Environmental gradients structure assemblages of Canidae across the planet

doi:10.21203/rs.3.rs-4473904/v1

Download PDF

Research Article

Environmental gradients structure assemblages of Canidae across the planet

https://doi.org/10.21203/rs.3.rs-4473904/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The phylogenetic structure of ecological communities carries the signature of ecological processes that influenced and are still influencing their assembly. However, identifying the mechanisms that shape communities is not simple, as they can vary geographically. Here, we investigate how the phylogenetic structure of Canidae communities across the globe is affected by the abiotic and biotic environment. We first identify phylogenetically clustered and overdispersed assemblages of canids over the planet. Then, we apply Structural Equation Models in these communities in order to identify the effect of six variables (current temperature, Last Glacial Maximum temperature, vegetation cover, human impact, Felidae richness, and a measure of canid body size dissimilarity) on the phylogenetic relatedness of canids. We found that South America and Asia present a high concentration of clustered communities, whereas Central America, Europe, and North America show phylogenetically overdispersed assemblages. Temperature from the Last Glacial Maximum is the most important variable in our models, indicating that as LGM temperature increases, assemblages become less overdispersed (clustered). Canidae community composition across the world thus presents patterns of clustering and overdispersion, which follow mainly the environmental gradient, suggesting habitat filtering as the main force acting on Canidae assemblages.

Community assembly

community phylogenetics

habitat filtering

NRI

overdispersion

phylogenetic clustering

The assembly of an ecological community is influenced by several biotic and abiotic factors (Webb 2000; Webb et al. 2002b; Westoby 2006), including competition and habitat filtering. On the one hand, the competitive exclusion principle predicts that ecologically similar species cannot coexist if resources are limiting, and under phylogenetic trait conservatism, this implies that communities will be phylogenetically overdispersed (Darwin 1859; Elton 1946; Leibold 1998; Webb et al. 2002b). On the other hand, phylogenetically closely related species likely share traits that allow them to tolerate a specific environment (Jarvinen 1982; Weiher et al. 1998; Webb et al. 2002b; Di Marco and Santini 2015), which implies phylogenetic clustering of species in an ecological community. However, the same clustered pattern can be generated by competition if there is filtering for traits involved in competition (e.g. tree height) (Mayfield and Levine 2010; HilleRisLambers et al. 2012). If one can rule out this possibility, patterns of phylogenetic overdispersion or clustering are indicative of competition and habitat filtering, respectively, if the relevant traits are phylogenetically conserved. If traits are not conserved but convergent, habitat filtering can lead to phylogenetic overdispersion, while competition is expected to lead to random dispersion (Webb 2000; Webb et al. 2002; Cavender-Bares et al. 2004; Kraft et al. 2007). Here we remark that Pigot & Etienne (2015) showed that the speciation process itself already creates a somewhat overdispersed pattern. Hence, absolute values for phylogenetic dispersion are difficult to interpret.

However, relative dispersion values can still be informative: one can still compare dispersion patterns between communities and ask why some communities are more overdispersed or clustered than others. Many studies have used the phylogenetic approach described by Webb (2000) and Webb et al. (2002) to understand how the phylogenetic information of clades is structured through space (Cavender-Bares et al. 2004; Helmus et al. 2007; Kress et al. 2009; Kraft and Ackerly 2010; Goberna et al. 2014; Yang et al. 2014; Miazaki et al. 2015; Cadotte and Tucker 2017; Pérez-Valera et al. 2017; Zhang et al. 2018; Kusumoto et al. 2019). The majority of these studies have shown that overdispersion dominates at small scales, while clustering better explains the structure of communities at large scales. These findings demonstrate that the interpretation of the mechanisms acting in a community is scale-dependent (Webb et al. 2008). In this light, it is interesting to note that a large number of studies on phylogenetic structuring of communities are at small scales (see Cardillo et al. 2008, Gittleman, & Purvis, 2008).

Another difficulty in studies on the phylogenetic structure of communities relates to how abiotic and biotic factors are treated as independent forces acting on a community. Although habitat filtering and competition are contrasted in their effect on community structure, they occur together in natural communities (Ackerly 2003; Cadotte and Tucker 2017). Over the last decade, several studies have attempted to understand how much each mechanism influences community structure (Kraft and Ackerly 2010; Goberna et al. 2014; Cadotte and Tucker 2017; Pérez-Valera et al. 2017; Zhang et al. 2018; Kusumoto et al. 2019). The results have been inconclusive. Cadotte & Tucker (2017) noted that “it is likely that most observational data reported as evidence for environmental filtering, in fact, reflect the combined effects of the environment and local competition.” To overcome these issues, one needs a detailed analysis of whether traits are conserved or not on a phylogenetic tree and of how patterns of phylogenetic dispersion vary with environmental variables.

To explore how the structure of assemblages is related to biotic and abiotic factors, one needs a well-resolved phylogenetic tree and biotic and abiotic factors that vary across continental geographic scales. Relatively few taxonomic groups satisfy these conditions and one of them is Canidae. Canids have a worldwide distribution, found in all continents, except in Antarctica (Wang and Tedford 2008; Wilson and Mittermeier 2009) and are hence exposed to a variety of abiotic factors. They also have a well-resolved phylogenetic tree (Porto et al., 2019). Among the extant families of carnivores, modern canids are the only family to have a truly worldwide distribution, and interact with several other species (Wang and Tedford 2008; Fleming et al. 2017). Many canids have distributions that span at least a whole continent. Red foxes (Vulpes vulpes) and grey wolves (Canis lupus) have the most extensive natural range of any land mammal (with the exception of humans), inhabiting the whole Northern Hemisphere, and being present in 81% of countries around the world (Sillero-Zubiri et al. 2004).

Understanding how assemblages of canids are structured is intriguing, because, as Porto et al. (2023) showed, the extant canids on Earth originated from three major diversification events, each one in a distinct continent. Thus, one may wonder if assemblages of canids are regulated by completely distinct biotic and abiotic forces based on the great ecomorphological variation that these three major events of diversification generated. Surprisingly, to our knowledge, there is no study of canids that addresses this topic, despite all the potential that this group presents to elucidate some of the mechanisms that are responsible for the maintenance of biodiversity.

Here, our goal is to explore how patterns of phylogenetic dispersion are structured around the globe, tackling the methodological issues mentioned above and focusing on the family Canidae. Using the data available for canids, together with a set of current biotic and abiotic variables, we demonstrate how the geographic distribution of Canidae today can be explained by environmental filters and competition forces.

