Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2025 Jan 20;28(1):e70061. doi: 10.1111/ele.70061

Weaker Plant‐Frugivore Trait Matching Towards the Tropics and on Islands

Xiao Huang 1,2, Bo Dalsgaard 3, Si‐Chong Chen 1,4,✉
PMCID: PMC11744342  PMID: 39829283

ABSTRACT

Biotic interactions play an important role in species diversification and maintenance and, thus, are regarded as the architecture of biodiversity. Since Darwin and Wallace, biologists have debated whether biotic interactions are stronger towards the tropics and on continents, when compared to temperate regions and islands. Here, based on 354 avian frugivory networks accounting for 22,199 interactions between 1247 bird species and 2126 plant species, we quantified trait matching strength, which reflects interaction strength and specificity, across gradients of latitude and insularity globally. We found that matching between beak size and fruit size was significantly stronger towards the poles and on continents, when compared with the tropics and on islands. As underlining ecological factors, trait matching was stronger with a larger proportion of frugivory (measured as the mean proportion of fruits in bird diets) and network‐level mean beak size, and with a smaller proportion of fleshy‐fruited species (measured as the proportion of fleshy‐fruited plant species in the botanical country where the network was located). These findings suggest that the latitudinal and insular patterns in trait matching are driven by biotic factors that may relate to trait co‐evolution between interacting species and optimal foraging for bird species.

Keywords: avian frugivore, biogeography, biotic interaction, co‐evolution, endozoochory, seed dispersal

This study investigates beak‐fruit size matching in frugivory to understand the strength of biotic interactions globally. Analysing 354 avian frugivory networks across different latitudes and insularity, the study finds that beak‐fruit size matching is stronger at higher latitudes and on continents compared to lower latitudes and islands. These results support Darwin's hypothesis that biotic interactions are weaker on islands but challenge Wallace's view that interactions are stronger in the tropics, with implications for ecological and evolutionary theories on global biodiversity patterns.

graphic file with name ELE-28-0-g005.jpg

1. Introduction

The co‐evolution between “Darwin's Orchid” (Angraecum sesquipedale) and the “Predicted Moth” (Xanthopan morganii praedicta) serves as a compelling example of morphological trait matching in nature (Darwin 1877). Trait matching, which describes the extent of how morphological traits are matched between interacting species, is prevalent in various types of biotic interactions, such as predation, pollination and frugivory (Chen and Moles 2015; Riede et al. 2011; Sonne et al. 2020). It influences species preference for potential partners and effectiveness in biotic interactions, thus playing an important role in species co‐evolution (Maglianesi et al. 2014; Vázquez et al. 2009). As a classic example of plant–animal mutualism, avian frugivory is crucial for seed dispersal of plants and energy acquirement of birds. In this mutualism, avian frugivores tend to consume fruits that closely match their beak size (Wheelwright 1985). Generally, birds with larger beaks are more likely to forage on larger fruits—a size matching phenomenon that has been recorded in several local‐scale observations (Donatti et al. 2011; Wheelwright 1985) and global syntheses (Chen and Moles 2015; McFadden et al. 2022). However, the global geographic pattern and the underlying mechanisms of trait matching in plant‐frugivore networks remain unclear, limiting our understanding of large‐scale mechanisms driving network construction and maintenance. This is due to the lack of a validated and standardised method to measure trait matching that properly accounts for sampling biases in networks compiled from various sources, and the incomplete trait information across a broad geographical range of plant‐frugivore networks. Here, we quantify the latitudinal and insular patterns of trait matching strength, and reveal the ecological factors underpinning these geographical patterns.

We first predict stronger trait matching in avian frugivory networks towards lower latitudes. Trait matching reflects the strength and specificity of biotic interactions (Dalsgaard et al. 2021; McFadden et al. 2022). Since Wallace, biotic interactions have been assumed to be more pronounced and specialised in the tropics (Wallace 1878), but the centuries of research testing this latitudinal interaction strength hypothesis remain inconclusive (Chen et al. 2017b; Dalsgaard et al. 2017; Moles and Ollerton 2016; Schleuning et al. 2012). Because tropical areas harbour higher biodiversity, interacting species may have closely‐matched traits to release competition pressure and enhance interaction effectiveness, facilitating species coexistence and community stability (Albrecht et al. 2018; but see Martin 1985). From an evolutionary perspective, bird species are more specialised in fruit diets in plant‐frugivore networks towards the tropics (Dalsgaard et al. 2017). Hence, bird species may gain more benefits from frugivory interactions and be under stronger mutualistic selection pressures in tropical regions. Tropical species also share a longer co‐evolutionary history under relatively stable environmental conditions, which might strengthen co‐evolution and matching of traits in biotic interactions (Schemske et al. 2009; Dalsgaard et al. 2011). Support for Wallace's idea of stronger interaction, and potentially trait matching, towards the tropics has surfaced in a few taxon‐specific groups, such as plant‐hummingbird pollination networks and co‐existing frugivorous birds and palms (McFadden et al. 2022; Sonne et al. 2020), however, we lack empirical evidence for community‐wide avian frugivory networks on a global scale.

Next, we predict stronger trait matching in networks on continents compared to those on islands. Darwin (1859) wrote that “on a small island, race for life will have been less severe”, suggesting weaker competition pressures and biotic interactions in island communities. On islands, biodiversity is lower, and biotic interaction networks are commonly more simplified in species composition and network structure than their counterparts on continents (Dalsgaard et al. 2018; Delavaux et al. 2024; González‐Castro, Traveset, and Nogales 2012; Traveset et al. 2016) because of geographic barriers and young ages of many islands (Losos and Ricklefs 2009). Thus, interspecies competition would be weak on islands, leading to the expansion of species niches and convergence of species traits (Olesen, Eskildsen, and Venkatasamy 2002; Si et al. 2022; Traveset et al. 2016). Within this context, we may suppose that generalised species on islands exhibit morphological features that enable them to interact with a wide range of partners, decreasing the extent of trait matching in interaction networks (Dalsgaard et al. 2021). Although there is a body of literature comparing the strength and specificity of biotic interactions on continents versus islands (González‐Castro, Traveset, and Nogales 2012; Schleuning et al. 2014; Traveset et al. 2016), little is known about how trait matching between interacting partners is shaped by insularity.

