Projecting consequences of global warming for the functional diversity of fleshy‐fruited plants and frugivorous birds along a tropical elevational gradient

Species in ecological communities are linked by biotic interactions. It is therefore important to simultaneously study the impacts of global warming on interdependent taxa from different trophic levels. Here, we quantify current and potential future associations of functional diversity (based on multiple traits) and functional identity (based on individual traits) between interacting taxa using projection models under climate change.

Species in ecological communities are linked by biotic interactions, such as trophic interactions between plants and animals (Bascompte & Jordano, 2007). Independent species' range shifts could cause spatial mismatches between interacting species if the interdependent species shift their ranges differently due to differences in their thermal tolerances or dispersal abilities (Schweiger, Settele, Kudrna, Klotz, & Kühn, 2008). Future spatial mismatches between currently interacting species might trigger secondary range contractions and extinctions if lost interaction partners cannot be replaced (Schleuning et al., 2016;Schweiger et al., 2008). Such indirect, interaction-mediated consequences of climate change have often been overlooked and should be considered when studying the potential impacts of climate change on biodiversity (Blois, Zarnetske, Fitzpatrick, & Finnegan, 2013).
According to the concept of ecological fitting (Janzen, 1985), trait matching between species determines who interacts with whom in ecological communities (Bender et al., 2018;Garibaldi et al., 2015). A widely documented example is the matching of bill length and corolla length in plant-hummingbird interactions (Maglianesi, Blüthgen, Böhning-Gaese, & Schleuning, 2014;. In plant-frugivore interactions, bill width and fruit width are closely related Moermond & Denslow, 1985;Wheelwright, 1985). At the community level, the variation and dominant value of such functional traits can be estimated by metrics of the functional diversity of multiple traits and the functional identity of single traits (Díaz et al., 2007;Gagic et al., 2015). Both, functional diversity and identity, have been shown to correspond between communities of interacting species. For instance, community means of the mouth length of pollinators and the nectar accessibility of their feeding plants improve predictions of plant reproduction (Garibaldi et al., 2015). Similarly, communities of fleshy-fruited plants and frugivorous birds show a high correspondence in their functional diversity and functional identity along elevational gradients Vollstädt et al., 2017). However, potential future range shifts of species in response to climate change might alter the composition of ecological communities (Graham, Weinstein, Supp, & Graham, 2017). Associated changes in the functional diversity and functional identity of ecological communities may trigger a disruption in the functional correspondence between interacting taxa. Therefore, integrative functional analyses are needed to identify ecological systems and regions that are prone to disruptions of functional correspondence between interacting taxa ( Figure 1).
Mountains are global biodiversity hot spots (Antonelli et al., 2018;Jetz, Rahbek, & Colwell, 2004) and are well suited to study functional associations between interacting taxa (Albrecht et al., 2018). Upslope range shifts of species in response to increasing temperatures might alter mountain biodiversity in distinct ways. The number of lowland species emigrating or going extinct from low elevations might exceed the number of persisting, warm-adapted species, resulting in a decline of diversity at low elevations (i.e., lowland biotic attrition; Colwell, Brehm, Cardelús, Gilman, & Longino, 2008).
Mid-elevations might experience elevated species turnover, due to immigrating lowland species and upslope shifts of species from midelevations (Corlett, 2011). At high elevations, diversity might accumulate as species from lower elevations immigrate to high elevations (Steinbauer et al., 2018), but high-elevation species might also face an increased extinction risk if their suitable environment contracts towards the mountaintop (Peters & Darling, 1985). So far, there is empirical evidence for upslope shifts of the lower and/or the upper elevational range limits for a variety of plant and animal taxa across the globe (Freeman, Lee-Yaw, Sunday, & Hargreaves, 2018;Freeman, Scholer, Ruiz-Gutierrez, & Fitzpatrick, 2018). Consequences of elevational range shifts for biodiversity may be particularly severe in the tropics, where latitudinal temperature gradients are shallow and elevational range shifts more likely (Colwell et al., 2008). Tropical mountains are, therefore, particularly suitable for studying the effect of climate change on interacting species.
Here, we propose a new integrative analysis of functional diversity and projection models ( Figure 1) to study potential future changes in the functional diversity of fleshy-fruited plant and frugivorous bird the future. Our approach of studying functional diversity of interacting taxa could be more widely applied to identify potential future mismatches between trophic levels.