2.1 Data compilation

Canids’ range maps were compiled from the IUCN Red List for all the 36 species (IUCN 2019). We processed these maps using ArcGIS (ESRI, 2019) to generate a presence-absence matrix of species within defined grids cells of 2º (hereafter, assemblages). Given the global scale of our study, degrees could represent a much smaller area near the poles than at the Equator, which could bias our results. Thus, to create a presence-absence matrix of canids, the projection used was the Lambert Azimuthal Equal Area (LAEA) (Snyder 1987). The LAEA projection maintains land features at their true relative sizes while simultaneously preserving a true sense of direction from the center of the globe (Snyder 1987). Additionally, we utilized the R package crsuggest 0.4 (Walker 2022), which suggests the best projection to be used in the set of maps we are working with. In our case, the LAEA projection was also indicated as the most appropriate.

We used 2º grid cells to maximize coexistence among canids, as we have a small diversity of extant canids around the planet, and few of them present geographical overlap. The 2º grid cells used here is the ideal scale in our case, as values smaller than this were indicated by Hurlbert and Jetz (2007) to result in an overestimation of the area of occupancy of individual species and mischaracterize spatial patterns of species richness. We opted against using a larger resolution (e.g., 3º grid cells) to avoid overestimating the overlap of canid species.

For each grid cell, we extracted average values for four environmental variables (mean current temperature, mean temperature during the Last Glacial Maximum, vegetation cover, and human footprint) (Center for International Earth Science Information Network 2005; Fick and Hijmans 2017; USGS 2019; Karger et al. 2021). Temperature, measured as the average for the years 1970–2000 (Fick & Hijmans, 2017), is an important environmental gradient at the global scale, and can strongly affect food-web structure and interaction strength of species (Bideault et al. 2019). In addition, temperature is one of the most commonly used variables in studies that evaluate species distributions (Porfirio et al. 2014). Temperature during the Last Glacial Maximum (LGM) was downloaded from CHELSA database (Karger et al. 2021). LGM temperature is used here as a measure of past abiotic factors influencing the distribution of canids. Vegetation cover was obtained from NDVI (Normalized Difference Vegetation Index) and is characterized as the percentage of global vegetation cover of the year 2019 (USGS 2019). This variable has been suggested to play a major role in species geographical distributions (Reino et al. 2013; Vasconcelos and Doro 2016). Human footprint, which consists of estimates of human population for the year 2015 (Center for International Earth Science Information Network, 2005), is included in the models following Di Marco & Santini (2015), who found that human impacts explain the distribution of terrestrial mammals better than biological traits. All the environmental variables we use here present a 2.5-minute spatial resolution (~ 4.5 km at the Equator). We also use two biotic variables, Canidae body size (head + body length) and Felidae richness within assemblages. Canid body sizes were obtained from the females of each species using the Handbook of the Mammals of the World (Wilson and Mittermeier 2009) and also from the Animal Diversity Web (ADW) (Myers et al. 2018) (Table S1). We use only the female body size to standardize the data due to sexual dimorphisms that a few species of canids present. The presence of felids within grid cells was obtained using range maps from the IUCN Red List (IUCN, 2019).

Other environmental variables are not taken into account for various reasons. Annual precipitation and primary productivity (Fick and Hijmans 2017) present, within the assemblages addressed here (grid cells), high correlation with vegetation cover and temperature, respectively (Figure S1), which we already used in our models. The elevation gradient, which is commonly used in studies for different groups, is not considered here because only two species of Canidae have a large part of their distributions in mountains (Lycalopex culpaeus and Canis simensis) (Wilson and Mittermeier 2009), so it is of little value in our study. Furthermore, including many more abiotic variables than biotic ones may bias the results toward environmental filtering. In general, one runs the risk of overfitting, when too many variables are included in the model (Hawkins 2004; Babyak 2004).

We then calculated, for each grid cell, the body size dissimilarity among the canids within each assemblage using the decoupled trait approach (dcFdist metric) proposed by De Bello et al. (2017). This approach calculates functional differences between species accounting for the shared evolutionary history. Thus, we can measure how different canids within communities are, based on their body size, while accounting for their shared ancestry. Analyses were performed in R 4.0–2. (R Development Core Team 2023) using the raster package 3.3–13 (Hijmans and Etten 2012) and the scripts for the dcFdist metric made available in the supplementary material of De Bello et al. (2017) paper.

The number of felid species present in each community is used as an additional explanatory variable as a measure of interaction with canids because some studies have suggested that Canidae and Felidae underwent intense episodes of competition over the last 12 Myr (Wang and Tedford 2008; Silvestro et al. 2015).

Evidently, community assembly is a process that requires time and hence our explanatory variables should ideally reflect past conditions. Yet, such historical data are not available at the scale we considered, and hence, all of our explanatory variables except LGM temperature are based on present-day data. Our analyses therefore largely assume that either 1) the time scale of community assembly is relatively short compared to the time scale of change of these variables, such that current values of the variables are representative, or 2) the current data are highly correlated with the historical data (Figure S1C).

2.2 Phylogenetic data analyses

We used the framework developed by Webb (2000) and Webb et al. (2002) to describe how communities can be tested for phylogenetic overdispersion or clustering by comparing the value of a community metric to that of a null distribution of randomized communities. One such metric is the net relatedness index (NRI), where negative values of NRI are indicative of overdispersion, while positive values indicate clustering (Webb, 2000; Webb et al., 2002).

The phylogeny presented by Porto et al. (2019), that includes all 36 extant canid species and a recently extinct one, is used here to obtain the NRI values. The phylogeny was constructed through Bayesian inference based on 23 genes and 68 osteological characters. We removed the extinct species Dusicyon australis from this tree because there is no environmental data available for the region where the species lived before the year of its extinction (1876). To investigate how body size is distributed over the phylogeny, we calculate its phylogenetic signal using the K-statistic (Blomberg and Garland 2003) with the R package Phytools 0.7–47 (Revell 2012). This statistic compares phylogenetic signal with that of a Brownian Motion (BM) model of trait evolution, where a trait value changes randomly, in both direction and distance, over time; K < 1 indicates that closely related species resemble each other less than expected under the BM model, whereas K > 1 means that closely related species are more similar than predicted by the BM model.