Latitudes and insularity are aggregates of diverse abiotic and biotic factors. Thus, we further quantify the relationships between trait matching strength and various ecological factors, aiming to identify the factors underpinning the geographic patterns of trait matching. For biotic factors, we predict stronger trait matching in networks with a higher proportion of fruits in bird diets (hereafter, frugivory proportion) and in regions with a higher proportion of fleshy‐fruited plant species. Frugivory evolution and mutualistic selection are pronounced in networks under such biotic conditions, thus frugivores may evolve to select better‐matched fruits to improve the efficiency of food processing and nutrient acquisition (Albrecht et al. 2018; Pedraza and Bascompte 2021). However, recent evidence also indicates that limited food resources might force frugivores to feed on morphologically matched fruits (Martins et al. 2024), thus the correlation between trait matching strength and fleshy‐fruited species proportion could be either positive or negative. We also predict that a higher level of trait differentiation, such as trait divergence and evenness of both bird communities and plant communities in networks, could reflect the more distinct partition of species niches, indicating stronger trait matching in the networks. Lastly, we predict that trait matching may be weaker with overall larger mean beak sizes or smaller mean fruit sizes (hereafter, network‐level beak size and network‐level fruit size, respectively). This weak matching may result from the fact that large‐beaked birds could feed on a wide range of fruit sizes, while small‐fruited plants could be consumed by birds with a wide range of beak sizes (Chen and Moles 2015).

For abiotic factors, we predict that traits are matched more strongly with higher mean annual temperature, precipitation, and actual evaporation. Ample heat, moisture and effective water cycling can support higher biodiversity and intensive biotic interactions (Coelho et al. 2023). Under such environments, trait co‐evolution is faster (McPeek 2017), and species tend to interact with better matched partners to increase interaction efficiency (Maglianesi et al. 2014; Vázquez et al. 2009). This tendency may further lead to strong trait matching in interaction networks (Sentis, Hemptinne, and Brodeur 2014). We also predict that trait matching strengthens under stable historical and current temperature and precipitation conditions. Stable environments benefit evolutionary processes (Dalsgaard et al. 2011), which favour the trait co‐evolution and morphological matching between interacting species (Fine 2015). Additionally, stable climates may lower the possibility of local extinctions (Sandel et al. 2011), thus prolonging the shared evolutionary history for species and facilitating the co‐existence of well trait‐matched interactions, which are generally more specialised and threatened due to climate changes (Dalsgaard et al. 2011; Whittaker et al. 2017).

Finally, after quantifying the potential direct effects of latitude, insularity and multiple ecological factors on trait matching strength, we aim to evaluate whether the effects of latitude and insularity are moderated by biotic and abiotic factors. For this, we performed structural equation models to test a hypothetical framework based on the literature evidence (Figure S1; Table S1–S4). We then identified the best‐fitted model to reveal the mechanisms driving geographical patterns in trait matching.

In this study, we compiled the most comprehensive dataset comprised of 354 avian frugivory networks accounting for 22,199 interactions between 1247 bird species and 2126 plant species. Our global dataset spans 96° of latitude from 43.75° S to 52.74° N, including 283 continental networks and 71 insular networks across a wide climatic gradient from major biogeographic realms (Figure 1). By quantifying the biogeographic pattern in size‐matching between bird beaks and plant fruits, we aim to test the hypothesis that trait matching is stronger towards the topics and on the mainland. Further, we seek to identify the main ecological factors driving the biogeographic variation in trait matching. The findings in this study expand our understanding of the processes underpinning the construction and maintenance of plant‐frugivore networks and biodiversity on a global scale.

FIGURE 1.

FIGURE 1

Open in a new tab

Geographic and climatic ranges of frugivory networks in this study. (a) Global locations of frugivory networks and trait matching strength. The inset shows the distributions of continental (yellow) and insular (grey) networks along latitudes. (b) Distribution of networks in Whittaker biomes according to the mean annual temperature and precipitation of sampled sites. (c) Proportion of continental and insular networks in major biogeographic realms.

2. Materials and Methods

2.1. Plant‐Frugivore Interaction Networks

We collected data on avian frugivory networks from several global syntheses (Chen and Moles 2015; Fricke and Svenning 2020; Martins et al. 2022), supplemented by recent publications before February 2024 from the Web of Sciences using the searching term:

(frugivor* OR endozoochor* OR interact* OR forag* OR network$ OR dispersal) AND (bird$ OR avian) AND (fruit$ OR seed$).

We included both binary networks (describing whether a given pair of bird and plant species interact with each other; n = 112) and weighted networks (describing interaction frequency for each pair of interacting species; n = 242) with species‐level interactions and excluded studies on specific taxa or amalgamated species. For interactions with unidentified bird or plant species, we removed those without trait information in the source papers. Only a small proportion of interactions were removed for this reason (less than 2% of the entire dataset). Plant species names were standardised according to the World Checklist of Vascular Plants (WCVP) (Govaerts et al. 2021), and bird species names were standardised according to the BirdLife International taxonomy, using the “gnr_solve” function in the R package taxize (Chamberlain and Szöcs 2017). In total, we acquired 354 networks from 160 studies ranging from 43.75° S to 52.74° N (Figure 1). These networks consist of 22,199 interactions between 1247 bird species and 2126 plant species. The flow diagram of literature synthesis is given in Figure S2. For each network, we retrieved the sampling location, time (year and month), duration (one season or multiple seasons), and method (faeces, observation, mixed methods or unknown method).

2.2. Beak Size and Fruit Size Data

We used beak width to represent the beak size of birds (horizontal width of beak at the anterior edge of nostrils), using the data from the AVONET database (Tobias et al. 2022). We decided to use beak width in our study because beak width data were comprehensively available in the AVONET database, and beak width is highly inter‐correlated with other measures of a bird beak such as gape size (Martins et al. 2024).

We used fruit length to represent the fruit size of plants, and extracted data from the source papers whenever possible (365 species, 22.3%). In other cases (1269 species, 77.7%), we collected fruit length data from published datasets (Bello et al. 2017; Kattge et al. 2020; Maitner et al. 2018; Sinnott‐Armstrong et al. 2018) and other online sources such as World Flora Online (https://www.worldfloraonline.org; last accessed in March 2024). Fruit length data from source papers and online databases were highly correlated (Figure S3). Overall, fruit length data coverage (proportion of interactions with empirical fruit length data in each network) varied from 14.4% to 100% across all networks, with a median of 93.0% (Figure S4). Among all networks, 88.2% of the interactions and 76.9% of the plant species had empirical fruit length data. Trait coverage shows no strong influence on latitudinal and insular patterns of trait matching in our main analyses (Text S1; Table S5).

We generated the phylogenetic tree from the megatree in the R package V. PhyloMaker2 (Jin and Qian 2022) for plant species with empirical fruit length information, and tested the phylogenetic signal of log‐transformed fruit length using the “phylosig” function in the phytools package (Revell 2012). Fruit length exhibited strong phylogenetic signal (Pagles' λ = 0.820, p < 0.001), thus we filled unavailable fruit length data using an imputation procedure accounting for the phylogenetic relationships among species via the R package missForests (Stekhoven and Bühlmann 2012). Before imputation, fruit length data were log‐transformed, centred, and scaled. The first 10 phylogenetic eigenvectors were included in the trait‐imputation matrix (Carmona et al. 2021). We repeated the imputation processes 1000 times to generate the mean of imputed values for the plant species without fruit length. Subsequently, we performed sensitivity analyses showing that our data imputation approach is reliable and robust (Text S1).

