Introduction

Habitat loss and fragmentation are among the main drivers of global biodiversity loss, contributing to the high number of species currently at risk of extinction (Betts et al. 2017). One of the consequences of habitat loss and fragmentation is the increase in edge effects, represented by changes in biotic and abiotic factors at the boundary between habitat and matrix areas (reviews in Ries et al. 2004, 2017). Edge effects may negatively or positively affect the local abundance of many species (Ries et al. 2004, 2017), in particular forest vertebrates (Banks-Leite et al. 2010; Pfeifer et al. 2017). While edge effects have a widespread impact on forest ecosystems and species globally (Lindell et al. 2007; Matuoka et al. 2020), identifying general patterns in how different species respond to forest edges has proven challenging. This challenge stems from the considerable variability in sampling designs and sample sizes across individual studies (Ries et al. 2004, 2017) and a dearth of fundamental data on species’ natural history, particularly in tropical regions (Lees et al. 2020). For these reasons, past efforts to synthesize quantitatively the existing research on edge effects have either focused on community-level response variables (i.e. species richness; Willmer et al. 2022) or on environmental explanatory factors only, such as resource availability at the forest-matrix interface (Ries et al. 2004, 2017) and the occurrence of historical disturbance in the study area (Betts et al. 2019).

There is no doubt that the environmental conditions at forest edges vary between regions (Mendes and Prevedello 2020), and that biotic communities respond differently to edge effects at different latitudes (e.g. Lindell et al. 2007; Betts et al. 2019; Weeks et al. 2023), especially due to varying levels of natural and/or anthropogenic disturbance (Betts et al. 2019; Weeks et al. 2023). These factors potentially contribute to the well-documented large variability in edge responses among species and regions (Ries et al. 2004, 2017). However, to identify general patterns on species-level responses, as well as to understand their underlying mechanisms, it is necessary to move beyond extrinsic factors alone, and determine whether and how different species traits mediate edge responses (e.g. Barbaro et al. 2014; Lopes et al. 2016; Matuoka et al. 2020). Accordingly, the number of individual studies focusing on how species with different traits are differently affected by edge effects has increased in recent years (e.g. Caitano et al. 2020; Jin et al. 2021; Zuñe-Da-Silva et al. 2022). Birds have been the most studied group in this regard (Matuoka et al. 2020), possibly due to their high abundance and richness in forests throughout the planet (e.g. Orme et al. 2005; Vale et al. 2018). Birds are also highly diverse in terms of species traits (see in Sekercioglu 2012), which can potentially lead to highly variable responses to edge effects. However, a quantitative synthesis testing how different bird species traits mediate edge responses is still lacking.

Many species traits may mediate bird responses to forest edges, in particular those related to diet, mobility and habitat preference. Diet is one of the best studied attributes in birds, but the literature on edge effects has shown contradictory results. Some studies show that the richness and abundance of insectivore and frugivore species can either increase (e.g. Rodewald and Brittingham 2002; Batáry et al. 2012; Terraube et al. 2016) or decrease (e.g. Watson et al. 2004; Albrecht et al. 2013; Rodrigues et al. 2018) near forest edges. Nectarivore, plant/seedeater, and omnivore species generally respond positively to edge effects (e.g. Dale et al. 2000; Deikumah et al. 2014; Khamcha et al. 2018). Flying ability, which may be inferred from the hand-wing index (HWI) (see in Sheard et al. 2020; Weeks et al. 2023), may affect the ability of birds to move between forest patches or use the matrix for foraging, thus potentially affecting edge responses (Weeks et al. 2023). Indeed, many migratory birds respond positively to forest edges (Rodewald and Brittingham 2007; Packett and Dunning 2009; Terraube et al. 2016). Additionally, species with lower HWI exhibited a weakly negative response to forest edges in the Brazilian Atlantic Forest (Bonfim et al. 2021). Small forest birds have lower area requirements and, therefore, could present a more positive response to edges than birds with higher body mass (Menke et al. 2012; Khamcha et al. 2018; Morante-Filho et al. 2018), as edges reduce habitat availability due to geometric constraints (Prevedello et al. 2013).