To analyze phylogenetic dispersion patterns over the planet, we first calculated the standardized effect size of mean pairwise distances in communities (MPD) for each assemblage using the package Picante 1.8-2 (Kembel et al. 2010). Species richness spatial gradient and distributions are shown in Figures S2 and S3, respectively.

To determine whether the observed MPD differs from that expected by chance, we implemented a null model that randomizes the community data matrix 999 times by drawing canids from the pool of species occurring in at least one community. Basically, our null model shuffles taxa labels and re-calculates MPD, keeping richness intact. The 999 null values constitute a null distribution to which the observed value was compared. For the null model, we use the global species pool (n = 36) rather than delimiting regional pools due to the low number of canids in distinct regions of the planet (e.g. South America = 10, North America = 9, Africa = 11, and Eurasia = 11), which would limit the number of different potential null communities (but see section on Sensitivity analyses). For example, for a regional pool of 12 species there are only 66 different combinations for a species pool of 2, and 220 for a community of 3 species, 495 for a community of 4 (the formula is n! / (r! *(n - r)! where n is the number of species in the regional pool and r is the number of species in the community). We then calculate the net relatedness index (NRI) from MPD as follows:

$$NRI=-1\times \frac{Observed MPD – mean\left(Expected MPD\right)}{SD\left(Expected MPD\right)}$$

Larger (positive) NRI values imply a more phylogenetically clustered assemblage while smaller (negative) NRI values indicate more phylogenetic overdispersion. We choose the MPD to measure the relatedness of canids because it is probably the most popular measure for computing the phylogenetic distance between a given group of species which makes it easier for our results to be compared with the findings from other biological groups.

The use of random permutation null models for testing for patterns of phylogenetic overdispersion has been criticized (Pigot & Etienne, 2015), but we note here that we do not use our NRI values for an absolute test whether there is phylogenetic clustering or overdispersion. We only look at relative values of NRI to see how they vary across the globe. Our method, in contrast to classical assemblage-based methods, explicitly takes into account phylogenetic relationships between species as it has been shown that ignoring such relationships can lead to different conclusions (Tamagnini et al. 2021).

2.3 Structural equation models

To test the influence of biotic and abiotic variables on NRI patterns we used Structural Equation Models (SEM) (Mitchel 1992; Wang et al. 2013). SEM allow testing the contribution of different variables while accounting for potential correlations between them. We do not specify a priori whether any of these correlations had positive or negative effects. Our predictor variables (observed variables) are mean current temperature, mean temperature at the LGM, vegetation cover, human footprint, Felidae richness, and body size dissimilarity. We assumed that all predictor variables affect NRI values and that there is a latent variable connected to our six observed ones to explain the covariation in the measurements. We tested six models (Fig. 1), which differ as follows. Model 1 assumes that the human footprint has an impact on global temperature, which in turn affects vegetation cover. Model 2 is similar to model 1, but the human footprint affects vegetation cover, and the temperature is influenced by both variables. Neither of these two models have the influence of environmental variables on body size, but Models 3 and 4 do. In Model 3, body size dissimilarity is assumed to be influenced by vegetation and temperature, because higher values of these variables indicate more resources and hence may affect the level of competition which in turn can drive body size dissimilarity. Vegetation influences temperature in areas with high human density, but temperature also influences vegetation cover, for example, higher temperatures are associated with tropical forests. In Model 4, human footprint is assumed to have an effect on temperature and vegetation due to urbanization, and different from Model 3, there is no influence of vegetation on temperature. Model 4 also assumes that temperature of the LGM influences body size dissimilarity in canids. Model 5 presents the influence of human footprint and vegetation cover over Felidae richness and assumes that the temperature of the LGM influences vegetation cover. In Model 6, the temperature of the LGM influences Felidae richness, which in turn influences body size dissimilarity in canids.

We conducted SEM analyses through maximum-likelihood estimation using the R package lavaan 0.6-7 (Rosseel 2012), and the models are compared using the Akaike information criterion (AIC) (Akaike 1973). Spatial autocorrelation is taken into account during the analyses using the package semTools 0.5-3 through the function “spatialCorrect” (Jorgensen et al. 2020). During the SEMs, the function takes a weighted matrix that reflects the intensity of the geographic relationship between observations and calculates Moran's I for the residuals of all variables. If there is spatial autocorrelation in our dataset, the function spatially corrects it via Moran's I.

2.4 Sensitivity analyses

To test the sensitivity of our models to different ways that our data could have been structured, we conducted four different sensitivity analyses for comparison with our main results. The first two tests involve applying the same Structural Equation Models to communities with three or two co-occurring canids and communities with four or more canids. Subsequently, we compared the models of both tests using AIC values to check if different sizes of Canidae assemblages can affect our findings.

Because SEM analyses on all assemblages (phylogenetically overdispersed and clustered) simultaneously could mask some effect related to the predictor variables, we did a third sensitivity test splitting assemblages with positive and negative NRI values and conducted the SEM separately on these assemblages.

The fourth sensitivity test we used aims to evaluate how our results are influenced by the species pool of our null model. As distributions of canids are not global, and some species are geographically restricted, historical processes (i.e. speciation and dispersal limitation) could influence our results if using geographically unrestricted null model. Thus, we set a null model with regional pools (instead of a global pool). For this step, we obtained NRI values splitting the world into 4 regional pools (South America = 10 species, North + Central America = 9 species, Africa = 12 species, and Eurasia = 11 species); only few canids inhabit more than one continent (N. + C. America + Eurasia: 2 species, Africa + Eurasia: 2 species, N. + C America + Eurasia + Africa: 1 species).

3.1 NRI patterns and SEM

Phylogenetic signal in body size is high across the phylogeny of canids (K = 1.31, p < 0.01). The NRI values distributed over the world show that there are phylogenetically overdispersed and clustered communities in different continents (Fig. 2A). South America and Asia generally present high positive NRI values, whereas Central America, Europe, a large part of North America, and some regions in Asia show negative NRI values. Mean temperature, temperature at the LGM, vegetation cover, human footprint, Felidae richness, and body size dissimilarity also vary considerably across the world (Figs. 2B − 2G).

Structural Equation Model 4 is selected as the best model, more than 4 AIC units better than the second-best model, model 3 (Table 1). The weight of model 4 is 0.742. Model 3 presents a weight of 0.257. Models 1 and 2, with no influence of environmental variables on body size dissimilarity, perform poorly (weights much less than 0.0001). In Model 4, all predictor variables influence NRI values, but also human footprint influences vegetation and temperature, which in turn influences vegetation cover and body size dissimilarity. This model also assumes vegetation cover and the temperature at the LGM influence body size dissimilarity.

