Climate and microhabitat shape the prevalence of endozoochory in the seed rain of tropical montane forests

Endozoochory, the dispersal of seeds by animal ingestion, is the most dominant mode of seed dispersal in tropical forests and is a key process shaping current and future forest dynamics. However, it remains largely unknown how endozoochory is associated with environmental conditions at regional and local scales. Here, we investigated the effects of elevation, climate, and microhabitat conditions on the proportion of endozoochorous plant species in the seed rain of the tropical Andes of southern Ecuador. Over 1 year, we measured seed rain in 162 seed traps on nine 1‐ha forest plots located at 1000, 2000, and 3000 m a.s.l. We recorded climatic conditions (mean annual temperature and rainfall) in each plot and microhabitat conditions (leaf area index and soil moisture) adjacent to each seed trap. In total, we recorded 331,838 seeds belonging to 323 morphospecies. Overall, the proportion of endozoochorous species in the seed rain decreased with elevation. The relative biomass of endozoochorous species decreased with increasing rainfall, whereas the relative seed richness of endozoochorous species increased with increasing temperature and leaf area index. These findings suggest an interplay between climate factors and microhabitat conditions in shaping the importance of endozoochorous plant species in the seed rain of tropical montane forests. We conclude that changing climatic and microhabitat conditions are likely to cause changes in the dominant dispersal modes of plant communities which may trigger changes in the current and future dynamics of tropical forests.

and biotic factors shape the prevalence of endozoochory in tropical montane forests is important for projecting the consequences of changing environmental conditions on ecological communities and ecosystem functioning in these vulnerable ecosystems (Bendix et al., 2021;Madani et al., 2018).
Seed dispersal by endozoochory plays an important role in several ecosystem processes. For instance, it influences the dynamics of plant populations such as avoiding negative density dependence (Howe & Smallwood, 1982;Janzen, 1970), reaching suitable microsites (Wenny, 2001), colonizing new habitats (Puerta-Piñero et al., 2013), and increasing the capacity of plants to track climate change (Fricke et al., 2022). Hence, seed dispersal by endozoochory is a key ecosystem service shaping the distribution of plant species, especially in tropical ecosystems (Aslan et al., 2019;Bello et al., 2015).
Endozoochory can be predicted from the morphological traits of seeds and fruits (Pérez-Harguindeguy et al., 2013;Rojas et al., 2022;van der Pijl, 1982). Endozoochorous plant species are characterized by thick-coated seeds inside edible fleshy fruits which are attractive to mutualistic animal partners (Snow, 1981;Stevenson et al., 2002). In contrast, in nonendozoochorous plant species, seed dispersal often depends on the attachment of seeds to animals' fur or on seed dispersal by abiotic factors such as wind, water, or gravity (Howe & Smallwood, 1982;van der Pijl, 1982). These plant species usually have dry diaspores and specific structures, such as hooked spines, arists, stiff hairs, or wings (Janson, 1983;Jara-Guerrero et al., 2011;Tovar et al., 2020;van der Pijl, 1982). Therefore, by studying the composition of seed and fruit morphological traits, it is possible to identify the main modes of seed dispersal in plant communities.
Since elevational gradients reflect gradients in climatic conditions, the dominance of certain seed dispersal modes along elevations may be driven by abiotic factors. Indeed, different abiotic factors are associated with the proportion of plant species dispersed by animals. For instance, species with fleshy fruits and large seeds are more common in wet forests than in dry forests (Almeida-Neto et al., 2008;Gentry, 1983;Tabarelli et al., 2003), corresponding to a general increase in endozoochory with increasing rainfall in the Neotropics (Correa et al., 2015). Despite this, the proportion of endozoochorous species was related to temperature rather than to rainfall in 64 Andean montane forests (Buitrón-Jurado & Ramírez, 2014). Generally, productive forest ecosystems with high rainfall and temperature appear to support more endozoochorous plant species given the high costs involved in the production of fleshy fruits (Bonte et al., 2012;Willson et al., 1989).
In addition to climatic factors, microhabitat conditions influence the distribution of plant species (Chanthorn et al., 2016;Stark et al., 2015) and also the seed deposition patterns of animal seed dispersers (García-Cervigón et al., 2018;Morán-López et al., 2020;Schupp et al., 2010). Within tropical forests, early and late successional plant species are favored by distinct microhabitat conditions related to canopy complexity and, to a lesser extent, to soil properties (Cheng et al., 2022). Wind-dispersed and light-demanding plant species are generally more abundant in early successional habitats (Tabarelli & Peres, 2002), characterized by an open canopy structure. These habitats can either be characterized by high soil moisture due to low rainfall interception and plant transpiration (Muscolo et al., 2014) or by low soil moisture due to high solar radiation and increased evaporation (Camargo & Kapos, 1995). On the contrary, endozoochory and seed size tend to increase in importance in late successional habitats (Bello et al., 2015). In these habitats, the canopy structure is more complex (Unger et al., 2013) and there may be high or low water retention in the soils (Camargo & Kapos, 1995;Muscolo et al., 2014). Variation between early and late successional habitats in fruit biomass and canopy complexity is also associated with habitat use and seed deposition patterns of frugivorous species, all of which affect the proportion of seeds dispersed to different habitats (Ferger et al., 2014;Holl, 1998;Hollander & Vander Wall, 2004;Loayza & Rios, 2014). So far, there are no studies testing how climate conditions at the large scale and microhabitat conditions at the small scale are related to the distribution of seed dispersal modes in the seed rain of tropical plant communities.
In this study, we investigated the importance of endozoochory in plant communities located along an elevational gradient in the tropical Andes in southern Ecuador. Based on seed rain data from 162 traps and across an entire year, we hypothesized that (a) the proportion of endozoochorous species in the seed rain of tropical montane forests decreases with increasing elevation because of the combined effects of high temperature and rainfall at low elevations that tend to favor fleshy-fruited species (Buitrón-Jurado & Ramírez, 2014); (b) endozoochory increases with increasing rainfall and temperature because productive environments support the high costs of producing fleshy fruits (Bonte et al., 2012;Willson et al., 1989); and finally, (c) a higher prevalence of endozoochory in microhabitats that favor late successional plant species with large seeds and endozoochorous seed dispersal (de Melo et al., 2006;Tabarelli & Peres, 2002).