K E Y W O R D S
biotic interactions, climate change, ecological communities, functional correspondence, global change, seed dispersal, trait matching, trophic interactions communities along an elevational gradient in the tropical Andes of southeast Peru. We investigate multi-trait functional diversity and single-trait functional identity of plant communities based on fruit dimensions, plant height and crop mass and those of bird communities based on bill dimensions, wing pointedness and body mass. These species characteristics are relevant for trait matching in plant-frugivore networks (Bender et al., 2018;; see further details in methods).
We consider three previously reported scenarios of how species might alter their elevational ranges under increasing temperatures (Freeman, Lee-Yaw, et al., 2018;Freeman, Scholer, et al., 2018;Figure 2a-c) and investigate changes in the projected functional correspondence of the two interacting species groups. We expect a decrease in functional diversity at low elevations under scenarios in which species shift their lower elevational range limit upslope (Colwell et al., 2008). Under scenarios in which species shift their upper elevational range limits upslope, we expect an increase in functional diversity at mid-and high elevations, because functionally more diverse species from low elevations might immigrate to mid-and high-elevation communities (Colwell F I G U R E 1 Projections of functional correspondence between interacting taxa from different trophic levels applying trait-based analyses. Shown are (a) hypothetical animal species (circles) and plant species (triangles) that co-occur in an ecological community. (b) Each species is defined by a set of functional traits that influence plant-animal interactions (e.g., fruit size and plant height and bill size and wing shape).  . We expect particularly pronounced changes in functional diversity of bird communities at mid-elevations because species turnover between neighbouring frugivorous bird communities is high at these elevations along the Manú gradient (Dehling, Fritz, et al., 2014). This might lead to a decoupling of plant and bird functional diversity and identity under future scenarios.
F I G U R E 2 Projected changes in (a-c) functional richness and (d-f) mean fruit and bill width for fleshy-fruited plant (dark grey) and frugivorous bird (medium grey) communities under RCP 8.5 for the year 2080. We applied three vertical dispersal scenarios for future projections of plant and bird communities along the Manú gradient, that is, range contraction, range expansion and range shift. Changes were computed as future functional richness minus current functional richness and future fruit and bill width minus current fruit and bill width, respectively. Given are mean values and standard deviation (error bars) over five general circulation models (CCSM4, HadGEM2-ES, MIROC 5, MRI-CGCM and NorESM). Dashed lines indicate low, mid-and high elevations. Functional identity of lowland communities could not be quantified under range contraction and shift scenarios because projected species richness was zero

| Study system
We studied communities of fleshy-fruited plants and frugivorous birds at seven elevations located every 500 m from 500 to 3,500 m a.s.l. in the Kosñipata valley in the Manú biosphere reserve in south-east Peru (hereafter "Manú gradient"). The entire study region ranges from approximately 250 to 3,750 m a.s.l. The forest along the Manú gradient comprises four main vegetation types (Patterson, Stotz, Solarit, Fitzpatrick, & Pacheco, 1998)

| Plant and bird communities
To identify fleshy-fruited plant species co-occurring along the  Figure S2). However, the plot data seemed to underestimate species' actual elevational range extents because about 80% of the analysed species were only recorded at a single elevation ( Figure   S1). Therefore, we decided to use the literature data that apparently provide a more comprehensive measure of the current elevational ranges of plants although this data source could slightly overestimate the current local ranges. We considered 392 plant species for which we had information on traits and elevational distribution in further analyses.
Information on co-occurring frugivorous bird species along the Manú gradient was compiled from local checklists (Merkord, 2010;Walker, Stotz, Pequeño, & Fitzpatrick, 2006) supplemented by field observations (Dehling, Fritz, et al., 2014;Dehling, Sevillano, & Morales, 2013). These occurrence data were compiled during repeated surveys along the Manú gradient over several years based on point counts, mist-netting or chance observations. If elevational ranges were derived from a country-wide guide book (Schulenberg, Stotz, Lane, O'Neill, & Parker, 2010), estimates of functional diversity of current bird communities were very similar to those obtained through the Manú checklist (Pearson r = 0.98, p < 0.001, n = 7 elevations; Figure S3). Because of its higher accuracy for the Manú gradient, we used range data from the Manú checklist. We considered all avian frugivore species that occur along the gradient, except ground-dwelling species (Tinamidae, Odontophoridae, Psophidae, Mitu). Ground-dwelling species have different fruit handling and foraging strategies than bird species that take fruit directly from the plant and their matching traits therefore differ compared to other guilds . Our final bird species pool included 217 frugivorous bird species.
We computed the current elevational range extent of each plant and bird species as the distance between its minimum and maximum elevational range limit (Tables A1 and A2 at DRYAD digital repository). Based on these data, we determined which plant and bird species occurred at each of the seven studied elevations (500-3,500 m a.s.l.) and on the lowest (250 m a.s.l.) and the highest elevations (3,750 m a.s.l.) of the Manú gradient. An elevational resolution of 500 m was appropriate for describing elevational turnover across this wide elevational gradient, as it corresponds to the resolution of the available distribution data and has been applied in previous work comparing patterns in plant and bird diversity along this gradient .