Table 1

Akaike information criterion ranking of the six Structural Equation Models tested for NRI values.
Models	df	AIC	ΔAIC	Weights
Model 1	12	38349.82	7936.36	0
Model 2	12	38346.82	7933.36	0
Model 3	16	30417.59	4.12	0.257
Model 4	15	30413.46	0	0.742
Model 5	18	43761.66	13348.19	0
Model 6	19	42578.58	12165.11	0

For NRI data (Fig. 3), model 4 shows that both temperature variables, from the LGM (p < 0.01, effect size = 0.345) and current mean temperature (p < 0.01, effect size = -0.261) are the most important to explain the spatial distribution of phylogenetic structure of canids (Table 2). The other two abiotic variables, vegetation cover and human footprint, present weak effects (effect size = 0.078 and − 0.074, respectively) on NRI values. Model 4 shows that body size dissimilarity is strongly influenced by environmental variables, such as current mean temperature (p < 0.01, effect size = -0.106) and vegetation cover (p < 0.01, effect size = -0.533). Both biotic variables (Felidae richness and body size dissimilarity), have significant effects on NRI. The strong and positive effect of LGM temperature on NRI values suggests that higher temperatures lead to higher NRI values, that is, more clustered communities.

Table 2

Results of the best-fitting Structural Equation Model (model 4) for NRI values indicating the effect size among biotic and abiotic variables.
Regressions		Estimate	Std. error	p	Effect
NRI ~	Body size dissimilarity	1.985	0.159	< 0.01	0.225
	Human footprint	-0.008	0.002	< 0.01	-0.074
	Temperature	-0.014	0.001	< 0.01	-0.261
	Vegetation cover	0.080	0.018	< 0.01	0.078
	Felidae richness	0.071	0.007	< 0.01	0.216
	LGM temperature	0.009	0.001	< 0.01	0.345
Temperature ~	Human footprint	0.754	0.035	< 0.01	0.353
Vegetation cover ~	Human footprint	0.114	0.002	< 0.01	0.129
Vegetation cover ~	Temperature	0.006	0.001	< 0.01	0.114
Body size dissimilarity ~	Temperature	-0.001	0	< 0.01	-0.106
	Vegetation cover	-0.062	0.002	< 0.01	-0.533
	LGM temperature	0	0	< 0.01	0.085
* P of the six models (< 0.01)

3.2 Sensitivity analyses

The results obtained throughout our sensitivity analyses indicate that our main findings are consistent and robust across various scenarios. Our first two sensitivity analyses also indicate model 4 as the best model for both assemblages’ sizes (with three or less species and with four or more canids) (Tables S2 and S3). For assemblages with three or less canids, environmental variables are still more important than biotic ones to explain phylogenetic distribution (Table S4), but vegetation cover presents the highest effect on NRI values in these communities (p < 0.01, effect size = 0.588). For more species-rich assemblages, with four or more species, both temperature variables present a strong effect on Canidae (Table S5), similar to our main analysis. In addition, the combined effect of abiotic variables is larger than that of biotic ones in our first two sensitivity analyses.

The third sensitivity test, splitting assemblages with positive and negative NRI values, presents similar results compared to our main analysis (with overall values). Model 4 is chosen as the best model for positive NRI assemblages (Table S6). Compared to our main results, in positive NRI assemblages we find a larger effect of current and LGM temperature on NRI values (effect size = -0.465 and 0.486, respectively), but the most important variable to explain NRI distribution still is LGM temperature (Table S7 and Figure S4). In assemblages with negative NRI values, model 4 is also the best model among the six tested (Table S8). In these assemblages, both temperature variables, current and LGM, are still the most important to explain NRI (effect size = -0.556 and 0.344, respectively), and Felidae richness presents a positive effect on NRI, similar to our main findings (effect size = 0.286) (Table S9). For negative NRI assemblages, abiotic variables are still more important than biotic ones to explain phylogenetic composition across the globe, as in our main analysis (Figure S5). Even though LGM and current temperature presented the strongest effects on NRI, current temperature presents a larger effect than LGM temperature (effect size = -0.556 and 0.344, respectively).

Our fourth sensitivity test, using regional pools, presents very similar results compared to our main analysis using a global pool. Model 4 is the model that performed better among the six models tested here (Table S10). Compared to our main results, regional pools yield very similar effects of biotic and abiotic variables on NRI values (Table S11 and Figure S6). Besides, NRI values obtained by a global pool or a regional pool present a strong association (R = 0.66) (Figure S7). Thus, for the 36 canids used here, the choice of pool does not matter that much as the main findings remain the same.

The composition of canid communities across the world shows both negative and positive values of NRI and is influenced by distinct factors across the world. However, environmental filters play a greater role than traits related to biotic factors (body size dissimilarity and Felidae richness) to explain the phylogenetic composition of assemblages. This pattern is consistent as it is supported not only for our main dataset, but also for different phylogenetic and spatial scales through our sensitivity analyses.

Temperature during the Last Glacial Maximum is the main force that shapes how Canidae lineages are phylogenetically distributed today. This result is in line with studies that suggested glacial oscillations as an important factor to contemporary biodiversity patterns through environmental filtering (Svenning et al. 2015a). Regional extinctions might have been triggered by paleoclimate shifts and limitations in species dispersal events, leaving behind lasting paleoclimate legacies that continue to impact modern biodiversity (Svenning et al. 2015b).

If oscillations in temperature over large time scales were indeed crucial for the distribution of Canidae, one would expect the current mean temperature to also have played a significant role. Indeed, our findings supported this notion, as the current mean temperature emerged as the second strongest force acting on NRI, but in the opposite direction to that of LGM temperature. The influence of LGM on the current distribution of canids can be elucidated by considering several key factors. Firstly, it is known that forests in Africa and South America, two major hotspot areas for Canidae (Porto et al. 2023), significantly contracted their areas due to global aridification during LGM (Clapperton 1993; Cowling et al. 2008a; Novello et al. 2019). Global aridification likely concentrated Canidae lineages into climatic refugia, fostering closer phylogenetic relationships into warm regions that offered suitable environmental conditions. This is in line with Carnaval and Moritz (2008), who proposed, through the use of paleoclimatic models, the presence of historical forest refugia in the Atlantic rain forest through the Pleistocene in South America. In contrast, the negative effect of current mean temperature suggests a different dynamic, possibly driven by the expansion and dispersal of species in response to contemporary climate changes.