| Study system
The study area was located in the Andes of southern Ecuador. This area is covered by mature forest along an elevational gradient ranging from 1000 to 3000 m a.s.l. Specifically, we established three study sites in Podocarpus National Park and San Francisco Reserve: (1) "Bombuscaro" around 1000 m a.s.l., located in the evergreen premontane forest (4°6′ S, 78°58′ W); (2) "San Francisco" around 2000 m a.s.l., located in the lower montane forest (3°58′ S, 79°4′ W); and (3) "Cajanuma" around 3000 m a.s.l., located in the upper montane forest (4°6′ S, 79°10′ W) (Homeier et al., 2008) (Figure 1a). The mean annual temperature decreases from 20 to 10°C with increasing elevation (Bendix et al., 2008), but temperature conditions are relatively stable throughout the year within each of the three study sites (see the monthly variation in temperature in Figure S1). Mean annual rainfall increases to up to 4500 mm/year at the upper montane forest, whereas rainfall is similarly high at mid (i.e., 2128.9 mm/ year at 2000 m a.s.l.) and low elevations (i.e., 2218.6 mm/year at 1000 m a.s.l.) (Bendix et al., 2008;Emck, 2007). The entire study area is subject to constant rainfall throughout the entire year (see the monthly variation in rainfall in Figure S1). Given these patterns, temperature decreases with increasing rainfall along the elevational gradient (r = −.82, p < .0001). At each elevation, we worked on three 1-ha plots (i.e., nine plots in total) established by the DFG research

| Seed rain and seed dispersal modes
At each 1-ha plot, we installed 18 regularly spaced seed traps (trap area = 0.36 m 2 , 1.5 mm nylon mesh, Figure 1b,c). The traps were hanging at approximately 90 cm above the ground and the content of each trap was recovered every 15 days for a period of 1 year (January 2019-January 2020). The material collected in the traps was dried, and all intact dispersal units ≥1 mm in length were counted and identified to the lowest possible taxonomic level using reference plant material from the forest, field guides, and the help of botanical experts. We classified seeds as endozoochorous or nonendozoochorous species based on the evidence from scientific literature, the consultancy with experts, and seed morphology (e.g., seeds of fleshy fruits or arillated seeds vs. ornamented seeds with wings, hair, or hooks). The seeds that could not be classified into endozoochorous or nonendozoochorous species were excluded from the analyses. In most cases, the dispersal units corresponded to seeds in a strict botanical sense. However, for some nonendozoochorous species, the dispersal units (diaspores) were made up of seeds and the surrounding fruit structure. All diaspores are referred to as seeds in the following. After drying seeds at

| Climatic conditions
For each plot, we recorded two climatic factors: mean annual rainfall (mm) and mean annual temperature (°C). Both variables were obtained by calculating the annual means of the total amount of rainfall and temperature recorded over a period of 3 years (2018-2020).
Climatic data were derived from an operational network consisting of automatic climate stations, remote sensing techniques, and a regionalization tool developed for the study area (see more details in Fries et al., 2009;Rollenbeck & Bendix, 2011).