| Plant and bird traits
For all plant and bird species, we collected morphological traits that have been shown to determine interactions between plants and frugivores (Bender et al., 2018;. Fruit width (mm) and fruit length (mm) of plants correspond to bill width (mm) and bill length (mm) of birds and relate to the size matching of bill and fruit Moermond & Denslow, 1985;Wheelwright, 1985). Fruit crop mass (i.e., the fresh mass (g) of a single fruit multiplied by the number of ripe fruits per plant) corresponds to avian body mass (g) and represents the matching between food resource availability and avian energy requirements, that is, larger frugivores tend to feed on plants that offer a higher resource amount as this minimizes their foraging cost (Albrecht et al., 2018; (Dunning, 2007). In our analyses, we used species means of all traits (see Tables A3 and A4 at DRYAD digital repository for trait values per plant and bird species and Appendix S1 for details on the museum collection and identity of bird specimens).

| Functional diversity and identity
We computed species richness, functional diversity and functional identity for each current and potential future plant and bird community. Species richness equals the number of species that occur in a community at a given elevation. Functional diversity was estimated as functional richness (Villéger, Mason, & Mouillot, 2008).

| Projections of species' elevational ranges
We approximated the current realized temperature niche of a species by its current elevational range and considered three scenarios of how the current elevational ranges of species could change under global warming (i.e., future temperature increases). We assumed that projected changes in species' elevational ranges are primarily driven by projected changes in temperature. Observed elevational range shifts of Andean species in response to recent (Feeley et al., 2011;Forero-Medina et al., 2011) and historic climate change (Hansen, Seltzer, & Wright, 1994) support this assumption. At the Manú gradient, precipitation might be a less limiting factor for species' occurrences, as precipitation is currently high (Girardin et al., 2010(Girardin et al., , 2013Rapp & Silman, 2012) and projected to increase (Hijmans et al., 2005;Stocker et al., 2014). We further assumed that plant and bird species shift their elevational ranges only upslope because the Manú gradient is entirely covered by rainforest (Patterson et al., 1998).
We computed species-specific vertical distances as the mean vertical distance across all of the studied elevations at which a species currently occurs (La Sorte & Jetz, 2010). The species-specific vertical distance approximates the distance a species would have to move upslope in order to track its current realized temperature niche under future global warming. In this computation, we considered vertical distances from nine grid cells at a resolution of 2.5 min that capture the entire elevational range of the Manú gradient (250-3,750 m a.s.l.; see Tables A1 and A2 at DRYAD digital repository for current elevational ranges and species-specific vertical distances of plant and bird species).
We implemented three vertical dispersal scenarios corresponding to observed changes in species' ranges on mountains (Freeman, Lee-Yaw, et al., 2018;Freeman, Scholer, et al., 2018). First, under the range contraction scenario (Figure 2a), we assumed that species cannot persist under temperatures that exceed their current realized temperature niche (i.e., species cannot tolerate or adapt to higher temperatures) and that species are unable to shift their ranges upslope. Therefore, species' lower elevational range limits shift upslope, while species' upper elevational range limits remain unchanged. For this scenario, we added the species-specific vertical distance to the lower limit of each species' current elevational range. Second, under the range expansion scenario (Figure 2b), we assumed that species can tolerate or adapt to temperatures that exceed their current realized temperature niche, but also that species are able to shift their elevational ranges upslope to track their current realized temperature niche. Therefore, species' lower elevational range limits remain unchanged, while species' upper elevational range limits shift upslope.
For this scenario, we added the species-specific vertical distance