2.3. Biotic Factors

We calculated frugivory proportion as the mean proportion of fruits in the diets of bird species in each network based on the bird diet information acquired from the EltonTrait database (Wilman et al. 2014). Because of the lack of complete plant species lists for the network sampling sites, we calculated fleshy‐fruited species proportion as the proportion of fleshy‐fruited species in botanical countries, i.e. World Geographical Scheme for Recording Plant Distributions (WGSRPD) Level 3 unit where the networks were located (Brummitt et al. 2001; Figure S5). This is the best available data of fleshy‐fruited species proportion we could access so far at the global scale. The data represents the overall proportion of available fleshy‐fruited resources for frugivorous birds, although this coarse estimation may cause less variation among networks. Continents and remote islands are generally delineated into distinct botanical countries in WGSRPD, and there are no continental and insular networks included within the same botanical country in our dataset. Data on the presence or absence of plant species for each botanical country are from the World Checklist of Vascular Plants (Govaerts et al. 2021). Identification of fleshy‐fruited species was conducted using the list provided in Sinnott‐Armstrong et al. (2018). We calculated network‐level beak size and network‐level fruit size as mean value of beak width and fruit length, respectively, for each network as potential biotic factors influencing trait matching. For functional diversity indices, in each network, we calculated functional evenness and divergence for bird and plant species, respectively, using the “REND” function in the R package TPD (Carmona et al. 2019). Functional evenness describes the regularity of trait value distribution, and functional divergence reflects the extent of trait value distribution that differs from the centralised pattern (Mason et al. 2005). These two metrics both represent the trait differentiation of species assemblies. We constructed null models by randomly selecting the same number of bird or plant species with each network from all the species included in our datasets and calculated the functional diversity indices 999 times. The observed functional diversity of networks was subtracted from the mean values of null models to get the standardised functional diversity, which was used in further analyses (Carmona et al. 2019).

2.4. Abiotic Factors

Mean annual temperature, temperature seasonality (standard deviation of the monthly mean temperatures), mean annual precipitation, precipitation seasonality (mean monthly precipitation coefficient of variation) during 1981–2010, and mean annual temperature and precipitation in the Last Glacial Maximum were from the CHELSA database (Karger et al. 2017, 2020). Historical temperature change and historical precipitation change were calculated as the differences between current and historical climates (Lim et al. 2020). Mean annual actual evaporation during 2001–2019 was calculated using monthly averaged data from the ETMonitor dataset (Zheng, Jia, and Hu 2022) The spatial resolution of all the climatic data is 1 km × 1 km, and we in all cases extracted the climatic information from the grid cell where the network was located.

2.5. Trait Matching Strength of Networks

To measure the trait matching strength in networks, we first standardised trait values by centring and scaling the beak size of bird species and fruit length of plant species within each network, and then calculated the absolute difference for each interaction pair (Sonne et al. 2020). We used standardised trait values instead of raw values due to two benefits: (1) since the interspecific variations of both beak sizes and fruit sizes of plants vary across communities, the standardised trait values can better reflect the relative sizes of beaks and fruits of species within each community (2) standardisation can decrease the influence of extreme trait values in the networks. We then calculated the mean trait dissimilarity of all interactions (MD) for each network. To account for the influence of network size and obtain an unbiased estimate of network trait‐matching strength, we generated 1000 random networks by constraining the generalisation of both bird and plant species (Bascompte et al. 2003). The standardised MD is calculated as MDstd = MD—MDnull, in which MDnull is the mean MD of 1000 randomised networks (Schleuning et al. 2012). We used the opposite values of MDstd to represent trait matching strength in networks. With this measurement, strong trait matching in a network indicates that birds with large beaks are more likely to feed on large fruits, while birds with small beaks are more likely to eat small fruits. We performed additional modelling to illustrate how MDstd can well reflect the trait matching patterns in networks, and are robust to scenarios in which we assume that all bird species are gulpers or mashers (Text S2; Figures S6 and S7). We also presented three empirical networks with different extents of trait matching to visually show how the metrics are changed according to the trait matching pattern of networks (Figure S8).

While mean values of trait dissimilarity reflect the overall trait matching strength of networks, different moments of the trait dissimilarity distribution (i.e. quantiles of trait dissimilarity) would provide additional information of the geographic patterns of trait matching. Hence, we also calculated 10%, 50% and 90% quantiles of trait dissimilarity in each network, and quantified their latitudinal and insular patterns. A higher quantile of trait dissimilarity reflects weaker trait matching, whereas a lower quantile of trait dissimilarity reflects stronger trait matching. The analyses on quantiles of trait dissimilarity in networks generally reflect consistent patterns with the main results, although there are slight variations across quantiles (Text S3; Figure S9; Table S6).

We measured trait matching strength (MDstd) based on the binary networks (hereafter, binary metrics) in the main analysis, thereby emphasising the ability of species to interact with partners with particular traits and maximise the geographic coverage of our study. Binary networks primarily indicate possible versus impossible (i.e. forbidden links) interactions, whereas weighted networks better capture the variety of fruit selection. To clarify how the geographical patterns of trait matching might be influenced by considering interaction frequency, we performed additional models including the best available weighted metrics data in the dataset (Text S4). We ran an ordinary least square (OLS) model to test the relationship between MDstd and WMDstd of quantitative networks. MDstd and WMDstd have a significantly positive relationship according to the model results (Figure S10).

2.6. Statistics

To quantify the latitudinal and insular patterns of trait matching, we fitted linear mixed‐effects models (LMM) using the lme4 package (Bates et al. 2015). We set random effects and covariates to account for the effects of different sample processes. In all LMMs, the random effects were studies in which the networks were extracted. The covariates were sampling duration, sampling method and sampling intensity of networks. According to the description in the original studies, we categorised sampling durations as a binary metric: sampling occurs in a specific season or year‐round. We categorised sampling methods into four types: (1) sampling from animal faeces; (2) sampling from focal or transect observations; (3) sampling from both animal faeces and observations; (4) sampling method unavailable in the data sources. Sampling intensity was measured as the square‐root of observed number of interactions divided by the square‐root of the product of the numbers of bird and plant species in each network (Schleuning et al. 2012). In the models testing the geographic patterns of trait matching strength, absolute latitude (scaled and centred) and geographic origin of networks (continents vs. islands) were set as fixed effects collectively and separately. The statistical significance was generated by the t‐value statistic test via Satterthwaite's degrees of freedom method using the lmerTest package (Kuznetsova, Brockhoff, and Christensen 2017). We applied Moran's I test on model residuals using the spdep package (Bivand 2022). The results showed that there was no significant spatial autocorrelation in the model (Moran's I = 0.015, p = 0.216).