| Microhabitat conditions
To characterize microhabitat conditions, we recorded canopy structure and soil conditions in nine 1 m 2 subplots per plot (i.e., each subplot was located adjacent to two seed traps characterized by similar canopy structures). For canopy structure, we measured the leaf area index which indicates how much of one-sided foliage area is projected per unit ground surface area (Chen & Black, 1992). To calculate the leaf area index, one hemispherical photo was taken per subplot with a Nikon FC-E8 Fish-Eye Converter (Nikon Corporation). The camera was placed 0.7 m above ground and oriented toward the sky. All photos were taken in the mornings with an overcast sky. Leaf area index values were estimated from hemispherical photos with Gap Light Analyser version 2.0 (Frazer et al., 1999). High leaf area index values indicate high foliage interception per unit of ground surface resulting in a more complex canopy structure (Chen & Black, 1992). For soil conditions, we recorded soil moisture which refers to water available for plants in the soils (Schmugge et al., 1980). We used a tensiometer (SM150 Kit, Delta-T Devices Ltd.) to measure soil moisture at five different points on the surface layer of soil in each subplot. Both soil moisture and leaf area index were recorded on the same dates (in October 2019). Since we were not able to measure the leaf area index close to two seed traps at 1000 m a.s.l., we extrapolated these two values from the closest traps. Soil moisture and leaf area index values were log-transformed prior to the analyses and were uncorrelated (n = 162 traps, r = .12, p = .12). Microhabitat and climatic conditions were only weakly related (n = 9 plots, rainfall vs. leaf area index, r = −.33, adjusted p = 1 (Holm's method), rainfall vs. soil moisture, r = .20, adjusted p = 1 (Holm's method); temperature vs. leaf area index, r = .76, adjusted p = .09 (Holm's method), temperature vs. soil moisture, r = −.12, adjusted p = 1 (Holm's method)) (see climate and microhabitat conditions recorded per elevation in Table S2).

| Statistical analysis
We tested our three hypotheses (a-c) with data on both seed rain biomass and seed richness by pooling seed rain samples across the entire study year, given the relatively constant environmental conditions throughout the year. First, for each seed trap, the relative seed rain biomass was calculated as the seed dry mass of endozoochorous species divided by the total seed biomass. This term, also understood as the proportion of endozoochorous species collected per trap along the year, was logit-transformed prior to the analyses (adjusting values of 0 and 1 by 0.025). Second, the relative seed rain richness was defined by the number of endozoochorous species sampled in a seed trap over the entire year divided by the total seed richness. Therefore, the relative seed rain richness refers to the proportion of endozoochorous species recorded per trap throughout the year. We tested relative seed rain richness with a binomial model including the classification of the two species groups as response variables ("yes" = endozoochorous species, "no" = nonendozoochorous species).
We tested the hypotheses (a-c) with generalized linear mixedeffects models (GLMMs) including the identity of study plots (n = 9) as a random factor. Hypothesis (a) was tested by including elevation  (Burnham & Anderson, 2002). The model with the lowest AICc value was selected as the most parsimonious model, but all models with a delta AICc value < 4 relative to the best model are shown (see Table S4 and S5). For all final models, we measured the proportion of variance explained by fixed factors (R 2 marginal) and random and fixed factors (R 2 conditional). All statistical analyses and graphs were performed with R statistical software version 4.1.2 (R Core Team, 2021), with the use of car (Fox & Weisberg, 2019), glmmTMB (Brooks et al., 2017), MuMIn (Barton, 2019), ggplot2 (Wickham, 2016), and ggpubr packages (Kassambara, 2018).

