Biotic homogenization is a process where communities increase in similarity over time or space, typically due to species invasions and/or native species extirpations (sensu McKinney & Lockwood, 1999).Biotic homogenization can have consequences for entire ecosystems like disrupting the provisioning of ecosystem services and reducing ecosystem resilience (Olden et al., 2004).It can occur at the taxonomic level as a decrease in β-diversity and at the functional level if the losses and gains of species are related to the roles or traits of those species (Olden & Rooney, 2006).The result is communities that are highly redundant in terms of species identities and/or individuals with the same roles or traits (Olden & Rooney, 2006).Biotic homogenization is frequently reported as an outcome of urbanization for plants and birds (Marcacci et al., 2021;McKinney, 2006), but comparable research for insects is still lagging behind (Lokatis & Jeschke, 2022).Piano et al. (2019) found evidence for taxonomic homogenization for multiple insect groups in the city of Belgium, but functional homogenization was not investigated.Kuussaari et al. (2021) inferred homogenization of urban butterfly communities from observed declines of less mobile species and habitat specialists but did not explicitly test for it.To our knowledge, only one previous study has explicitly evaluated both taxonomic and functional homogenization in urban butterfly communities and they detected increases in community similarity at the functional, but not taxonomic level (Merckx & Van Dyck, 2019).
Here, we investigate how the composition of butterfly communities around Montréal, Quebec, Canada, changes across an urban gradient by conducting field surveys of adult butterflies and quantifying the amount of urban cover in the landscape surrounding butterfly habitats.Our first objective is to determine whether butterfly richness, abundance and evenness (i.e., components of α-diversity) are impacted by urban cover.We predict that they will be negatively affected, in accordance with past research.Our second objective is to test for trends in the dominance of trait values (hereafter 'functional dominance') for six key traits (Table 1) by evaluating patterns in community-weighted means for each trait.We consider these traits 'functional' in the sense that they impact fitness by influencing a species' response to environmental change, rather than a species' role in ecosystem function per se (functional effect vs. functional response traits; Wong et al., 2019).We predict that butterflies with larger wingspans, increased voltinism, longer flight periods and those that are larval diet generalists, overwinter as adults and oviposit in clusters will be more dominant in urban areas (Table 1).Our third objective is to identify taxonomic and functional homogenization by testing for declines in β-diversity (taxonomic) and community-weighted standard deviations for the same six traits (functional level) across the urban gradient.We predict that butterfly communities in urban areas will be both taxonomically and functionally homogenized.

Study system
Field surveys of butterfly communities were conducted in 44 grassland or savanna sites situated within a 7300 km 2 study area around the city of Montréal in Quebec, Canada (Figure 1).
The region is comprised of mainly urban and agricultural land, but also includes remnant patches of mixed wood forests (deciduous and coniferous) and a system of waterways including the St. Lawrence River (Wiken, 1986).Montréal is the second largest city in Canada with over 4 million inhabitants (Statistics Canada, 2017).Most of the population lives on Montréal Island, but the surrounding municipalities (i.e., suburbs) are also extensively developed and have seen the largest gains in population sizes since 1966 (Dupras et al., 2016).
T A B L E 1 Predicted patterns in functional response traits for butterflies with increasing urbanization of their habitats and the potential mechanisms at play based on past research.Field sites were grasslands or savannas that were chosen to span a gradient from more rural (e.g., old fields, forest clearings) to more urban areas (e.g., city parks, vacant lots).Thus, none were completely unaffected by human activities (e.g., nature preserves).Sites were selected by visually assessing local site features (e.g., evidence of human presence) during a springtime visit and surrounding land cover using satellite imagery.Site area was measured post-surveys using satellite imagery and drawing polygons around continuous tracts of habitat.Sites were located at least 7 km apart and were visited three to four times from May to August in 2017 (n = 39 sites) or 2018 (n = 5 sites).The five additional sites surveyed in 2018 were selected to increase the number of urban sites in our sample since in 2017 we surveyed relatively more sites (69%; 27/39 of sites) that were surrounded by primarily non-urban land (at 300 m scale).