Next, we quantified the effects of different ecological factors on trait matching and revealed the influential factors using linear mixed‐effects models. All biotic and abiotic factors were set as fixed effects (i.e. mean annual temperature, mean annual precipitation, actual evaporation, temperature seasonality, precipitation seasonality, historical temperature change, historical precipitation change, frugivory proportion, fleshy‐fruited species proportion, network‐level beak size, network‐level fruit size, bird trait divergence, bird trait evenness, plant trait divergence, plant trait evenness). The covariates were sampling duration, sampling method and sampling intensity. The random effects were studies in which the networks were extracted. All the predictors were scaled and centred before fitting the models. To identify the possible existence of multicollinearity of these predictors, we calculated the variance inflation factor (VIF). The factors all had VIF smaller than 3 (Table S7), indicating no collinearity. There was also no significant spatial autocorrelation in this full model (Moran's I = −0.007, p = 0.5797). To account for model uncertainty and improve the robustness of model predictions, we applied a multi‐model inference and averaging approach for model selection and parameter estimation using the R package MuMIn (Bartoń 2023). We initially fitted a set of models with all possible combinations of the predictors using the “dredge” function and sorted the candidate models by their Akaike Information Criterion (AIC) values. Then, we selected the models with ΔAIC smaller than two and generated averaged model coefficients and corresponding P of each fixed effect using the “model.avg” function (Table S8). To further explore the relative importance of different ecological factors to the variation of trait matching strength, we applied a hierarchical partition method to split the influence of fixed effects on the response variable using the R package glmm.hp (Lai et al. 2022).

Finally, to clarify how the latitudinal and insular patterns of trait matching are moderated by biotic and abiotic factors directly and indirectly, we performed piecewise structural equation models using the R package piecewiseSEM (Lefcheck 2016). The initial conceptual model was constructed based on previous evidence of the relationships among our predictors, and the hypothetical effects of these predictors on trait matching. (see details in Figure S1 and Tables S1–S4). With the initial model, we subsequently aimed at identifying the best‐fitted model to the data, by stepwise eliminating non‐significant pathways until the model showed an overall good fit through Fisher's C statistic (Lefcheck 2016). We also tested if there were additional paths not included in our initial models or correlated error structures between predictors that could improve the fitness of the model. Because we aimed to quantify the effects of predictors on trait matching strength, a given predictor was removed if it was not directly or indirectly linked with trait matching strength in the model. The structure of the final model was selected from the set of candidate models with non‐significant Fisher's C value (p > 0.05) according to their AICs (Lefcheck 2016).

All analyses were performed using R 4.3.2 (R Core Team 2023).

3. Results

3.1. Latitudinal and Insular Patterns of Trait Matching

Contradictory with our hypothesis, trait matching was significantly stronger towards the poles (Slope = 0.015, S.E. = 0.007, t = 2.160, p = 0.033, accounting for the effect of insularity, Figure 2a; Slope = 0.016, S.E. = 0.007, t = 2.226, p = 0.022, not accounting for the effect of insularity; Table S9). That is, bird beak size is better matched with plant fruit size at higher latitudes. The pattern was consistent when including weighted metrics in the model (p = 0.021), but non‐significant when only using the smaller dataset of weighted networks (p = 0.224; Table S10).

FIGURE 2.

FIGURE 2

Open in a new tab

Geographic patterns of trait matching in avian frugivory networks. Yellow symbols illustrate continental networks, and grey symbols illustrate insular networks. (a) Trait matching strength along absolute latitudes. The solid line represents model prediction (continental networks as the reference level), and grey‐shaded area represents the 95% confidential interval. (b) Trait matching strength on continents vs. islands. Violin plots represent the distribution of trait matching strength. Dots and error bars show the mean value and 95% confidential interval, respectively. Detailed model results are given in Table S9.

Consistent with our hypothesis, plant‐frugivore trait matching was significantly stronger on continents when compared to islands (Slope = −0.040, S.E. = 0.017, t = −2.318, p = 0.022, accounting for the effect of latitudes, Figure 2b; Slope = −0.041, S.E. = 0.017, t = −2.380, p = 0.019, not accounting for the effect of latitudes; Table S9). That is, bird beak size is better matched with plant fruit size on continents. This tendency was non‐significant when including weighted metrics in the model (p = 0.308), and when only using the smaller dataset of weighted networks (p = 0.644; Table S10).

3.2. Effects of Ecological Factors on Trait Matching

Among all ecological factors considered, trait matching was significantly positively related to network‐level beak size of bird species (p = 0.016) and negatively related to fleshy‐fruited species proportion (p < 0.001, Figure 3a; Table S11). Frugivory proportion conducted positive effect on trait matching, which was marginally significant (p = 0.065; Figure 3a; Table S11). Fleshy‐fruited species proportion, frugivory proportion and network‐level beak size were the three most important factors, which explained 34.9%, 14.0% and 13.2% of the variance of trait matching, respectively (Figure 3b; Table S12). Functional diversity of bird and plant species, and current and historical climate all showed non‐significant effects on trait matching strength (Figure 3b; Table S12).

FIGURE 3.

FIGURE 3

Open in a new tab

Effects of ecological factors on trait matching strength. (a) Model‐averaged coefficients of ecological factors. Points and error bars represent mean values and 95% confidential interval of the coefficients. (b) The relative importance of ecological factors in explaining variances in trait matching strength. The proportion of variance is calculated as the individual contribution of each factor to the total variances (marginal R 2) of the linear mixed‐effects model with trait matching strength as the response variable and all ecological factors as fixed effects. Detailed model results are given in Tables S11 and S12.

3.3. Geographic Patterns of Trait Matching Moderated by Ecological Factors

The final piecewise structural equation models had a good fit to the data (Fisher's C = 13.03, p = 0.367, Figure 4a). Latitude directly influenced trait matching, but the direct effect of insularity was not significant (Table S13). The latitudinal and insular patterns of trait matching were partly explained by latitudinal and insular influences on biotic factors, either directly or indirectly via climates (Figure 4a). These variations of biotic factors influenced trait matching directly and differently. Towards lower latitudes, there was a larger proportion of fleshy‐fruited species in the botanical countries, frugivory proportion and network‐level beak size of bird communities. The frugivory proportion was smaller on islands compared with continents, while islands tended to harbour a larger proportion of fleshy‐fruited plant species. While frugivory proportion and mean beak size had a positive effect on trait matching strength, the fleshy‐fruited species proportion had a negative effect. They together resulted in the negative indirect effects of latitudes on trait matching, and led to the positive indirect effects of insularity (Figure 4b; Table S14). This indicates that geographic patterns of trait matching are mainly controlled by biotic factors that may drive trait co‐evolution between interacting species and optimal foraging for bird species. Mean annual temperature and temperature seasonality only conducted indirect and weak effects on trait matching (Figure 4; Table S14).

FIGURE 4.