These contrasting effects of temperature during LGM and current times may have influenced Canidae distribution in distinct ways, perhaps on different continents with varying degrees of overdispersion. This nuanced interplay underscores the complex and varied impacts of Quaternary climatic oscillations on the phylogenetic structure of Canidae worldwide. In fact, Quaternary major oscillations have been suggested to drive aspects of biodiversity such as species endemism and functional structures of communities at continental scales (Sandel et al. 2011; Brown et al. 2017).

However, we can also argue that Canids, as carnivores, are intrinsically linked to their prey (Vettorazzi et al. 2022), and this relationship could have played a prominent role in their distribution patterns (i.e. canid distributions are determined by a biotic factor, namely their prey, but the prey distributions in turn are determined by abiotic factors). As temperatures dropped and ice sheets advanced during the LGM, many herbivorous species sought refuge in forested areas (Cowling et al. 2008b). Canids might have naturally followed the distribution of their prey into these forested refugia, which is in line with the findings of Rowan et al. (2016) that demonstrated the distribution of Ungulate communities were strongly influenced by paleoclimate in Africa during the LGM, which in turn seems to have been followed by the distributions of carnivores.

The positive relationship between the LGM and NRI indicates that higher temperatures led to increased phylogenetic clustering. This observation strongly suggests the presence of environmental filtering, particularly prevalent in South America and Africa. It is likely that the drastic changes in vegetation in these regions created a robust filter that dictated which canid species could inhabit the remaining refuge areas. Furthermore, we highlight the negative effect that vegetation cover has on body size dissimilarity, showing that as areas become more forested, species turn out to be more similar in their sizes (less dissimilarity).

One can argue that body size dissimilarity, while not necessarily stronger than abiotic variables, plays an interesting role in structuring Canidae phylogenetic composition. When examining the distribution of this variable on the map (Fig. 2D), most of its variation occurs in the Northern Hemisphere, a region known to have undergone significant aridification during the LGM (Clark 1995). Therefore, in these areas, is it probable that ecological refuges may not have been as abundant, and competitive exclusion likely played a more pivotal role. In this sense, we can consider body size dissimilarity among current species as a proxy for past competition. Finarelli (2007) suggests that trends in body size were shaped by interactions between species throughout the evolutionary history of Canidae. Consequently, it is reasonable to assume that competition was most intense between species of similar body sizes, and if body size is evolutionarily conserved (as assumed here), competition led to the displacement of phylogenetically related species. But in the end, abiotic factors together were still the most important.

The observed relationship between Felidae richness and the assembly composition of canids unveils intriguing ecological dynamics, potentially rooted in several compelling explanations. Firstly, interspecific competition emerges as a potential force shaping these communities, wherein areas abundant in felids may exert stronger selective pressures on canids. This heightened competition could propel convergent evolution among canid species, fostering phylogenetic similarity as a means to mitigate overlap in ecological niches. Such mechanism could be similar to the one already suggested by Pires et al. (2017), indicating an incumbent effect in the Bering Strait. Their findings point to Felidae imposing a filter over extinct Canidae lineages, allowing only one clade of canids to cross it to Eurasia. Moreover, the diversity and specialization of prey that would reduce the competition against felines, could direct canids towards comparable evolutionary paths, further accentuating the observed clustering. However, canids inhabiting assemblages with fewer felines could explore diverse niches, fostering phylogenetic overdispersion.

Human footprint presents a low effect on NRI values. This is surprising because Di Marco & Santini (2015) demonstrate that human impacts influence the geographical distribution of lineages of terrestrial mammals better than the biological traits of species. Anthropogenic impacts seem to be important to understand the spatial distribution of the phylogenetic information of Canidae, considering the reports of endemic canids in the Americas constantly losing habitat and being killed by humans (Hoffmann et al. 2011). We hypothesize that the strong influence of human footprint on assemblages, found by Di Marco & Santini (2015), was not corroborated in our study due to the lower number of species that were analyzed here compared to their paper, corresponding to only 0.73% of all the terrestrial mammals they took into account. Moreover, Di Marco & Santini (2015), different from our study, explored several groups of mammals together, and apparently canids do not follow the general rule they found for all terrestrial mammals.

Our sensitivity analyses performed here yield similar results compared to our main framework, highlighting the robust pattern of temperature variables (mainly from the LGM) strongly influencing the geographic distribution of Canidae phylogenetic information.

We emphasize the better performance that models with environmental variables influencing body size dissimilarity had compared to the ones that did not. This points towards environmental control over body size dissimilarity. Our findings suggest a pattern of habitat filtering prevalent in warmer regions across the globe, such as deserts and tropical forests, with increased competition more pronounced in colder climates. This outcome stands in contrast to what was proposed by Dobzhansky (1950), who argued that biotic factors are more limiting in the tropics, while abiotic conditions are more important at higher latitudes. However, our results may not be entirely surprising given that the earliest ancestors of extant canids likely exhibited a pack-hunting lifestyle with medium body sizes around 70 cm (Wang and Tedford 2008; Porto et al. 2019). Deserts and tropical forests may act as a strong filter to canids that still present such a lifestyle because food is scarce in deserts and pack hunting is difficult inside dense forests.

Even though canids are not dispersal limited, because they can travel long distances (Wang and Tedford 2008; Wilson and Mittermeier 2009), they may still tend to be found in higher numbers near to their center of diversification than far from it. This can also help understand the phylogenetic structure of the Middle East + Northern Africa, and South America, which, based on the fossil records and biogeographical models, had major diversification events of foxes and South American canids, respectively (Wang and Tedford 2008; Porto et al. 2023). The ecological refugia created for canids within these regions during the LGM might have generated species’ phylogenetic clustering.

The majority of studies of the phylogenetic structure within communities concerns plants or focus on small geographic scales (Cadotte and Tucker 2017). Nevertheless, some studies with vertebrates at large spatial scales have demonstrated contrasting patterns of phylogenetic composition. Cooper, Rodríguez, & Purvis, (2008), also using NRI values, found for the first time in mammals a widespread tendency of overdispersion in assemblages of New World monkeys, Australasian possums, and North American ground squirrels, and debate the importance of competitive exclusion in these communities. However, Yan et al. (2016), applying a distinct metric to understand the phylogenetic structure of assemblages, found phylogenetic clustering for Mammalia, Aves, Reptilia, and Amphibia from China. Cardillo (2011), using NRI values and distinct null models for each region studied, demonstrated an unstructured phylogenetic pattern on African carnivore assemblages. Here we have presented a case where both patterns are important to understand community composition across the planet, depending on the region studied.