Territoriality remains relatively unexplored in the literature on edge effects (refer to Ries et al. 2017; Pfeifer et al. 2017), despite its potential impact on area requirements and dispersal ability, consequently influencing edge responses (Stratford and Robbins 2005; Sheard et al. 2020). Indeed, non-territorial birds appear to be more numerous in forest edges (Rodewald and Brittingham 2004). Finally, habitat preference may also strongly mediate edge responses, as the ability to use complementary or supplementary resources from the matrix may increase species abundances at edges (Ries et al. 2004). Indeed, many studies have shown that forest-dependent birds are more vulnerable to landscape changes than species with other habitat preferences (e.g. Carvajal-Cogollo and Urbina-Cardona 2015; Matuoka et al. 2020; Oliveira-Silva et al. 2022), and that this vulnerability could be higher in tropical regions, due to lower levels of historical disturbance (Betts et al. 2019; Weeks et al. 2023).

Here, we quantify the impacts of edge effects on bird species’ abundances globally, to identify how different species traits mediate edge responses. We focus on six species traits presumed to affect edge responses, which are available for different bird species globally, namely diet, foraging stratum, body mass, HWI (hand-wing index), territoriality, and habitat preference. We also consider edge contrast and latitude as additional, extrinsic (environmental) variables, as they are known to affect community-level responses to edges at a global scale (Willmer et al. 2022). Based on the hypotheses from previous studies (see previous paragraph), we tested four predictions: (i) generalist species, as well as species from omnivore, nectarivore, and plant/seedeater diet and broad foraging stratum, respond more positively to edge effects; (ii) birds with higher mobility (higher HWI) and lower area requirements (non-territorial, lower mass) respond more positively to edges; (iii) species that prefer open and semi-open habitats respond more positively to edges than forest species; and (iv) birds in areas with higher latitude and lower edge contrast respond more positively to edges.

Materials and methods

Literature review

The global scientific literature on quantitative studies comparing bird abundance on forest edges and interior was reviewed using SCOPUS (www.scopus.com), Web of Science (www.webofknowledge.com), Scielo (https://scielo.org/) and the Periodicos CAPES platform (https://www.periodicos.capes.gov.br) in English, Spanish, and Portuguese. The search covered publications from 1970 to December 22nd, 2021. The following expressions were used: [(“avian” OR “Bird”) AND (“edge” OR “edge forest” OR “edge effect” OR “edge influence”)]. These words were searched for in the title, abstract, keywords, and references.

Studies containing only presence/absence data were discarded. The studies (59, see in Table S1) retained for analysis contained at least one of the following abundance proxies: average or total abundance, capture percentage, average capture ratio, total number of individuals, or relative density (data deposited in Zenodo repository, 10.5281/zenodo.10622352). Data from four studies were extracted manually from graphs using the WebPlotDigitizer 4.5 software (Rohatgi 2021). Unpublished original data from three studies by the authors of the current study were also used (data to be deposited in Dryad repository). Values from studies in which two sampling methods were used (e.g., nets and counting spots), or two types of areas were sampled (e.g., primary or secondary forests, or patches with high or low edge contrast), were considered separately for the analysis. Similarly, data from two or more types of edge contrast for a same study were analyzed separately (DeGraaf 1992; da Silva and Silva 2020). As the utilization of mist net sampling is an inherently selective and potentially biased technique in regard to the assortment of species captured within distinct habitats, we performed a sequence of analyses excluding data acquired through this method. Our objective was to scrutinize whether the application of mist nets exerted a discernible influence on the comparative assessment of avian communities at the forest-edge and interior. The exclusion of these data returned similar results as the full dataset (see Supplementary information S2).