FIGURE 4

Open in a new tab

Results of the piecewise structural equation model. (a) Final model showing direct and indirect effects of latitude, insularity and main ecological factors on trait matching strength. Solid arrows indicate significant positive (blue) and negative (red) effects, respectively (p < 0.05). Dashed arrows indicate non‐significant effects. Double‐headed arrows indicate correlation between factors. Values along each path are standardised model coefficients of the piecewise structural equation model. R 2 shows the explained variance of the response variable. (b) Summary of direct, indirect and total effects of latitudes, insularity, and ecological factors on trait matching strength. Bars and error bars represent mean values and their 95% confidential intervals. Detailed model results are given in Tables S13 and S14.

4. Discussion

We provide a global assessment of trait matching strength using a comprehensive range of empirical data on avian frugivory networks. The size matching between beaks of birds and fruits of plants tends to be stronger towards higher latitudes and on continents. The large‐scale patterns of trait matching were primarily shaped by biotic factors, and only weakly influenced by climatic factors. These findings offer new insights into the potential mechanisms explaining the global gradients of interaction strength and specificity. Moreover, we highlight regions with stronger trait matching in frugivory networks that may face greater cascading extinction risks in a world characterised by biodiversity decline.

Contrary to our prediction, trait matching is stronger towards higher latitudes (Figure 2). This finding contributes to a growing body of evidence challenging the traditional view that biotic interactions are more pronounced and specialised towards the tropics (Moles and Ollerton 2016). Frugivorous birds may optimise fruit size consumed to meet their energetic needs when fleshy fruit resources are limited, as predicted by optimal foraging theory. Such conditions may exist at the geographical range edges of birds (Martins et al. 2024) and, as shown in our study, at high latitudes where fleshy‐fruited species proportion is low (Chen et al. 2017). Besides limited fruit resources, the harsher living conditions for birds at higher latitudes likely exert stronger selection pressures, forcing them to feed on better‐matched fruits to enhance foraging efficiency. This may explain the direct effect of latitude on trait matching, alongside its indirect effects via fleshy‐fruited species proportion (Figure 4). Our findings are supported by two previous studies, which also show that frugivory network structures are more specialised at temperate latitudes (Dalsgaard et al. 2017; Schleuning et al. 2012). However, our results contrast with a study on hummingbird‐flower pollination networks in Central and South America (Sonne et al. 2020) and a study focused on modelled coexistence of palms and frugivorous birds (McFadden et al. 2022)—two systems where trait matching is most likely driven by co‐evolution. As a highly diffuse interaction type, frugivory may be more influenced by ecological processes such as optimal foraging than by evolutionary processes such as trait co‐evolution, which is also indicated by the stronger impact of fleshy‐fruited species proportion (i.e. resources) over frugivory proportion (Figure 3). These two biotic factors show parallel changes along latitudes, and their opposing effects on trait matching strength jointly shape geographic patterns. Nevertheless, some inland subtropical regions, such as western Argentina and India, exhibit low fleshy‐fruited species proportions but high frugivory proportions, potentially due to the dry climates in these regions (Chen et al. 2017a). Taken together, the effects of fruit resources and other potential ecological constraints on trait matching outweigh the influence of frugivory evolution, as indicated by frugivory proportion and network‐level beak sizes. This leads to the overall increasing pattern of morphological matching towards higher latitudes, suggesting that ecological processes play a more important role than evolutionary processes in driving this pattern for plant‐frugivore networks.

The significantly weaker trait matching in insular networks (Figure 2) aligns with island theory and recent evidence that island communities tend to be more generalised (Dalsgaard et al. 2021). On islands, the low frugivory proportion suggest a weaker evolutionary force of morphological matching between fruit and beak size. Meanwhile, the high fleshy‐fruited species proportion allow bird species to feed on fruits of various sizes to achieve energy needs. However, the non‐significant mediating effects of network‐level trait values on insular pattern of trait matching contradicts the prediction that insular communities exhibit traits enabling species to interact with a wide range of mutualistic partners. Still, our findings support the long‐standing idea that interaction strength declines on islands, which is perhaps derived from the distinct ecological and evolutionary features of islands, such as lower biodiversity and dominance of endemic “super generalists” (Dalsgaard et al. 2021; Delavaux et al. 2024; Olesen, Eskildsen, and Venkatasamy 2002). The stronger negative effect of insularity using the 90% quantile of trait dissimilarity in networks also indicates the crucial role of generalists in shaping the insular pattern of trait matching, compared with the results from the 10% and 50% quantiles (Figure S6). This result is consistent with the fact that island communities are more seriously impacted by non‐native species, which are typically generalised and occupy broad niches (Peralta, Perry, Vázquez, Dehling, and Tylianakis 2020; Peralta et al. 2020; Vollstädt et al. 2022). Furthermore, island communities harbour unique species diversity, and are fragile to anthropogenic stresses and climate change (Gonçalves et al. 2024). Although weaker trait matching indicates a lower possibility of cascading co‐extinctions in insular networks, it may potentially signify that considerable species loss and turnover have already transpired, with large and specialised frugivores more easily lost in this process (Heinen et al. 2023; Soares et al. 2021; Whittaker et al. 2017).

The stronger trait matching observed with larger network‐level beak sizes (Figure 3) challenges our hypothesis that bird species with large beaks exhibit weaker trait matching because they potentially interact with a wider range of fruit sizes. Large‐beaked frugivores can easily swallow fruits and likely rely more on a fruit diet in their co‐evolutionary history with plants. This is evident in our result that network‐level beak size is positively correlated with frugivory proportion (Figure 4a), suggesting an adaptation towards feeding on fleshy fruits in frugivorous bird communities with larger beaks. Hence, these bird communities are likely more selective in their fruit choices and under stronger mutualistic selection by fruit sizes. This selection pressure may outweigh the trait barriers affecting the co‐evolution of small‐beaked frugivores with fruit sizes, leading to stronger trait matching among species with large beaks.

The geographical patterns of trait matching are overall consistent between our main model and the model that includes both weighted and binary metrics. Previous evidence has also shown that the “binary reduction” (i.e. treating weighted networks as binary) does not significantly alter global patterns for interaction network indices (Corso et al. 2015). The binary metrics can maximise the geographic coverage of our analyses and may be less influenced by ecological factors, such as species abundance, phenology and other specific environmental conditions during the sampling period, especially for generalist species (Peralta, Perry, Vázquez, Dehling, and Tylianakis 2020; Peralta et al. 2020), which have advantages in our global quantification of trait‐matching using network data from various studies. This may account for the greater variation in weighted trait matching metrics and the non‐significant patterns in the model only using the smaller dataset of weighted networks, especially for the generalised island networks impacted more by ecological factors (Dalsgaard and Temeles 2024). In accordance with this, the models using the most comprehensive dataset including binary networks show consistent patterns of weaker plant‐frugivore trait matching towards the tropics and on islands.