Butterfly surveys
We estimated butterfly abundance during each visit to a site using the transect walk method which is a standardized protocol used in many butterfly monitoring schemes around the world (van Swaay et al., 2008).Two observers walked a spatially fixed 300 m long route, divided into five separate transects of 60 m each that were arranged side by side and 15 m apart.Transects were walked at a consistent speed of 10 m/min, and all butterflies observed within 5 m in front of or to either side of the observers were identified and counted by sight.This ensured the search area and time were constant, despite sites varying in their total area.Following standard protocols, we only conducted surveys when butterflies were fully active: 9 am to 4 pm, temperatures of 13-40 C or >17 C when cloud cover exceeded 75%, wind below 30 km/h, and no rain.For the analysis, butterfly data were combined across the three to four visits for each site.

Landscape data
To quantify the amount of urban land in the surrounding landscape of each site, we used a 30 m resolution 2017 land cover map (Agriculture and Agri-Food Canada, 2017; Figure 1).Urban land was defined as built-up or developed land used for human infrastructure like roads, buildings and industrial sites.We calculated the proportional cover of urban land in circular buffers of varying radii (range = 100-5000 m, at 100 m intervals) surrounding each site.The proportion of urban land surrounding sites varied at different scales (e.g., 0%-100% at 300 m scale, whereas 0%-85% at 2600 m scale).

Trait data
To evaluate the role of traits in mediating species' responses to urbanization, we collected data on six key traits (Table 1) for the 36 species we observed from several North American and Canadian field guides (Layberry et al., 1998, Douglas & Douglas, 2005, Wagner, 2005, Hall et al., 2014; Table S1).When it was available, we recorded data that was specific to our study region (i.e., mixed wood ecozone of southeastern Canada).For the European Common Blue (Polyommatus icarus), which was recently introduced to Canada from Europe, we used data from both the native (Nygren et al., 2008;Thomas & Lewington, 1991) and introduced ranges (Rivest & Kharouba, 2021).
Data for oviposition style were more difficult to obtain, so we consulted two additional online databases (BAMONA, 2023;Mass Audubon, 2023) for this trait.
Wingspan was calculated as the median of the range reported.
Voltinism was measured as the average number of generations per year across the species' Canadian or North American range.Flight period length was calculated as the number of months when adult butterflies could be observed in the study region.Larval diet breadth Map of the study region in Northeastern North America (left map) with a red square showing the extent of the study area (right map) in Montréal where the survey sites were located (n = 44 sites; red dots).Map of Montréal is overlaid on a 30 m resolution land cover map (Agriculture and Agri-Food Canada, 2017) showing the extent of urban (dark grey), non-urban (white) and water (light grey).Maps were created using R 4.2.3 (R Core Team, 2023).
was scored from one to five based on whether larvae were known to feed on (1) one plant species, (2) one plant genus, (3) one plant family, (4) one plant order or (5) multiple plant orders.Overwintering stage was scored from one to four based on whether species were known to enter diapause as (1) an egg, (2) larva, (3) pupa or (4) adult and excluded species that do not spend the winter in Canada (i.e., seasonal migrants; n = 4 species).Oviposition style was scored as 0 or 1 based on whether females were known to lay eggs (0) singly or (1) in clusters.For each butterfly species, we calculated the average value for each trait across all sources (Table S1).For butterflies that could only be identified to genus in the field (n = 3 genera), this calculation was done across all potential species (e.g., for Colias spp., we averaged across C. philodice, C. interior and C. eurytheme).