| Statistical analyses
We quantified projected changes in plant and bird functional diversity by subtracting the current functional diversity of a community from its projected functional diversity. We compared these projections to analogous projections of plant and bird species richness. Furthermore, we computed projected changes in the functional identity of plant and bird communities by subtracting the current mean trait value of a community from its projected mean trait value. Projected changes were computed separately for each global circulation model.
To test for a correspondence between the functional diversity and functional identity of plant and bird communities along elevation , we fitted linear regression models with the functional diversity or functional identity of the plant communities as predictor and the functional diversity or functional identity of the corresponding bird communities as response. We did this for the current situation and for the three dispersal scenarios. These models were based on functional diversity and identity values that were averaged across the five general circulation models.

| Current patterns of functional diversity and identity
Current plant species richness ranged from 276 (see Table A5 at DRYAD digital repository for species richness and functional diversity of plant communities at each elevation) at low elevations (500 m a.s.l.) to 76 species around the treeline (3,500 m a.s.l.). The current bird species richness decreased from 126 species (see Table A6 at showed largely similar patterns ( Figures S7-S9a).

| Projected changes in plant and bird functional richness
At low elevations, plant and bird functional diversity were projected to decrease under range contraction and range shift scenarios. If temperature increases by about 3-4°C, all fleshy-fruited plant and frugivorous bird species were projected to be lost from low elevations. Such a projected decline is in line with the hypothesis of lowland biotic attrition (Colwell et al., 2008). However, lowland biotic attrition might be mitigated by species from warmer regions that immigrate to lowland communities (Anderson et al., 2012). This requires the existence of such warm-adapted species (Colwell et al., 2008) and their ability to disperse to the respective regions (Anderson et al., 2012).
The strength of lowland biotic attrition will further depend on species' thermal tolerances (Feeley & Silman, 2010). Occurrence-derived thermal tolerances of tropical lowland species might underestimate thermal tolerances as species possibly tolerate higher temperatures than currently realized (Feeley & Silman, 2010). A global meta-analysis on past distributional changes in vascular plants and endo-and ectothermic animals indeed shows that most lowland species rather extend their ranges upslope than emigrating away from low elevations (Freeman, Lee-Yaw, et al., 2018). Furthermore, changes in the composition of lowland plant communities in France lag behind observed temperature increases between 1965 and 2008 (Bertrand et al., 2011). Nevertheless, a decrease in plant species richness was recently reported at low elevations in the Alps (Scherrer, Massy, Meier, Vittoz, & Guisan, 2017), suggesting that lowland biotic attrition is happening in some mountain areas.
At mid-elevations, functional richness increased under range expansion and, to a lesser extent, under range shift, indicating that species with functionally more extreme trait combinations were projected to immigrate to higher elevations. The current decline in functional richness with elevation suggests a filtering of species with extreme trait values, for example, large-billed and large-fruited species, towards high elevations (Dehling, Fritz, et al., 2014;. Our projections of upslope dispersal suggest that communities at high elevations might become functionally more diverse in the future. Empirical studies indeed report increasing plant species richness at mid-and high elevations under climate change in North America (Savage & Vellend, 2015;Sproull, Quigley, Sher, & González, 2015) and Europe (Steinbauer et al., 2018). If global warming continues, a likely scenario would be a gradual increase in functional diversity of plant and bird communities at mid-and high elevations of the Manú gradient and on other tropical mountains.   Jankowski et al., 2013) might explain the strong effect of range contraction on these bird communities. Accordingly, many montane bird species are projected to be threatened if they are vulnerable to increasing temperatures and limited in their vertical dispersal (La Sorte & Jetz, 2010). In current plant-frugivore communities across the Andes, certain functional types of frugivorous birds interact with certain functional types of fleshy-fruited plants (Bender et al., 2018).