In conclusion, our study has quantified the global patterns of trait matching between the size of bird beaks and plant fruits in avian frugivory networks, which supports Darwin's idea that biotic interactions on islands are weaker than those on continents, but rejects Wallace's assumption that interactions are stronger towards the tropics. The geographic variation of trait matching is mainly affected by biotic factors related to the optimal foraging for bird species and trait co‐evolution in communities. Our findings indicate that trait matching in avian frugivore‐plant networks are shaped by both evolutionary and ecological processes, while ecological processes seem to a play more important role. By comparing our findings with previous studies, we highlight that the relative effects of different mechanisms in determining trait matching may be varied according to how morphological matching is measured and the studied interaction types. Our study provides novel evidence and insights into the long‐standing debates of global patterns of interaction strength and specificity, especially for highly diffuse mutualistic systems such as plant‐frugivore networks. Additionally, the methodological and theoretical frameworks presented here can be further applied to other interaction types to understand the large‐scale evolution and maintenance of biotic interaction networks.

Author Contributions

Si‐Chong Chen and Xiao Huang designed the study. Xiao Huang collected, analysed and visualised the data with input from Si‐Chong Chen and Bo Dalsgaard. Xiao Huang drafted the initial manuscript. Si‐Chong Chen and Bo Dalsgaard reviewed and edited the manuscript.

Conflicts of Interest

The authors declare no conflicts of interests.

Peer Review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/ele.70061.

Supporting information

Data S1.

ELE-28-0-s001.docx (3MB, docx)

Acknowledgements

S‐CC was supported by the National Natural Science Foundation of China (32371612), and the start‐up research grant from Wuhan Botanic Garden (E1559902). BD was supported by Independent Research Fund Denmark (0135‐00333B). We thank Yang Liu for the beneficial discussions. We thank authors of the data source papers who collected the empirical data on the avian seed dispersal network and traits of bird and plant species that are used in this study.

Editor: Marlee A Tucker

Funding: Si‐Chong Chen was supported by the National Natural Science Foundation of China (32371612), and the start‐up research grant from Wuhan Botanic Garden (E1559902). BD was supported by Independent Research Fund Denmark (0135‐00333B).

Data Availability Statement

Data and codes are available at the figshare repository (https://doi.org/10.6084/m9.figshare.25451611).