“Edge” was defined as the intersection between matrix and forest. For each study, we extracted the data nearest to this intersection to represent the Edge, which could include any sampling point located between 0 and 50 m into the forest interior, depending on the study. Studies with data for both the forest and the matrix were considered, but matrix data were discarded, as our focus was the comparison between forest edge and forest interior. The “Interior” was defined as the farthest distance to the edge sampled in each study, which ranged from 75 to 1500 m across studies. Only five studies presented as “interior” data those obtained from less than 100 m from the edge; exclusion of these data returned similar results as the full dataset (see Supplementary information S3).

Response variable

The response variable in our analysis (i.e., “effect size”) was the “response ratio” (RR), calculated as the ratio between the abundance on the edge and the abundance in the forest, for a given species in a given study. RR values higher than 1 indicate higher abundance on the forest edge, while values lower than 1 indicate the opposite. When a species sampled in a given study was present only at either the edge or the interior (i.e., abundance = 0 at one of the two environments), a small constant value (e.g., 0.1) was added to edge and interior abundances for all species from the same studies. Such small constant value was defined separately for each study, corresponding to the minimum value observed for the least abundant species in the study. The addition of this value allowed calculation of RR values for species that were absent in forest interiors, as the presence of zeros in the denominator precludes the calculation of the RR. Despite RR values could be calculated when the numerator (edge abundance) was zero, we still added the constant value for the edge abundance, to keep consistency and avoid potential biases in favor of forest edges or interiors. An equal sampling effort was assumed for forest edge and interior for studies in which sampling effort was not specified. When the sampling effort differed between edge and interior, we standardized abundance values prior to the calculation of the RR, by dividing abundance by sampling effort separately for edge and interior, thus obtaining corrected abundance values that are directly comparable between edges and interiors.

Explanatory variables

The two extrinsic (environmental) variables, namely latitude and edge contrast, were extracted directly from the scientific articles reviewed. Edges were classified according to their contrast with the surrounding matrix as either: (1) soft (clearings in the middle of forest, advanced age silviculture, shrubs or immature ecological succession, narrow unpaved roads through forest), (2) intermediate (wide/paved roads through forest, natural grasslands, railroads in forest, mixed areas with unspecified edges) or (3) abrupt (agriculture, roads delimiting forest, cultivated pasture, urban or suburban areas). Despite a recent review (Willmer et al. 2022) has classified edges in two types only (“hard” or “soft”), we also included an intermediate category for a finer classification. When the study included more than one type of edge contrast, the data were treated separately in the analyses; thus, some studies provided more than one data set (DeGraaf 1992; Antos and White 2004; da Silva and Silva 2020). The potential non-independence of multiple data from a same study was controlled by using the identity of each study as a random factor (see section “Data analysis”).

Each species was classified based on the following intrinsic traits, extracted or modified from different sources: (1) diet; (2) foraging stratum; (3) body mass; (4) hand-wing index (HWI); (5) territoriality; (6) habitat preference (see Table 1 for details). In the absence of data for recently described species, the attributes were extracted from the “Birds of the World” database (Billerman et al. 2022). Bird species predominantly associated with water bodies were excluded from our analyses (e.g., water birds, shorebirds, non-forest rails), as we focused on terrestrial (forest-matrix) edge effects. All compiled data are available from repository (data deposited in https://zenodo.org/records/10622352).