| RE SULTS
Overall, we recorded a total of 331,838 seeds over the 1-year sampling period along the elevational gradient. Seeds were collected in 160 seed traps because two traps never contained seeds (one at 1000 m a.s.l., one at 3000 m a.s.l.). About 76.4% of the seeds (61.3% of the morphospecies) could taxonomically be identified, others remained classified as undetermined morphospecies. Overall, the seed rain consisted of 323 species/morphospecies distributed over 51 plant families (see the complete list of species/morphospecies in Table S3). Melastomataceae and Lauraceae were the most widespread families of the endozoochorous species, while Asteraceae and Melastomataceae were the most dominant families of the nonendozoochorous plant species. The mean seed dry mass per species (mean seed mass = 0.06 g, SD = 0.22, n = 300) ranged from tiny seeds with 0.005 μg (Family Ericaceae) to heavy seeds such as those of Anomospermum reticulatum (Family: Menispermaceae) with 2.6 g.
Overall, endozoochorous species represented 83% of the total seed rain richness.
The relative seed rain biomass and richness of endozoochorous species were both associated with elevation. At low and mid elevations, the relative seed rain biomass of endozoochorous species was similar, but it declined in the upper montane forest (Table 1,  Table 2). The second-best model included rainfall and soil moisture as predictors (delta AICc relative to the best model = 1.14, all model combinations with delta AICc < 4 are shown in Table S4).
Relative seed rain biomass of endozoochorous plant species was negatively associated with rainfall (Table 2, Figure 3a). According to the most parsimonious model, relative seed rain richness was related to temperature and leaf area index ( Table 2). The second-best model included temperature, leaf area index, and soil moisture (delta AICc relative to the best model = 1.93, all model combinations with delta AICc < 4 are shown in Table S5). In the two most parsimonious models, temperature and leaf area index were positively related to the relative seed rain richness of endozoochorous plant species (

| DISCUSS ION
We studied the importance of endozoochory in the seed rain of plant communities located along an elevational gradient in the tropical Andes of southern Ecuador. Increasing elevation led to a decrease in the relative seed rain biomass and richness of endozoochorous plant species. Relative seed rain biomass of endozoochorous species decreased with increasing rainfall, whereas relative seed rain richness of endozoochorous species increased with increasing temperature TA B L E 1 Generalized linear mixed effect models (GLMMs) of the effect of elevation on (a) relative seed rain biomass and (b) relative seed richness of endozoochorous species Note: Given are estimates, standard errors (SE), z values, and p values. Marginal R 2 (R 2 m ) and conditional R 2 (R 2 c ) are shown in the final columns. The sample size corresponds to 160 traps. Identities of plots (n = 9) were included as random effects in both models.

F I G U R E 2
Relative seed rain biomass and richness across the elevational gradient in southern Ecuador. (a) Relative seed rain biomass is defined as the proportion of dry biomass of endozoochorous species relative to the total dry biomass of species per trap over the entire year; (b) relative seed rain richness is defined as the proportion of endozoochorous species relative to the total number of species per trap over the entire year. Boxes show 25th and 75th percentiles, with the median indicated, whiskers show data range and gray points are outliers. and higher leaf area index. Our analyses suggest an interplay be- This trend might be driven by the fruit production of abundant and large-seeded plant species contributing a large fraction to seed rain biomass. We suspect that similar values of the relative seed rain biomass of endozoochorous species at low and intermediate elevations may derive from the high records of seed rain of abundant smallseeded species (e.g., Moraceae, Melastomataceae spp.) as well as of TA B L E 2 Generalized linear mixed effect models (GLMMs) of the effects of climatic and microhabitat conditions on (a) relative seed rain biomass and (b) seed rain richness of endozoochorous species Note: Given are estimates, standard errors (SE), z values, and p values for each best model. Estimates are directly comparable because all predictors were scaled to zero mean and unit variance prior to the analysis. Marginal R 2 (R 2 m ) and conditional R 2 (R 2 c ) are shown in the two last columns. Identities of plots (n = 9) were included as random effects. The sample size was 160 traps.

F I G U R E 3
Effects of climatic factors on the seed rain of endozoochorous species across the elevational gradient in southern Ecuador. (a) Relative seed rain biomass of endozoochorous species in relation to mean annual rainfall, and (b) relative seed rain richness of endozoochorous species in relation to mean annual temperature. Every point represents mean values of seed rain at the respective nine 1-ha plot levels, and the vertical lines are the corresponding standard deviations in seed rain biomass and richness across the 18 seed traps in each plot.