References

  1. Albrecht, J. , A. A., Vollstädt M. G. R., et al. 2018. “Plant and Animal Functional Diversity Drive Mutualistic Network Assembly Across an Elevational Gradient.” Nature Communications 9: 3177. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bartoń, K. 2023. “MUMIn: Multi‐model inference. R Package Version 1.47.5.”, https://cran.r‐project.org/web/packages/MuMIn/index.html.
  3. Bascompte, J. , Jordano P., Melián C. J., and Olesen J. M.. 2003. “The Nested Assembly of Plant–Animal Mutualistic Networks.” Proceedings of the National Academy of Sciences 100: 9383–9387. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bates, D. , Mächler M., Bolker B., and Walker S.. 2015. “Fitting Linear Mixed‐Effects Models Using lme4.” Journal of Statistical Software 67: 1–48. [Google Scholar]
  5. Bello, C. , Galetti M., Montan D., et al. 2017. “Atlantic Frugivory: A Plant–Frugivore Interaction Data Set for the Atlantic Forest.” Ecology 98: 1729. [DOI] [PubMed] [Google Scholar]
  6. Bivand, R. 2022. “R Packages for Analyzing Spatial Data: A Comparative Case Study With Areal Data.” Geographical Analysis 54: 488–518. [Google Scholar]
  7. Brummitt, R. K. , Pando F., Hollis S., and Brummitt N. A.. 2001. World Geographical Scheme for Recording Plant Distributions. 2nd ed. Pittsburgh: Hunt Institute for Botanical Documentation. [Google Scholar]
  8. Carmona, C. P. , Bueno C. G., Toussaint A., et al. 2021. “Fine‐Root Traits in the Global Spectrum of Plant Form and Function.” Nature 597: 683–687. [DOI] [PubMed] [Google Scholar]
  9. Carmona, C. P. , de Bello F., Mason N. W. H., and Lepš J.. 2019. “Trait Probability Density (TPD): Measuring Functional Diversity Across Scales Based on TPD With R.” Ecology 100: e02876. [DOI] [PubMed] [Google Scholar]
  10. Chamberlain, S. A. , and Szöcs E.. 2017. “Taxize: Taxonomic Search and Retrieval in R.” F1000Research 2: 191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Chen, S.‐C. , Cornwell W. K., Zhang H.‐X., and Moles A. T.. 2017a. “Plants Show More Flesh in the Tropics: Variation in Fruit Type Along Latitudinal and Climatic Gradients.” Ecography 40: 531–538. [Google Scholar]
  12. Chen, S.‐C. , Hemmings F. A., Chen F., and Moles A. T.. 2017b. “Plants Do Not Suffer Greater Losses to Seed Predation Towards the Tropics.” Global Ecology and Biogeography 26: 1283–1291. [Google Scholar]
  13. Chen, S.‐C. , and Moles A. T.. 2015. “A Mammoth Mouthful? A Test of the Idea That Larger Animals Ingest Larger Seeds.” Global Ecology and Biogeography 24: 1269–1280. [Google Scholar]
  14. Coelho, M. T. P. , Barreto E., Rangel T. F., et al. 2023. “The Geography of Climate and the Global Patterns of Species Diversity.” Nature 622: 537–544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Corso, G. , Torres Cruz C. P., Pinto M. P., de Almeida A. M., and Lewinsohn T. M.. 2015. “Binary Versus Weighted Interaction Networks.” Ecological Complexity 23: 68–72. [Google Scholar]
  16. Dalsgaard, B. , Kennedy J. D., Simmons B. I., et al. 2018. “Trait Evolution, Resource Specialization and Vulnerability to Plant Extinctions Among Antillean Hummingbirds.” Proceedings of the Royal Society B: Biological Sciences 285: 20172754. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Dalsgaard, B. , Magård E., Fjeldså J., et al. 2011. “Specialization in Plant‐Hummingbird Networks Is Associated With Species Richness, Contemporary Precipitation and Quaternary Climate‐Change Velocity.” PLoS One 6: e25891. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dalsgaard, B. , Maruyama P. K., Sonne J., et al. 2021. “The Influence of Biogeographical and Evolutionary Histories on Morphological Trait‐Matching and Resource Specialization in Mutualistic Hummingbird–Plant Networks.” Functional Ecology 35: 1120–1133. [Google Scholar]
  19. Dalsgaard, B. , Schleuning M., Maruyama P. K., et al. 2017. “Opposed Latitudinal Patterns of Network‐Derived and Dietary Specialization in Avian Plant‐Frugivore Interaction Systems.” Ecography 40: 1395–1401. [Google Scholar]
  20. Dalsgaard, B. , and Temeles E. J.. 2024. “Hurricanes Threaten Species and Alter Evolutionary Trajectories on Tropical Islands.” Current Biology 34: R1115–R1120. [DOI] [PubMed] [Google Scholar]
  21. Darwin, C. 1859. On the Origin of Species by Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life. UK, London: John Murray. [PMC free article] [PubMed] [Google Scholar]
  22. Darwin, C. 1877. The Various Contrivances by Which Orchids Are Fertilised by Insects. UK, London: John Murray. [Google Scholar]
  23. Delavaux, C. S. , Crowther T. W., Bever J. D., Weigelt P., and Gora E. M.. 2024. “Mutualisms Weaken the Latitudinal Diversity Gradient Among Oceanic Islands.” Nature 627: 335–339. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Donatti, C. I. , Guimarães P. R., Galetti M., Pizo M. A., Marquitti F. M. D., and Dirzo R.. 2011. “Analysis of a Hyper‐Diverse Seed Dispersal Network: Modularity and Underlying Mechanisms: Modularity in a Seed Dispersal Network.” Ecology Letters 14: 773–781. [DOI] [PubMed] [Google Scholar]
  25. Fine, P. V. A. 2015. “Ecological and Evolutionary Drivers of Geographic Variation in Species Diversity.” Annual Review of Ecology, Evolution, and Systematics 46: 369–392. [Google Scholar]
  26. Fricke, E. C. , and Svenning J.‐C.. 2020. “Accelerating Homogenization of the Global Plant–Frugivore Meta‐Network.” Nature 585: 74–78. [DOI] [PubMed] [Google Scholar]
  27. Gonçalves, F. , Farooq H., Harfoot M., et al. 2024. “A Global Map of Species at Risk of Extinction due to Natural Hazards.” Proceedings of the National Academy of Sciences 121: e2321068121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. González‐Castro, A. , Traveset A., and Nogales M.. 2012. “Seed Dispersal Interactions in the Mediterranean Region: Contrasting Patterns Between Islands and Mainland.” Journal of Biogeography 39: 1938–1947. [Google Scholar]
  29. Govaerts, R. , Lughadha E. N., Black N., Turner R., and Paton A.. 2021. “The World Checklist of Vascular Plants, a Continuously Updated Resource for Exploring Global Plant Diversity.” Scientific Data 8: 215. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Heinen, J. H. , Florens F. B. V., Baider C., et al. 2023. “Novel Plant–Frugivore Network on Mauritius Is Unlikely to Compensate for the Extinction of Seed Dispersers.” Nature Communications 14: 1019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Jin, Y. , and Qian H.. 2022. “V.PhyloMaker2: An Updated and Enlarged R Package That Can Generate Very Large Phylogenies for Vascular Plants.” Plant Diversity 44: 335–339. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Karger, D. N. , Conrad O., Böhner J., et al. 2017. “Climatologies at High Resolution for the earth's Land Surface Areas.” Scientific Data 4: 170122. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Karger, D. N. , Nobis M. P., Normand S., Graham C. H., and Zimmermann N. E.. 2020. “CHELSA‐TraCE21k: Downscaled Transient Temperature and Precipitation Data Since the Last Glacial Maximum.” EnviDat. 10.16904/envidat.211. [DOI] [Google Scholar]
  34. Kattge, J. , Bönisch G., Díaz S., et al. 2020. “TRY Plant Trait Database—Enhanced Coverage and Open Access.” Global Change Biology 26: 119–188. [DOI] [PubMed] [Google Scholar]
  35. Kuznetsova, A. , Brockhoff P. B., and Christensen R. H. B.. 2017. “lmerTest Package: Tests in Linear Mixed Effects Models.” Journal of Statistical Software 82: 1–26. [Google Scholar]
  36. Lai, J. , Zou Y., Zhang S., Zhang X., and Mao L.. 2022. “Glmm.Hp: An R Package for Computing Individual Effect of Predictors in Generalized Linear Mixed Models.” Journal of Plant Ecology 15: 1302–1307. [Google Scholar]
  37. Lefcheck, J. S. 2016. “piecewiseSEM: Piecewise Structural Equation Modelling in R for Ecology, Evolution, and Systematics.” Methods in Ecology and Evolution 7: 573–579. [Google Scholar]
  38. Lim, J. Y. , Svenning J.‐C., Göldel B., Faurby S., and Kissling W. D.. 2020. “Frugivore‐Fruit Size Relationships Between Palms and Mammals Reveal Past and Future Defaunation Impacts.” Nature Communications 11: 4904. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Losos, J. B. , and Ricklefs R. E.. 2009. “Adaptation and Diversification on Islands.” Nature 457: 830–836. [DOI] [PubMed] [Google Scholar]