Table 1 Species traits and respective categories used as explanatory variables of bird responses to edge effects. Sources: diet, foraging stratum and body mass—primarily Wilman et al. (2014), complemented with Billerman et al. (2022); HWI—Sheard et al. (2020); territoriality and habitat—(Tobias et al. 2016)

Data analysis

We used variance inflaction factor (VIF) to check the collinearity of each variable and all are low (See in Fig. S4). All quantitative response and explanatory variables (RR, body mass, HWI) were logarithmized (ln) for the analysis, for normality. We built a Linear Mixed- Effects Model (LMM) with the lme4 package (Bates et al. 2015) in R 4.1.2 (R Core Team 2021). The model included all explanatory variables as fixed factors, and the identity of each study as a random factor (Zuur et al. 2009), to control for the potential non-independence of multiple RRs from a same study. We also included two interaction terms in the model, latitude × edge contrast and latitude × habitat preference, based on the literature (Betts et al. 2019; Weeks et al. 2023). We also included the genus nested within family of each species as a separate random factor to control for phylogenetic non-independence among species (see in Bird et al. 2020). This approach allowed to consider evolutionary convergence among certain families, such as Bucerotidae/Ramphastidae or Trochilidae/Nectariniidae. This phylogenetic structure (family/genus) was more plausible than alternative structures, as inferred from a model selection approach (see Table S5). All inferences and figures regarding the fixed explanatory variables were based on the full model (containing all variables), as our focus was on understanding whether and how all species traits and extrinsic variables mediate edge responses, rather than comparing their relative importance (which would require model selection procedures). All figures were produced in R 4.2.2 (R Core Team 2022) except for the map, which was produced in QGIS software (QGIS Development Team 2022).

Results

Our review returned 2,981 interior/edge comparisons for 756 bird genera and 1,414 species, corresponding to 12.8% of the bird species on the planet (Gill et al. 2022), from 59 studies with 63 different data sets (Fig. 1). The large majority of these studies was conducted in forests below 1,000 m altitude. Only five studies (8,5%) were carried out above 1,000 m (Restrepo and Gomez 1998; Raman and Sukumar 2002; Becker et al. 2008), with a single study conducted above 3000 m (Cahill and Matthysen 2007) (data deposited in https://zenodo.org/records/10622352).

Fig. 1
figure 1

Distribution of the 63 datasets that compared responses of bird species to edges across global forests (green areas). The type of edge contrast is also shown (i.e., abrupt, intermediate, or soft edge)

The overall response ratio (RR; edge/interior abundance) across all species and studies did not differ from 0 (mean = 0.06; CI= − 0.11 to 0.23; p = 0.45). Similar overall estimates were obtained after excluding studies that sampled < 100 m from edges (mean = 0.05; CI = − 0.14 to 0.23; p = 0.61) or that used mist nets (mean = 0.06; CI = − 0.13 to 0.26; p = 0.53).

The RRs were significantly affected by diet (Table 2; Fig. 2) and the interaction between habitat preference and latitude (Table 2; Fig. 3). In addition, RRs were significantly affected by edge contrast and sampling method (Table 2; Fig. 3). Response ratios were significantly positive for grani-folivores (Fig. 2). Response ratios decreased as latitude increased for open habitat species, but slightly increased with latitude for dense and semi-open forest species (Table 3; Fig. 3). Finally, RRs tended to be mostly negative when transects were used for sampling, but positive for mist net data (Fig. 2). These results were generally robust to the exclusion of datasets that sampled less than 100 m from the edge (Supplementary material S.3) or used mist nets (Supplementary material S.5).

Table 2 Effects of species traits, edge contrast and latitude on bird response ratios (edge/interior abundance)
Fig. 2
figure 2

Forest-plot summarizing bird responses to forest edges, based on a global analysis of 2,981 responses of 1,414 bird species, obtained from 63 datasets. The vertical line at zero represents a neutral response of birds to edges (i.e., equal abundance in forest edge and interior). Positive response ratios (on the right side) represent positive responses to edges (i.e., higher abundance in edges), while negative response ratios (on the left side) represent negative responses to edges (i.e., lower abundance in edges). The significant categories/variables were: Granivores-Folivores: 0.26∗; Mist net: 0.47∗∗; Transect: − 0.30∗; Aerial 0.66∗: (∗∗∗p < 0.001; ∗∗<0.01; ∗<0.05)