Larger geographical scales are expected to generate more phylogenetic clustering than overdispersion because the rate of speciation increases with more space available (Losos and Schluter 2000) due to higher habitat heterogeneity (Kneitel and Chase 2004), and thus sister species are more likely to co-occur at larger scales. Even though we find many areas with phylogenetic clustering (Fig. 2A), there are also many areas with overdispersed patterns, indicating that the scale we use (2º grid cells) is appropriate.

Sandel (2018) suggests that, if there is substantial variation in species richness along the geographical gradient, the effects that biotic and abiotic factors have on community assembly should not be assessed by only comparing index values, because NRI may depend on species richness. However, this is not the case here, as the species richness spatial gradient for Canidae does not present considerable variation over the planet (Figure S2).

The mechanisms that influence the assembly of a community can act in complex ways (Ricklefs 1987, 2015). In this study we use both a phylogenetic approach and a method based on environmental variables and traits. We demonstrate that canid community composition across the world presents substantial patterns of clustering and overdispersion that follow mainly the environmental gradient, suggesting habitat filtering as the main force acting on Canidae assemblages.

Acknowledgements

This work was financially supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brazil (CAPES) and by the Netherlands Organization for Scientific Research (NWO-VICI grant awarded to R.S.E.).

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Open Research statement

All data used here are already published and publicly available, with those items properly cited in this submission. All data generated during this study are included in this published article.

Competing interests

We have no competing interests.

Funding

L.M.V.P. was supported by CAPES and by the University of Groningen. R. M. is supported by UFRGS, CAPES, and CNPq (406497/2018-4). R.S.E. is supported by the Dutch Research Council (NOW) through a VICI grant.

Authors' contributions

L.M.V.P. conceived and designed the study and analyses. L.M.V.P. performed the analyses. L.M.V.P. wrote the first draft of the manuscript. R.S.E. and R.M. commented on the methods and contributed to substantial revisions on the draft.