Overall approach
We structured the analysis into three sections and fit a series of regression models in each.Specifically, we analysed the effects of the proportional cover of urban land surrounding sites on: (1) the components of α-diversity (richness, abundance and evenness), (2) the functional dominance of six traits (Table 1) and (3) variables describing biotic homogenization at two levels: taxonomic (β-diversity) and functional (community-weighted standard deviations for the same six traits).
Model fitting followed the same approach in all three sections.
Where data met the assumptions, we used linear models (LM); otherwise, we used generalized linear models (GLM).We compared model fits using Akaike Information Criterion (AIC) or, for models with quasilikelihoods (e.g., quasibinomial), F-tests (Burnham & Anderson, 1998;Zuur et al., 2009).We tested for effects of year and site area and included these variables as covariates when they improved model fit (ΔAIC>2 or p < 0.05 from an F-test).When included, we evaluated collinearity between all predictor variables using variance inflation factors (i.e., correlated if VIF >5) from the package car (Fox & Weisberg, 2019).We also explored non-linear effects of urban land by visually checking plots and comparing model fits.Final models were validated by visually assessing residual plots.For GLMs, residual plots were produced using the package DHARMa which uses a simulation-based approach to produce scaled residuals (between 0 and 1) that are readily interpretable for generalized models (Hartig, 2018).Spatial autocorrelation was evaluated using Moran's I tests.To evaluate the contribution of each covariate to model fit, we used likelihood ratio tests (type II ANOVA) or F-tests.All statistical analyses were performed using R 4.2.3 (R Core Team, 2023).
Spatial scale is central to determining associations between species and their environment (Brennan et al., 2002;Levin, 1992), so we used a multi-scale approach in all three analysis sections.We fit models at scales between 100 and 5000 m at 100 m intervals and determined the scales at which the species-landscape relationships were strongest (i.e., 'scale of effect'; Martin & Fahrig, 2012).We selected the best scale for each group of response variables by maximizing model fit (i.e., lowest AIC or, for models with quasi-likelihoods, highest R 2 ; Figure S1).Groups here refer to the same type of response variables, for example, the best scale for total richness was also used for analyses of native richness and non-native richness.
Likewise, the best scales for trait means were also used for analyses of trait standard deviations.The relationship between mean larval diet breadth and proportion of urban land was not significant at any spatial scale, so we used 400 m because it was the average scale for the other trait means (range = 200-700 m), and our results for this trait were not sensitive to differences in scale (Table S2).

Components of α-diversity
To evaluate the effects of urban land on butterfly community structure, we calculated richness and abundance for the entire community, native species only and non-native species only for each site (n = 44 sites).We also calculated Shannon's evenness (J'), or Pielou's J, for each site using the package vegan (Oksanen et al., 2019).This index ranges from 0 to 1, with 1 indicating complete evenness (i.e., abundance evenly distributed amongst species; Pielou, 1966).We analysed changes in richness (discrete; >0) using GLMs with the Poisson probability distribution from the package MASS (Venables & Ripley, 2002).Abundance (discrete count; >0) was overdispersed (mean: variance = 1:48), so we used GLMs with the negative binomial probability distribution.For richness and abundance models, the number of visits to each site was included as a weighting factor using an offset term equal to the log of the number of visits.Finally, we analysed changes in evenness (continuous; [0,1]) using GLMs with the beta probability distribution from the package mgcv (Wood, 2011).Since evenness sometimes took on the value of one (n = 2 of 44) but the beta distribution assumes values in the open interval (0,1), we rescaled values using the transformation: y' = (y*(n À 1) + 0.5)/n, where n is the sample size (Smithson & Verkuilen, 2006).

Functional dominance of butterfly traits
To evaluate the effects of urban land on the dominance of the six traits, we calculated community-weighted means for each trait based on species-specific abundances at each site (n = 44 sites).These weighted means are commonly used to indicate the overall dominance of trait values within a community (e.g., Roscher et al., 2012).We analysed changes in trait means (continuous) using LMs for all traits except oviposition style (discrete proportion) which we analysed using GLMs with the quasibinomial probability distribution.This variable represented the proportion of the abundance in each site that was made up of individuals that laid eggs in clusters (0 = singly, 1 = clusters) and was underdispersed (mean: variance = 1:0.12).Since non-native P. icarus was by far the most common and abundant species observed during our surveys (see Results), we repeated these functional dominance analyses without P. icarus to evaluate whether trends were being driven by just this species.