Fig. 3
figure 3

Predicted effects of habitat preference interacting with latitude on bird responses to forest edges. The zero horizontal line indicates a neutral response to edges (i.e., equal abundance in forest edge and interior); positive response ratios indicate positive responses to edges (i.e. higher abundance in edges), and negative response ratios indicate negative responses to edges (i.e. lower abundance in edges)

Table 3 Model coefficients, confidence intervals (CI), and p-value of the categories of variables tested in the model. Significant variables (P ≤ 0.05) are highlighted in bold. ∗∗∗p < 0.001; ∗<0.05

Discussion

We found that two bird species’ traits, namely diet and habitat preference, the latter interacting with latitude, mediate species’ responses to forest edges globally. These findings support at least partly two of our four initial predictions (namely, predictions iii and iv), confirming and expanding the findings of previous individual studies on habitat fragmentation (Bregman et al. 2014; Keinath et al. 2017). The type of edge contrast also affected bird responses to edges, despite this effect was not as clear as the effects of the intrinsic traits. Therefore, our results show that both, species traits (in particular) and environmental factors (secondarily), are relevant to explain differences in the abundance of bird species between forest edges and interiors, and should be considered in explanatory models of edge effects. Our prediction relative to diet categories was also mostly supported, as omnivore, nectarivore, and (in particular) plant/seedeater species responded positively to edges. Nonetheless, the predictions concerning HWI, body mass, territoriality, and foraging stratum were not supported. The primary discovery from our study is that edge responses of species with distinct habitat preferences are contingent upon latitude. This interaction shows that birds inhabiting open areas benefit from edges in tropical regions, while forest-dwelling species tend to respond negatively to edges in the same regions, consistent with previous studies (Pfeifer et al. 2017; Betts et al. 2019; Weeks et al. 2023). These patterns were only detectable in our study due to the standardization of traits attributed to all avian species on a global scale. This contrasts with several previous articles that lacked a consistent classification of attributes(e.g. Restrepo and Gomez 1998; Dale et al. 2000; Laurance 2004). It is possible that this lack of standardization of traits is more common in the tropics, where biodiversity is very high, and a large research gap of species traits exists (Hortal et al. 2015), as well as a lack of basic knowledge of many species (Lees et al. 2020). Our results indicate that the use of a uniform classification of species’ traits can help in explaining the variability of responses of bird species to edges in fragmented landscapes globally.

Species traits

Forest birds are widely known to respond more negatively to edge effects, especially in the tropics (Weeks et al. 2023). As we anticipated, most open-habitat birds appear to benefit from forest edges (Kattan 1992; Thomas et al. 2014; Rodrigues et al. 2018) and habitat fragmentation in general (Keinath et al. 2017). Similar patterns appear to hold for other taxa, such as plants (Tabarelli et al. 2010; Benchimol and Peres 2015; Shumi et al. 2019). Nevertheless, forest (dense-habitat) bird species did not exhibit a significant response to edge effects, suggesting the presence of forest species that either exploit or complement their resource needs along the forest edge or in the matrix (Ries and Sisk 2004). We emphasize that there is a higher concentration of data with negative responses towards the border in tropical regions, while positive outcomes increase in subtropical areas.