Ackerly DD (2003) Community Assembly, Niche Conservatism, and Adaptive Evolution in Changing Environments. Int J Plant Sci 164:S165–S184. https://doi.org/10.1086/368401
Akaike H (1973) Information Theory and an Extension of the Maximum Likelihood Principle. Springer, New York, NY
Babyak MA (2004) What You See May Not Be What You Get: A Brief, Nontechnical Introduction to Overfitting in Regression-Type Models. Psychosom Med 66:411–421. https://doi.org/10.1097/01.psy.0000127692.23278.a9
Bideault A, Loreau M, Gravel D (2019) Temperature Modifies Consumer-Resource Interaction Strength Through Its Effects on Biological Rates and Body Mass. Front Ecol Evol 7:. https://doi.org/10.3389/fevo.2019.00045
Blomberg SP, Garland T (2003) Testing for phylogenetic signal in comparative data : behavioral traits are more labile. Evolution (N Y) 57:717–745
Brown LE, Khamis K, Wilkes M, et al (2017) Functional diversity and community assembly of river invertebrates show globally consistent responses to decreasing glacier cover. Nat Ecol Evol 2:325–333. https://doi.org/10.1038/s41559-017-0426-x
Cadotte MW, Tucker CM (2017) Should Environmental Filtering be Abandoned? Trends Ecol Evol 32:429–437. https://doi.org/10.1016/j.tree.2017.03.004
Cardillo M (2011) Phylogenetic structure of mammal assemblages at large geographical scales: Linking phylogenetic community ecology with macroecology. Philos Trans R Soc B Biol Sci 366:2545–2553. https://doi.org/10.1098/rstb.2011.0021
Cardillo M, Gittleman JL, Purvis A (2008) Global patterns in the phylogenetic structure of island mammal assemblages. Proc R Soc B Biol Sci 275:1549–1556. https://doi.org/10.1098/rspb.2008.0262
Carnaval AC, Moritz C (2008) Historical climate modelling predicts patterns of current biodiversity in the Brazilian Atlantic forest. J Biogeogr 35:1187–1201. https://doi.org/10.1111/j.1365-2699.2007.01870.x
Cavender‐Bares J, Ackerly DD, Baum DA, Bazzaz FA (2004) Phylogenetic Overdispersion in Floridian Oak Communities. Am Nat 163:823–843
Center for International Earth Science Information Network (2005) Gridded Population of the World, Version 3 (GPWv3): Population Density Grid. https://sedac.ciesin.columbia.edu/data/set/gpw-v3-population-density
Clapperton CM (1993) Nature of environmental changes in South America at the Last Glacial Maximum. Palaeogeogr Palaeoclimatol Palaeoecol 101:189–208. https://doi.org/10.1016/0031-0182(93)90012-8
Clark DH (1995) Extent, timing, and climatic significance of latest Pleistocene and Holocene glaciation in the Sierra Nevada, California. Oak Ridge, TN (United States)
Cooper N, Rodríguez J, Purvis A (2008) A common tendency for phylogenetic overdispersion in mammalian assemblages. Proc R Soc B Biol Sci 275:2031–2037. https://doi.org/10.1098/rspb.2008.0420
Cowling SA, Cox PM, Jones CD, et al (2008a) Simulated glacial and interglacial vegetation across Africa: implications for species phylogenies and trans-African migration of plants and animals. Glob Chang Biol 14:827–840. https://doi.org/10.1111/j.1365-2486.2007.01524.x
Cowling SA, Cox PM, Jones CD, et al (2008b) Simulated glacial and interglacial vegetation across Africa: implications for species phylogenies and trans-African migration of plants and animals. Glob Chang Biol 14:827–840. https://doi.org/10.1111/j.1365-2486.2007.01524.x
Darwin C (1859) On the origin of species. John Murray, London, UK
De Bello F, Šmilauer P, Diniz-Filho JAF, et al (2017) Decoupling phylogenetic and functional diversity to reveal hidden signals in community assembly. Methods Ecol Evol 8:1200–1211. https://doi.org/10.1111/2041-210X.12735
Di Marco M, Santini L (2015) Human pressures predict species’ geographic range size better than biological traits. Glob Chang Biol 21:2169–2178. https://doi.org/10.1111/gcb.12834
Dobzhansky T (1950) Evolution in the tropics. Am Sci 38:209–211
Elton C (1946) Competition and the Structure of Ecological Communities. J Anim Ecol 15:54–68
ESRI (Environmental Systems Resource Institute) (2019) ArcGIS Desktop Help 10.2. Geostatistical Analyst.
Fick SE, Hijmans RJ (2017) New 1-km spatial resolution climate surfaces for global land areas. Int J Climatol
Finarelli JA (2007) Mechanisms behind Active Trends in Body Size Evolution of the Canidae (Carnivora: Mammalia). Am Nat 170:876–885. https://doi.org/10.1086/522846
Fleming PJS, Nolan H, Jackson SM, et al (2017) Roles for the Canidae in food webs reviewed: Where do they fit? Food Webs 12:14–34. https://doi.org/10.1016/j.fooweb.2017.03.001
Goberna M, Navarro-Cano JA, Valiente-Banuet A, et al (2014) Abiotic stress tolerance and competition-related traits underlie phylogenetic clustering in soil bacterial communities. Ecol Lett 17:1191–1201. https://doi.org/10.1111/ele.12341
Hawkins DM (2004) The Problem of Overfitting. J Chem Inf Comput Sci 44:1–12. https://doi.org/10.1021/ci0342472
Helmus MR, Bland TJ, Williams CK, Ives AR (2007) Helmus et al. 2007. American Naturalist.pdf. Am Nat 169:69–83
Hijmans RJ, Etten J van (2012) raster: Geographic analysis and modeling with raster data. R package version 2.0-12
HilleRisLambers J, Adler PB, Harpole WS, et al (2012) Rethinking Community Assembly through the Lens of Coexistence Theory. Annu Rev Ecol Evol Syst 43:227–248. https://doi.org/10.1146/annurev-ecolsys-110411-160411
Hoffmann M, Belant JL, Chanson JS, et al (2011) The changing fates of the world’s mammals. Philos Trans R Soc B Biol Sci 366:2598–2610. https://doi.org/10.1098/rstb.2011.0116
Hurlbert AH, Jetz W (2007) Species richness, hotspots, and the scale dependence of range maps in ecology and conservation. Proc Natl Acad Sci 104:13384–13389. https://doi.org/10.1073/pnas.0704469104
IUCN (2019) The IUCN Red List of Threatened Species. Version 2019-2. http://www.iucnredlist.org
Jarvinen O (1982) Species-To-Genus Ratios in Biogeography: A Historical Note. J Biogeogr 9:363–370
Jorgensen TD, Pornprasertmanit S, Schoemann AM, Rosseel Y (2020) semTools: Useful tools for structural equation modeling. R package version 0.5-3
Karger DN, Nobis MP, Normand S, et al (2021) CHELSA-TraCE21k v1. 0. Downscaled transient temperature and precipitation data since the last glacial maximum. Clim Past Discuss 1–27
Kembel SW, Cowan PD, Helmus MR, et al (2010) Picante: R tools for integrating phylogenies and ecology. Bioinformatics 26:1463–1464. https://doi.org/10.1093/bioinformatics/btq166
Kneitel JM, Chase JM (2004) Trade-offs in community ecology: linking spatial scales and species coexistence. Ecol Lett 7:69–80
Kraft NJB, Ackerly DD (2010) Functional trait and phylogenetic tests of community assembly across spatial scales in an Amazonian forest. Ecol Monogr 80:401–422. https://doi.org/10.1890/09-1672.1
Kraft NJB, Cornwell WK, Webb CO, et al (2007) Trait Evolution , Community Assembly, and the Phylogenetic Structure of Ecological Communities. Am Nat 170:271–283
Kress WJ, Erickson DL, Jones FA, et al (2009) Plant DNA barcodes and a community phylogeny of a tropical forest dynamics plot in Panama. Proc Natl Acad Sci U S A 106:18621–18626. https://doi.org/10.1073/pnas.0909820106
Kusumoto B, Villalobos F, Shiono T, Kubota Y (2019) Reconciling Darwin’s naturalization and pre‐adaptation hypotheses: An inference from phylogenetic fields of exotic plants in Japan. J Biogeogr 1–12. https://doi.org/10.1111/jbi.13683
Leibold MA (1998) Similarity and local co-existence of species in regional biotas. Evol Ecol 12:95–110