Biotic homogenization
To evaluate whether butterfly communities were taxonomically homogenized in urban areas relative to rural areas, we first divided sites into two groups according to the proportion of urban land surrounding sites at the 300 m spatial scale: urban (≥50% urban, n = 16 sites) and rural (<50% urban, n = 28 sites).We chose 300 m since this was the best spatial scale for models of butterfly abundance (Data S1).Within each group, we calculated Bray-Curtis dissimilarity indices (i.e., β-diversity) for all pairwise site comparisons (n = 498 comparisons) based on species abundances (Bray & Curtis, 1957) using the package vegan (Oksanen et al., 2019).This index ranges from 0 to 1 and represents the proportion of the abundance that is different between two communities (i.e., low β-diversity, high homogenization).Bray-Curtis is more robust to sampling errors than other measures of β-diversity (e.g., Sorenson, Simpson) when abundance data are available (Schroeder & Jenkins, 2018).We analysed changes in β-diversity (continuous; [0,1]) based on site grouping (urban-urban vs. rural-rural comparisons) using GLMs with the beta probability distribution from the package mgcv (Wood, 2011).We included distance between sites as a covariate to account for increasing taxonomic similarity with increasing proximity.We also fit this model including two crossed random effects representing the first and second sites included in the site pairing, but our results did not change so we excluded the random terms (Table S3).
To evaluate whether butterfly communities were functionally homogenized in urban areas relative to rural areas, we calculated community-weighted standard deviations for each functional trait based on species-specific abundances at each site (n = 44 sites).
These weighted measures can be used to indicate the overall amount of variation in trait values within a community (Merckx & Van Dyck, 2019), thus, lower values would be indicative of greater functional homogenization.We analysed changes in trait standard deviations (continuous) using LMs.We tested for effects of species richness in all models since more diverse communities could have higher amounts of trait variation.Spatial autocorrelation was detected for the wingspan model, so we included a spatial dependence correlation structure (Gaussian) within the model using the package nlme (Pinheiro et al., 2023).

RESULTS
We counted 2794 butterflies from 36 species across the 2 years of surveys (Figure S2).Per site, we observed an average of 63.5 individuals (range = 5-202) from 9.5 species (range = 4-17).Only three species in this region are non-native, but combined they accounted for 59% (1642/2794) of the total abundance (Figure S2).The most common and abundant species was non-native P. icarus, which we counted 1128 times in 80% (35/44) of sites.

Components of α-diversity
Urban sites supported a few, highly abundant butterfly species that were typically non-native (Table 2; Figure 2).First, richness decreased with the proportion of urban land (β = À0.70 (0.21SE), χ 2 = 11.85,p < 0.001, df = 1; Figure 2a), declining by 50% across the entire urban gradient.This translated to a loss of approximately six species per site.
Second, abundance did not change across the urban gradient (β = 0.46 (0.31SE), χ 2 = 2.52, p = 0.11, df = 1; Figure 2d).This was because the abundance of native species decreased by 57% across the urban gradient (β = À0.87 (0.38SE), χ 2 = 5.59, p = 0.018, df = 1; Figure 2e) while the abundance of non-native species increased by 336% (β = 1.47 (0.36SE), χ 2 = 19.74,p < 0.001, df = 1; Figure 2f).This translated to a loss of 17 native individuals and a gain of 46 non-native individuals per site.Increases in nonnative abundance across the urban gradient were attributable to gains in just one species, P. icarus, since there was no trend in nonnative abundance when this species was excluded (Figure S3).Thus, P. icarus was very abundant in more urban sites.The survey year also impacted abundance ( p < 0.001 all models; Table 2) with more individuals counted per site in 2018 compared to 2017 (mean difference in total abundance at 50% urban land = 120 individuals).
F I G U R E 2 Effects of urban land on butterfly community structure showing relationships with (a-c) richness, (d-f) abundance and (g) evenness of all species, native species only and non-native species only using GLMs (n = 44 sites).Shown are raw data (dots), model predictions (orange lines), 95% confidence interval (grey ribbons) and p-values from likelihood ratio tests.The year of survey had a significant effect on abundance (d-f), so model predictions are shown for 2017 and raw data are coloured by year (2017 = black; 2018 = white).

DISCUSSION
Urbanization-driven biotic homogenization has been well documented for plants and birds but is understudied for insects like butterflies.
Here, we demonstrate that urbanization in Montréal, Canada, has led to the biotic homogenization of butterfly communities.We show that there were clear differences in the structure of urban versus rural butterfly communities: a few, highly abundant, non-native species were characteristic of urban areas.These shifts were defined by losses of native species richness and abundance combined with increases in non-native species abundance.We also found that butterflies with longer flight periods were more common in urban areas.Lastly, we show that increasing urban land surrounding butterfly habitats was linked to taxonomic homogenization and reduced variation in butterfly wingspan, larval diet breadth and oviposition style, demonstrating functional homogenization.