Non-sensitive groups, such as plant/seedeaters, have been the focus of attention in numerous individual studies examining edge effects and human disturbances in general (e.g., Deikumah et al. 2014; Wiącek et al. 2015a; Navarro et al. 2021). Contrary to our initial prediction, vertivore species exhibited a positive response to edge effects. It is noteworthy that mist-net sampling methods may not be universally suitable for all bird groups; for example, these methods may underestimate vertivore abundance, as vertivores are usually larger. Another possible explanation is a higher detection of vertivores along forest edges (Neate-Clegg et al. 2016), where it is easier to observe relatively large birds. Another factor is the potential attraction of these species to areas where more prey or more resources are available (Ries and Sisk 2004; Packett and Dunning 2009). Furthermore mist-net sampling may not be an appropriate method to compare birds of different body masses and different vegetation structures, which may partially explain the tendency for positive RRs for mist-net data. The mesh-size used in most studies (36 mm or smaller) avoids the capture of larger birds and favor the sample of smaller ones (Pardieck and Waide 1992; Piratelli 2003). We observed a tendency (though not statistically significant) for birds with lower body mass to exhibit positive RRs (Fig S6.1). It is possible that mist-net data may have underestimated vertivores, which typically consist of larger birds. Additionally, forest edges usually have denser understory, resembling young-forest habitats, in which mist nets are more effective in detecting canopy and sub-canopy bird species (Pagen et al. 2002). Hence, the sampling at the edges may result in an oversampling of these groups compared to the forest interior. Finally, to validate these outcomes, future investigations could employ additional sample data and variables either in situ or through a review-based approach.

Species specialized in their diet (frugivores and invertivores) can respond differently to edges, and it is challenging to find a consistent pattern with the variables used. Divergent responses to the edge may be due to turnover among species with similar dietary preferences in landscapes where densities have not increased excessively (Boesing et al. 2018; Jones et al. 2021). These various responses have the potential to have a disproportionate impact on forest species within these groups. Forest edge expansion may have a disproportionate impact on forest species within these groups. Responses of invertivore species to forest edges were less positive than for other diets (see Fig. 2), probably because they are more sensitive to habitat fragmentation and/or edge effects globally (Laurance 2004; Rodrigues et al. 2018; Khamcha et al. 2018). However, most studies which found that invertivores benefit from forest edges were conducted in subtropical areas (Barbaro et al. 2014; Bereczki et al. 2015; Terraube et al. 2016). Two possible explanations may account for these positive responses: (1) The resources were habitat-specific, concentrated in certain vegetation types (Ries and Sisk 2004; Packett and Dunning 2009); and (2) Due to a higher plant resource variability in edges, as a result of vegetation structure and plant diversity, which may attract insects (Wiącek et al. 2015). Resource availability may attract some invertivore species, leading to the replacement of species that are able to tolerate forest edges (Powell et al. 2015), as occurs with many bird species in some families (e.g. Tyrannidae, Phylloscopidae, Picidae). In addition, certain invertivore species on forest edges might exhibit a more generalist behavior by consuming a variety of food items. Invertivore species typically comprise small birds (< 31 g) with diverse responses to the edge (Pfeifer et al. 2017). Frugivores responded in a similar way to invertivores, i.e., without a significant response to edges. However, in previous individual studies, many frugivore birds responded negatively to edge effects (Restrepo and Gomez 1998; Albrecht et al. 2013), especially large-bodied birds in certain families (e.g. Cotingidae, Paradisaeidae, Bucerotidae) which are highly dependent on specific resources (Kattan 1992), such as the quality of fruits. Birds with larger beaks are among the frugivores more negatively affected by edge effects (Menke et al. 2012; Saavedra et al. 2014), as well as those associated exclusively with a specific stratum in the forest (Galetti et al. 2003; Benjara et al. 2021). On the other hand, small-bodied species without a specialized diet responded positively to edge effects in previous studies (Deikumah et al. 2014; Rąkowski and Czarnocki 2019), while in our study, nectarivores responded positively, but not significantly, similarly to previous studies (Dale et al. 2000; Ikin et al. 2015). Nectarivores may, however, show different patterns throughout the year (Wardell-Johnson and Williams 2000). The absence of statistically positive responses to edge effects for nectarivores in our study may be attributed to high niche diversity, particularly in tropical regions (Sherry et al. 2020). This observation may also be influenced by various environmental, physiological, and/or phylogenetic factors.