  40. Maglianesi, M. A. , Blüthgen N., Böhning‐Gaese K., and Schleuning M.. 2014. “Morphological Traits Determine Specialization and Resource Use in Plant–Hummingbird Networks in the Neotropics.” Ecology 95: 3325–3334. [Google Scholar]
  41. Maitner, B. S. , Boyle B., Casler N., Condit R., Donoghue J., and Durán S. M.. 2018. “The Bien r Package: A Tool to Access the Botanical Information and Ecology Network (BIEN) Database.” Methods in Ecology and Evolution 9: 373–379. [Google Scholar]
  42. Martin, T. E. 1985. “Resource Selection by Tropical Frugivorous Birds: Integrating Multiple Interactions.” Oecologia 66: 563–573. [DOI] [PubMed] [Google Scholar]
  43. Martins, L. P. , Stouffer D. B., Blendinger P. G., et al. 2022. “Global and Regional Ecological Boundaries Explain Abrupt Spatial Discontinuities in Avian Frugivory Interactions.” Nature Communications 13: 6943. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Martins, L. P. , Stouffer D. B., Blendinger P. G., et al. 2024. “Birds Optimize Fruit Size Consumed Near Their Geographic Range Limits.” Science 385: 331–336. [DOI] [PubMed] [Google Scholar]
  45. Mason, N. W. H. , Mouillot D., Lee W. G., and Wilson J. B.. 2005. “Functional Richness, Functional Evenness and Functional Divergence: The Primary Components of Functional Diversity.” Oikos 111: 112–118. [Google Scholar]
  46. McFadden, I. R. , Fritz S. A., Zimmermann N. E., et al. 2022. “Global Plant‐Frugivore Trait Matching Is Shaped by Climate and Biogeographic History.” Ecology Letters 25: 686–696. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. McPeek, M. A. 2017. “The Ecological Dynamics of Natural Selection: Traits and the Coevolution of Community Structure.” American Naturalist 189: E91–E117. [DOI] [PubMed] [Google Scholar]
  48. Moles, A. T. , and Ollerton J.. 2016. “Is the Notion That Species Interactions Are Stronger and More Specialized in the Tropics a Zombie Idea?” Biotropica 48: 141–145. [Google Scholar]
  49. Olesen, J. M. , Eskildsen L. I., and Venkatasamy S.. 2002. “Invasion of Pollination Networks on Oceanic Islands: Importance of Invader Complexes and Endemic Super Generalists.” Diversity and Distributions 8: 181–192. [Google Scholar]
  50. Pedraza, F. , and Bascompte J.. 2021. “The Joint Role of Coevolutionary Selection and Network Structure in Shaping Trait Matching in Mutualisms.” Proceedings of the Royal Society B: Biological Sciences 288: 20211291. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Peralta, G. , Perry G. L. W., Vázquez D. P., Dehling D. M., and Tylianakis J. M.. 2020. “Strength of Niche Processes for Species Interactions Is Lower for Generalists and Exotic Species.” Journal of Animal Ecology 89: 2145–2155. [DOI] [PubMed] [Google Scholar]
  52. Peralta, G. , Vázquez D. P., Chacoff N. P., Lomáscolo S. B., Perry G. L. W., and Tylianakis J. M.. 2020. “Trait Matching and Phenological Overlap Increase the Spatio‐Temporal Stability and Functionality of Plant–Pollinator Interactions.” Ecology Letters 23: 1107–1116. [DOI] [PubMed] [Google Scholar]
  53. R Core Team . 2023. “R: A Language and Environment for Statistical Computing.”, https://www.r‐project.org/.
  54. Revell, L. J. 2012. “Phytools: An R Package for Phylogenetic Comparative Biology (And Other Things).” Methods in Ecology and Evolution 3: 217–223. [Google Scholar]
  55. Riede, J. O. , Brose U., Ebenman B., et al. 2011. “Stepping in Elton's Footprints: A General Scaling Model for Body Masses and Trophic Levels Across Ecosystems.” Ecology Letters 14: 169–178. [DOI] [PubMed] [Google Scholar]
  56. Sandel, B. , Arge L., Dalsgaard B., et al. 2011. “The Influence of Late Quaternary Climate‐Change Velocity on Species Endemism.” Science 334: 660–664. [DOI] [PubMed] [Google Scholar]
  57. Schemske, D. W. , Mittelbach G. G., Cornell H. V., Sobel J. M., and Roy K.. 2009. “Is There a Latitudinal Gradient in the Importance of Biotic Interactions?” Annual Review of Ecology, Evolution, and Systematics 40: 245–269. [Google Scholar]
  58. Schleuning, M. , Böhning‐Gaese K., Dehling D. M., and Burns K. C.. 2014. “At a Loss for Birds: Insularity Increases Asymmetry in Seed‐Dispersal Networks.” Global Ecology and Biogeography 23: 385–394. [Google Scholar]
  59. Schleuning, M. , Fründ J., Klein A.‐M., et al. 2012. “Specialization of Mutualistic Interaction Networks Decreases Toward Tropical Latitudes.” Current Biology 22: 1925–1931. [DOI] [PubMed] [Google Scholar]
  60. Sentis, A. , Hemptinne J.‐L., and Brodeur J.. 2014. “Towards a Mechanistic Understanding of Temperature and Enrichment Effects on Species Interaction Strength, Omnivory and Food‐Web Structure.” Ecology Letters 17: 785–793. [DOI] [PubMed] [Google Scholar]
  61. Si, X. , Cadotte M. W., Davies T. J., et al. 2022. “Phylogenetic and Functional Clustering Illustrate the Roles of Adaptive Radiation and Dispersal Filtering in Jointly Shaping Late‐Quaternary Mammal Assemblages on Oceanic Islands.” Ecology Letters 25: 1250–1262. [DOI] [PubMed] [Google Scholar]
  62. Sinnott‐Armstrong, M. A. , Downie A. E., Federman S., Valido A., Jordano P., and Donoghue M. J.. 2018. “Global Geographic Patterns in the Colours and Sizes of Animal‐Dispersed Fruits.” Global Ecology and Biogeography 27: 1339–1351. [Google Scholar]
  63. Soares, F. C. , Leal A. I., Palmeirim J. M., and de Lima R. F.. 2021. “Niche Differences May Reduce Susceptibility to Competition Between Native and Non‐native Birds in Oceanic Islands.” Diversity and Distributions 27: 1507–1518. [Google Scholar]
  64. Sonne, J. , Vizentin‐Bugoni J., Maruyama P. K., et al. 2020. “Ecological Mechanisms Explaining Interactions Within Plant–Hummingbird Networks: Morphological Matching Increases Towards Lower Latitudes.” Proceedings of the Royal Society B: Biological Sciences 287: 20192873. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Stekhoven, D. J. , and Bühlmann P.. 2012. “MissForest—Non‐parametric Missing Value Imputation for Mixed‐Type Data.” Bioinformatics 28: 112–118. [DOI] [PubMed] [Google Scholar]
  66. Tobias, J. A. , Sheard C., Pigot A., Devenish A. J. M., Yang J., and Sayol F.. 2022. “AVONET: Morphological, Ecological and Geographical Data for all Birds.” Ecology Letters 25: 581–597. [DOI] [PubMed] [Google Scholar]
  67. Traveset, A. , Tur C., Trøjelsgaard K., et al. 2016. “Global Patterns of Mainland and Insular Pollination Networks.” Global Ecology and Biogeography 25: 880–890. [Google Scholar]
  68. Vázquez, D. P. , Blüthgen N., Cagnolo L., and Chacoff N. P.. 2009. “Uniting Pattern and Process in Plant–Animal Mutualistic Networks: A Review.” Annals of Botany 103: 1445–1457. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Vollstädt, M. G. R. , Galetti M., Kaiser‐Bunbury C. N., et al. 2022. “Plant–Frugivore Interactions Across the Caribbean Islands: Modularity, Invader Complexes and the Importance of Generalist Species.” Diversity and Distributions 28: 2361–2374. [Google Scholar]
  70. Wallace, A. R. 1878. Tropical Nature and Other Essays. UK, London: Macmillan & Co. [Google Scholar]
  71. Wheelwright, N. T. 1985. “Fruit‐Size, Gape Width, and the Diets of Fruit‐Eating Birds.” Ecology 66: 808–818. [Google Scholar]
  72. Whittaker, R. J. , Fernández‐Palacios J. M., Matthews T. J., Borregaard M. K., and Triantis K. A.. 2017. “Island Biogeography: Taking the Long View of nature's Laboratories.” Science 357: eaam8326. [DOI] [PubMed] [Google Scholar]
  73. Wilman, H. , Belmaker J., Simpson J., de la Rosa C., Rivadeneira M. M., and Jetz W.. 2014. “EltonTraits 1.0: Species‐Level Foraging Attributes of the world's Birds and Mammals.” Ecology 95: 2027. [Google Scholar]
  74. Zheng, C. , Jia L., and Hu G.. 2022. “Global Land Surface Evapotranspiration Monitoring by ETMonitor Model Driven by Multi‐Source Satellite Earth Observations.” Journal of Hydrology 613: 128444. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Data S1.

ELE-28-0-s001.docx (3MB, docx)

Data Availability Statement

Data and codes are available at the figshare repository (https://doi.org/10.6084/m9.figshare.25451611).

Articles from Ecology Letters are provided here courtesy of Wiley

RESOURCES