Components of α-diversity
Increasing the proportion of urban land surrounding butterfly habitats had a negative effect on native butterfly richness and abundance.This generally (Fenoglio et al., 2020;Vaz et al., 2023).Our result is not surprising as urban areas are associated with multiple factors that could negatively impact native insects such as lower habitat amount, connectedness and quality (Fenoglio et al., 2021).Critically, future research should focus on determining the relative importance of different urban drivers (e.g., temperature, pollution, connectivity), so that managers can better mitigate threats to native butterflies by prioritizing which ones to focus on.
In contrast, increasing the proportion of urban land had a positive effect on the abundance of one non-native species: P. icarus (see Results).This was expected given previous work showing that P. icarus populations in this region are concentrated in urban areas where its main host plants (i.e., non-native Fabaceae) are commonly available (Dexheimer & Despland, 2023;Rivest & Kharouba, 2021).This result is also consistent with past work that shows some non-native species can reach very high abundances in cities for a variety of reasons

Functional dominance of butterfly traits
Apart from P. icarus, the communities in urban areas were made up of a greater proportion of butterflies with longer flight periods.This pattern has also been observed in butterfly species occurrence records from across Europe (Callaghan et al., 2021).One hypothesis for this pattern is that butterflies with longer flight periods are exposed to a wider range of climatic conditions over their lifespan which could translate to a greater tolerance overall for the varied environmental conditions found in cities (Callaghan et al., 2021).For example, temperature can vary over relatively short distances in cities, resulting in both cold and hot spots despite being warmer overall than surrounding rural areas (McGlynn et al., 2019).Our result is also consistent with reports of lengthened flight periods due to local adaptation in butterflies in urban environments compared to rural ones (Dennis et al., 2017;Merckx et al., 2021).This is thought to be related to the warmer temperatures and longer growing seasons caused by the urban heat island effect, potentially allowing butterflies to emerge earlier, develop faster and/or produce additional generations towards the end of the season (Altermatt, 2010;Dennis et al., 2017;Merckx et al., 2021).As urban areas continue to expand and as the rate of climate change increases, these patterns are likely to become even more common and thus require additional scrutiny.
Other than flight period length, we found no other trends in functional dominance across the urban gradient after excluding P. icarus.This is inconsistent with past studies on butterflies (Table 1) and for insects in general (Buchholz et al., 2020;Merckx et al., 2018;Piano et al., 2017).However, our results do align with research that found no trends across urban gradients in butterfly larval diet breadth (Merckx & Van Dyck, 2019) and oviposition style (Callaghan et al., 2021).The lack of trends we found here could be related to the low interspecific variation in some of the traits we examined.For instance, most species in this community (47%; 17/36) show moderate diet generalism, meaning that their larvae feed from multiple genera of plants from a single family (level 3) whereas only one species was an extreme diet specialist (level 1; Table S1).It is possible that most butterflies that are extreme specialists are already absent from this landscape since this region has a long history of human activities in addition to urbanization (e.g., agriculture; Statistics Canada, 2014).
The way we measured traits may have also influenced our ability to detect trends.We obtained trait values by averaging across a butterfly species' entire Canadian range and annual cycle (i.e., species level).This may be less informative than data collected at the individual level given that local adaptation to urbanization has been demonstrated (Dennis et al., 2017, Merckx et al., 2021).That said, previous studies that found trends in butterfly traits across urban gradients used an approach similar to ours (Kuussaari et al., 2021;Merckx & Van Dyck, 2019).Given the challenges and effort in collecting ontogenetic trait data for all individuals in a population, it will be critical to determine when and where this approach is needed.For example, this approach may be needed for species assemblages that are not as diverse, since trait analyses for such assemblages are known to be less robust to the presence of intraspecific variation (Gentile et al., 2021).
If the goal is to detect interspecific trait variation, it may be helpful to include surveys of a greater variety of non-urban sites (e.g., agricultural, wetland, forest) to help increase overall insect diversity and thus the robustness of future trait analyses.