Losos JB, Schluter D (2000) Analysis of an evolutionary species–area relationship. Nature 408:847–850
Mayfield MM, Levine JM (2010) Opposing effects of competitive exclusion on the phylogenetic structure of communities. Ecol Lett 13:1085–1093. https://doi.org/10.1111/j.1461-0248.2010.01509.x
Miazaki AS, Gastauer M, Meira-Neto JAA (2015) Environmental severity promotes phylogenetic clustering in campo rupestre vegetation. Acta Bot Brasilica 29:561–566. https://doi.org/10.1590/0102-33062015abb0136
Mitchel RJ (1992) Testing Evolutionary and Ecological Hypotheses Using Path Analysis and Structural Equation Modelling Author. Funct Ecol 6:123–129
Myers P, Espinosa R, Parr CS, et al (2018) The Animal Diversity Web
Novello VF, Cruz FW, McGlue MM, et al (2019) Vegetation and environmental changes in tropical South America from the last glacial to the Holocene documented by multiple cave sediment proxies. Earth Planet Sci Lett 524:115717. https://doi.org/10.1016/j.epsl.2019.115717
Pérez-Valera E, Goberna M, Faust K, et al (2017) Fire modifies the phylogenetic structure of soil bacterial co-occurrence networks. Environ Microbiol 19:317–327. https://doi.org/10.1111/1462-2920.13609
Pigot AL, Etienne RS (2015) A new dynamic null model for phylogenetic community structure. Ecol Lett 18:153–163. https://doi.org/10.1111/ele.12395
Pires MM, Silvestro D, Quental TB (2017) Interactions within and between clades shaped the diversification of terrestrial carnivores. Evolution (N Y) 71:1855–1864
Porfirio LL, Harris RMB, Lefroy EC, et al (2014) Improving the use of species distribution models in conservation planning and management under climate change. PLoS One 9:1–21. https://doi.org/10.1371/journal.pone.0113749
Porto LMV, Maestri R, Duarte LDS (2019) Evolutionary relationships among life-history traits in Caninae (Mammalia: Carnivora). Biol J Linn Soc 128:. https://doi.org/10.1093/biolinnean/blz069
Porto LM V, Etienne RS, Maestri R (2023) Evolutionary radiation in canids following continental colonizations. Evolution (N Y) 77:971–979. https://doi.org/10.1093/evolut/qpad015
R Development Core Team (2023) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria.
Reino L, Beja P, Araújo MB, et al (2013) Does local habitat fragmentation affect large-scale distributions? The case of a specialist grassland bird. Divers Distrib 19:423–432. https://doi.org/10.1111/ddi.12019
Revell LJ (2012) phytools: an R package for phylogenetic comparative biology (and other things). Methods Ecol Evol 3:217–223. https://doi.org/10.1111/j.2041-210X.2011.00169.x
Ricklefs RE (1987) Community diversity: relative roles of local and regional processes. Science (80- ) 235:167–171
Ricklefs RE (2015) Intrinsic dynamics of the regional community. Ecol Lett 18:497–503
Rosseel Y (2012) lavaan: An R Package for Structural Equation Modeling. J Stat Softw 48:1–36
Rowan J, Kamilar JM, Beaudrot L, Reed KE (2016) Strong influence of palaeoclimate on the structure of modern African mammal communities. Proc R Soc B Biol Sci 283:20161207. https://doi.org/10.1098/rspb.2016.1207
Sandel B (2018) Richness-dependence of phylogenetic diversity indices. Ecography (Cop) 41:837–844. https://doi.org/10.1111/ecog.02967
Sandel B, Arge L, Dalsgaard B, et al (2011) The Influence of Late Quaternary Climate-Change Velocity on Species Endemism. Science (80- ) 334:660–664. https://doi.org/10.1126/science.1210173
Sillero-Zubiri C, Hoffmann M, Macdonald D (2004) Canids: Foxes, Wolves, Jackals and Dogs. Information Press, Oxford
Silvestro D, Antonelli A, Salamin N, Quental TB (2015) The role of clade competition in the diversification of North American canids. https://doi.org/10.1073/pnas.1502803112
Snyder JP (1987) Map Projections: A Working Manual. US Geol Surv Prof Pap Washington, DC United States Gov Print Off
Svenning J-C, Eiserhardt WL, Normand S, et al (2015a) The Influence of Paleoclimate on Present-Day Patterns in Biodiversity and Ecosystems. Annu Rev Ecol Evol Syst 46:551–572. https://doi.org/10.1146/annurev-ecolsys-112414-054314
Svenning J-C, Eiserhardt WL, Normand S, et al (2015b) The Influence of Paleoclimate on Present-Day Patterns in Biodiversity and Ecosystems. Annu Rev Ecol Evol Syst 46:551–572. https://doi.org/10.1146/annurev-ecolsys-112414-054314
Tamagnini D, Canestrelli D, Meloro C, et al (2021) New Avenues for Old Travellers: Phenotypic Evolutionary Trends Meet Morphodynamics, and Both Enter the Global Change Biology Era. Evol Biol 48:379–393. https://doi.org/10.1007/s11692-021-09545-x
USGS (2019) Global Land Cover Facility digital database Available. https://www.usgs.gov/
Vasconcelos TS, Doro JLP (2016) Assessing how habitat loss restricts the geographic range of Neotropical anurans. Ecol Res 31:913–921. https://doi.org/10.1007/s11284-016-1401-8
Vettorazzi M, Mogensen N, Kaelo B, Broekhuis F (2022) Understanding the effects of seasonal variation in prey availability on prey switching by large carnivores. J Zool 318:218–227. https://doi.org/10.1111/jzo.13013
Walker K (2022) crsuggest package 0.4
Wang IJ, Glor RE, Losos JB (2013) Quantifying the roles of ecology and geography in spatial genetic divergence. Ecol Lett 16:175–182. https://doi.org/10.1111/ele.12025
Wang X, Tedford RH (2008) Dogs: Their Fossil Relatives and Evolutionary History. Columbia University Press
Webb CO (2000) Exploring the phylogenetic structure of ecological communities: An example for rain forest trees. Am Nat 156:145–155. https://doi.org/10.1086/303378
Webb CO, Ackerly DD, Kembel SW (2008) Phylocom: Software for the analysis of phylogenetic community structure and trait evolution. Bioinformatics 24:2098–2100. https://doi.org/10.1093/bioinformatics/btn358
Webb CO, Ackerly DD, Mcpeek MA, et al (2002a) Phylogenies and Community Ecology. Annu Rev Ecol Syst 33:475–505
Webb CO, Ackerly DD, Mcpeek MA, Donoghue MJ (2002b) Phylogenies and Community Ecology. 33:475–505. https://doi.org/10.1146/annurev.ecolsys.33.010802.150448
Weiher E, Clarke GDP, Keddy PA (1998) Community assembly rules, morphological dispersion, of plant species the coexistence. Oikos 81:309–322
Westoby M (2006) Phylogenetic ecology at world scale, a new fusion between ecology and evolution. Ecology 87:163–165
Wilson DE, Mittermeier RA (2009) Handbook of the Mammals of the World – Volume 1. Lynx Edicions
Yan C, Xie Y, Li X, et al (2016) Species co-occurrence and phylogenetic structure of terrestrial vertebrates at regional scales. Glob Ecol Biogeogr 25:455–463. https://doi.org/10.1111/geb.12428
Yang J, Zhang G, Ci X, et al (2014) Functional and phylogenetic assembly in a Chinese tropical tree community across size classes, spatial scales and habitats. Funct Ecol 28:520–529. https://doi.org/10.1111/1365-2435.12176
Zhang Q, Goberna M, Liu Y, et al (2018) Competition and habitat filtering jointly explain phylogenetic structure of soil bacterial communities across elevational gradients. Environ Microbiol 20:2386–2396. https://doi.org/10.1111/1462-2920.14247

Supplementarymaterial.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Environmental gradients structure assemblages of Canidae across the planet

Status:

Version 1

Abstract

Figures

1. Introduction

2. Material and methods

2.1 Data compilation

2.2 Phylogenetic data analyses

2.3 Structural equation models

2.4 Sensitivity analyses

3. Results

3.1 NRI patterns and SEM

3.2 Sensitivity analyses

4. Discussion

Declarations

References

Supplementary Files

Status:

Version 1