| Potential future differences between plant and bird functional diversity and identity
For instance, large-beaked bird species mostly feed on large fruits, species with pointed wings tend to forage in the canopy and species with a large body mass prefer plants with a large crop mass Wheelwright, 1985). A higher plant than bird functional richness, as found at mid-elevations under the range contraction scenario, suggests that certain functional types of plants might lack seed dispersers with matching functional traits in future communities (Bender et al., 2018;. This is in line with the projected mismatch between fruit and bill width at 2,000 m, where mean bill width was projected to decrease in the future. At this elevation, the dispersal of particular plant species might decrease because the loss of large-bodied frugivores reduces, particularly, the dispersal of large-fruited plant species (Markl et al., 2012). The projected increase in avian bill width at high elevations might have less severe consequences as broad-billed bird species tend to be flexible in their fruit choice (Bender et al., 2017;Moermond & Denslow, 1985;Wheelwright, 1985). Generally, avian flexibility to switch to other fruit resources could mitigate the risk to lose seed-dispersal functions under climate change. In the long run, however, a mismatch between plant and bird traits would likely trigger evolutionary changes in plant communities, for example, at the expense of large-fruited plant species (Onstein et al., 2018).
Under range expansion, projected differences between plant and bird functional richness and identity were smallest. This suggests that losing seed-dispersal functions under climate change is less likely if species are able to shift their upper elevational range limits upslope. Accordingly, the projected range loss of montane bird species under climate change was least severe if species were able to disperse vertically (La Sorte & Jetz, 2010). Indeed, many species across the globe already shifted their upper elevational range limits upslope (Freeman, Lee-Yaw, et al., 2018;Freeman, Scholer, et al., 2018). Projected changes were less conclusive under a range shift scenario. Associations between functional richness and, in particular, individual functional traits were projected to decrease under range shift. However, variation in projections was also highest under this scenario, indicating that future correspondence between plant and bird communities was most difficult to project if both lower and upper elevational ranges of species were assumed to change.

| Integrative analysis of functional diversity under climate change
Our approach provides a new way to integrate species' occurrences, functional traits and projection models to assess consequences of global warming for the functional correspondence between interacting taxa (Figure 1). The proposed approach can yield new information on how functional associations between taxa are likely to change under future climate change and is useful to identify under which scenarios functional mismatches between taxa are most likely to happen. Such analyses could be applied to a wide range of other taxa that are linked by trophic interactions and only require occurrence data for both taxa as well as information on the functional traits that mediate their interactions, such as traits that determine the probability of plant-pollinator or predator-prey interactions (Webb & Shine, 1993;. Such trait-based approaches are timely as more and better trait data are increasingly becoming available . Ultimately, a formal integration of such trait-based approaches into projection models is desirable and could improve how biotic interactions are accounted for in species distribution models (Dormann et al., 2018).
Our study is a first step towards comparative analyses of interacting taxa under climate change. We particularly encourage future studies that use more fine-scale occurrence data than those that are available for our study system. Moreover, the future projection models we apply rely on many assumptions and a better integration of species-specific responses in terms of physiological thermal tolerances (Khaliq, Hof, Prinzinger, Böhning-Gaese, & Pfenninger, 2014;Londoño, Chappell, Jankowski, & Robinson, 2017) and different dispersal capacities of species (Grewe, Hof, Dehling, Brandl, & Brändle, 2013) could improve model projections. A promising way forward could also be the use of time-series data that document how occurrences of interacting taxa have changed over time (Burkle, Marlin, & Knight, 2013), in order to quantify temporal variation in the functional associations between taxa.

| CON CLUS IONS
To our knowledge, this is the first study that simultaneously investigates how projected range change might influence the functional diversity of interacting taxa under climate change. Our results suggest that under a scenario in which species are sensitive to increasing temperatures and dispersal-limited, the potential for future functional mismatches between fleshy-fruited plants and frugivorous birds is highest. For conserving tropical mountain biodiversity, it might therefore be important to enable species to shift their ranges upslope and to plan for movement corridors along elevational gradients (Moore, Robinson, Lovette, & Robinson, 2008). Our approach of integrating functional diversity analyses and projection models can be widely applied to a range of interacting taxa linked by trophic interactions and could help to identify future scenarios under which biotic interactions between taxa are most vulnerable.