Extrinsic (environmental) factors

Latitude significantly influences bird responses to habitat edges, playing a pivotal mediating role that (Mendes and Prevedello 2020). Additionally, latitude is associated with many ecological variables, including species richness (Willmer et al. 2022), as well as bird abundance in habitat fragmentation studies(Bregman et al. 2014; Keinath et al. 2017; Weeks et al. 2023). The interplay between latitude and habitat preference emerges as a critical factor in the examination of species responses to forest edges. Species thriving in open areas may replace forest-dependent avian species in comparable numbers at forest edges, thereby contributing to an overall increase in species abundance and richness at edges (Tabarelli et al. 2012; Filgueiras et al. 2021).

Edge contrast had a significant effect on edge responses. In our study, bird species had lower abundances near soft than hard edges, in agreement with a recent review that documented lower species richness near soft edges (Wilmer et al. 2022). Although this result appears counter-intuitive at first sight, it is in accordance with the predictive model of Ries and Sisk (2004). This model assumes that two highly contrasting habitats may provide complementary resources, particularly for generalist species that can utilize both habitats. Comparative studies between edge/interior with different matrix types are still scarce (see the Methods section), but in the case of positive responses the contrast between intermediate and abrupt edges may result in higher richness and abundance of bird species, as they find complementary resources along the forest edge (Ries and Sisk 2004). Matrix species are generally open-area specialists, most of which do not go into the forest (Ewers and Didham 2006). Few generalist birds use, with similar frequency, the matrix, forest edge, and forest interior (Watson et al. 2004), while some species respond to vegetation structure and others do not (Ries et al. 2004). Despite the global geographical representation of edge effect studies (Ries et al. 2004), there remain important knowledge gaps that need to be addressed concerning the response of vertebrates to edge effects (Betts et al. 2017; Pfeifer et al. 2017), such as the “Parkerian” gap (i.e. lack of knowledge about basic natural history, which is key to understanding how species might respond to environmental challenges; Lees et al. 2020).

We were unable to evaluate landscape variables in detail due to the low number of replicates on more specific aspects such as edge location in regard to cardinal points, slope, or the amount of surrounding vegetation cover. Edge-effect studies covering large regions are needed to test for possible differences in local-level edge responses among macroecological scales, including different ecosystems or vegetation types (Banks-Leite et al. 2022). Studies on edge effects in particular, and habitat fragmentation in general, in high-altitude forests are also lacking. Our review revealed a higher number of publications on lower altitude areas than on high altitude areas. However, species confined to specific altitudinal strips tend to demonstrate increased specialization and possess lower dispersal capacity, as indicated by Cadena et al. (2012). The characteristics of species may functionally vary in ecosystems with different levels of degradation (Saavedra et al. 2014; Weeks et al. 2023), for example, species tend to better tolerate disturbance by humans in areas more prone to natural disturbance (Jahn et al. 2013; Harper et al. 2015), such as flooding or isolation in the Amazon or in southeast Asia forests, or recurring cyclones in northern Australia (Neate-Clegg et al. 2016). The concentration of threatened species is therefore probably higher in areas not subject to natural disturbance (Betts et al. 2019), as in some biodiversity hotspots (e.g. the Atlantic Forest in Brazil, and Choco-Darien-Tumbes in northwest of South America, (see in Jenkins et al. 2010; Mittermeier et al. 2011; Vale et al. 2018).

We have shown that species traits can play an important mediating role in the response of birds to forest edges. Our analyses reveal that there are general patterns in the response of species to edges, which may help to explain the well-known large variability in edge responses among species (Ries et al. 2004, 2017). As edge effects are one of the consequences of habitat fragmentation, our findings may also help to understand why habitat fragmentation impacts on biodiversity vary among species and landscapes (Fahrig 2017; Fletcher et al. 2018). In addition, our study assists in the identification of which groups may actually benefit from edge effects. In particular, open-area species are more likely to respond positively to forest edges in the tropics, as well as folivore-granivores guilds.