Biotic homogenization
Urbanization led to the taxonomic and functional homogenization of butterfly communities in Montréal, Quebec.To our knowledge, only one previous study has evaluated both taxonomic and functional homogenization in urban butterfly communities (Merckx & Van Dyck, 2019), making comparisons with the literature difficult.Merckx and Van Dyck (2019) concluded that functional homogenization was occurring, but not taxonomic.Our findings are consistent with studies on the urban-driven taxonomic homogenization of insects in general (e.g., leafhoppers, beetles, spiders; Knop, 2016, Piano et al., 2019), the taxonomic and functional homogenization of moth communities (Merckx & Van Dyck, 2019) and the taxonomic homogenization of butterfly and moth communities across other types of anthropogenic gradients (e.g., management intensity in agricultural systems; Ekroos et al., 2010).Our findings add to a growing body of literature showing support for the urbanization-driven biotic homogenization hypothesis across multiple taxa and regions.
We found reduced trait variation in urban habitats for half of the traits we examined (wingspan, larval diet breadth and oviposition style) indicating functional homogenization.Merckx and Van Dyck (2019) also detected homogenization for their measure of larval diet specificity, but not for wingspan (oviposition style was not examined).
In our study, functional homogenization likely occurred because the more urban sites were species poor and dominated by P. icarus, resulting in lower trait variation in urban areas.For instance, butterflies with both very small (<30 mm) and very large (>40 mm) wingspans (full range 28-53 mm) were absent in urban sites whereas there was greater variation in wingspan in more rural areas (Figure 3a).This is consistent with past work showing that biotic homogenization typically results from species invasions and/or native species extirpations (McKinney & Lockwood, 1999).
An important direction for future research will be to evaluate whether functional homogenization, like we have reported here, could have higher order implications for ecosystem functions and services.
Past research suggests that homogenized communities can have reduced ecosystem functionality because they have fewer species with distinct functions or roles (Cadotte et al., 2011;Olden et al., 2004;van der Plas et al., 2016).For butterflies, typical ecosystem functions include pollination and nutrient cycling which, if homogenized, could have impacts on species that interact with butterflies (e.g., predators) or species occupying the same habitats (e.g., plants).For example, functional homogenization in butterfly wingspan could result in communities that include only large butterflies or only small butterflies.This could have cascading impacts on predators like birds which feed on butterflies of different sizes, thus impacting ecosystem functions like nutrient cycling.Future studies should quantify ecosystem functions provided by butterflies in urban compared to rural environments to better understand the consequences of functional homogenization.

CONCLUSIONS
Our findings indicate that the preservation and restoration of native butterfly communities should be the focus of conservation efforts in cities. Continued urban growth as it is currently managed will lead to further reductions in the amount and quality of butterfly habitat, potentially contributing to future declines and/or extirpations of native butterflies.However, it remains difficult to make specific recommendations about best practices for managing urban butterfly communities (e.g., design of green spaces) without a better understanding of the mechanisms at play.Future studies should identify what specific urban characteristics (e.g., floral diversity; Gordon & Kerr, 2022) have the greatest benefits for butterflies so habitats can be improved and threats mitigated.Overall, our results highlight the need for more research aimed at making cities more habitable places for native species to live.

T A B L E 2
Results of final models analysing the effects of proportion of urban land on (a) the components of α-diversity (richness, abundance, evenness), (b) functional dominance for six traits (means) and (c) measures of taxonomic (β-diversity) and functional (standard deviations (SD)) homogenization using LMs and GLMs (n = 44 sites).Shown are spatial scales (m), estimates, standard errors (SE), degrees of freedom (df), as well as test statistics (χ 2 or F-value) and their corresponding p-values.'SAC' indicates spatial autocorrelation was accounted for in the model.Significant results are in bold.
finding is in agreement with past research on butterflies (Ramírez-Restrepo & MacGregor-Fors, 2017) and terrestrial arthropods more F I G U R E 3 Effects of urban land on the dominance of functional butterfly traits showing relationships with community-weighted means for (a,b) wingspan, (c,d) voltinism, (e-f) flight period length, (g,h) larval diet breadth, (i,j) overwintering stage and (k,l) oviposition style for the entire butterfly community as well as when P. icarus is excluded using LMs and GLMs (n = 44 sites).Shown are raw data (dots), model predictions (black lines), 95% confidence interval (grey ribbons) and p-values from (a-h) likelihood ratio tests or (k,l) F-tests.

F
I G U R E 4 Biotic homogenization of butterfly communities in response to urbanization showing relationships between (a) β-diversity (Bray-Curtis dissimilarity index) and site grouping (urban vs. rural), and (b-g) trait standard deviations (SD) and urban land using LMs and GLMs ([a] n = 498 comparisons; [b-g] n = 44 sites).Shown are raw data (dots), model predictions ([a] white bars; [b-d] black lines), 95% confidence intervals ([a] error bars; [b-d] grey ribbons) and p-values from likelihood ratio tests.(e.g., enhanced resource supply, lower competition; Cadotte et al., 2017).