F I G U R E 4
The relationship between relative seed rain richness and leaf area index (LAI). Points represent values for each seed trap (n = 160), and the line shows the linear trend in the relationships (see Table 2 for the results from generalized linear mixed effect models). LAI is log-transformed. The hemispherical photo on the low right side illustrates an LAI of 4.15 at 1000 m a.s.l.
Second, relative seed rain richness of endozoochorous species systematically decreased with increasing elevation. The decrease of endozoochorous plant species with increasing elevations might be due to a systematic shift in plant species composition at the high elevation. Decreased productivity at high elevations (Malhi et al., 2017;Tanner et al., 1998) leads to a reduction of plant taxa producing fleshy fruits at the highest elevation. In addition, herbaceous plant species, which increase in dominance at high elevations, are often dispersed by abiotic agents rather than by biotic vectors (Arbeláez & Parrado-Rosselli, 2005;Armesto & Rozzi, 1989;Tovar et al., 2020).
The continuous decline in the importance of endozoochory is also consistent with a decline of avian frugivores along the elevational gradient (Quitián et al., 2018). This study in the same montane forests found lower diversity of plant-frugivore interactions at high elevations, corresponding to a decreased number of frugivorous bird species in the upper montane forest.
In our study, climatic conditions were associated with the prevalence of endozoochory in seed rain. Relative seed rain biomass of endozoochorous plant species decreased with increasing rainfall. This finding is not consistent with the positive association between rainfall and the abundance of endozoochorous plants that have been found in other Neotropical (Almeida-Neto et al., 2008;Correa et al., 2015) and Paleotropical forests .
These contrasting results might be associated with the relationship between water availability and productivity in our study system. Generally, rainy environments have high productivity that supports the high costs involved in the production of fleshy fruits (Bonte et al., 2012;Willson et al., 1989). In many tropical mountains, however, extremely high rainfall is associated with a decrease in soil fertility and lowered productivity because of the leaching of soil nutrients (Malhi et al., 2017). Lowered productivity may result in reduced production of fleshy fruits at the highest elevation in our study system. As proposed by Willson et al. (1989), fleshy fruit production may be driven by productivity rather than by rainfall in tropical montane forests.
The relative seed rain richness of endozoochorous plant species increased with increasing temperatures. The proportion of endozoochorous species was also positively correlated with the mean annual temperature for other montane forests in the northern Andes (Buitrón-Jurado & Ramírez, 2014). One potential explanation for the temperature effect on endozoochory could be related to the continuous decrease in productivity with decreasing temperature (Pau et al., 2018). In addition, low temperatures may damage fleshy fruits (Burke et al., 1976) which are characterized by high levels of organic compounds and water (Coombe, 1976). A previous study has shown that endozoochorous species growing at high elevations tolerate higher temperatures than nonendozoochorous species (Tovar et al., 2020). Therefore, it is likely that the prevalence of endozoochorous plant species is determined by a combination of physiological constraints imposed by temperature and involved in the production of fleshy fruits.
An additional explanation for the increase of endozoochorous species towards warmer environments is that temperature modulates the diversity of animal seed dispersers (Peters et al., 2016) and thereby indirectly affects the relative richness of endozoochorous plant species. In line with that, positive effects of temperature have been reported on the diversity of frugivorous birds (Santillán et al., 2020) and indirect associations between plant and bird diversity have been shown to be more important than direct climatic effects in another tropical mountain system (Vollstädt et al., 2017).
This suggests that a higher number of endozoochorous plant species can be found in environments with a more diverse community of animal frugivores.
In addition to the climatic gradients across the mountain, microhabitat conditions were also related to the relative richness of endozoochorous species in the seed rain of tropical montane forests.
In line with our hypothesis, the prevalence of endozoochory in seed rain richness increased with increasing leaf area index, while endozoochory was only weakly related to soil moisture. It is likely that late successional habitats, characterized by increased canopy complexity (i.e., with denser foliage in the canopy, Unger et al., 2013), harbor numerous endozoochorous species, most of which produce rather large seeds. We find further support for this explanation in the covariation between seed dispersal mode and other plant traits. For instance, it has been shown that animal-dispersed plants in Atlantic forests tended to be tall trees with a high wood density (Bello et al., 2015).

| CON CLUS IONS
We show that the importance of endozoochory in the seed rain of tropical montane forests is shaped by an interplay of climatic and microhabitat conditions at large and small spatial scales.
Changing climatic and microhabitat conditions are therefore likely to jointly lead to compositional changes in the dominant dispersal modes of tropical plant communities. Given the importance of seed dispersal for ecological processes and ecosystem functions (Bello et al., 2015;Fricke et al., 2022), such changes in species composition will need to be considered in predictive models of the dynamics of tropical forest ecosystems under global change (Bendix et al., 2021).

CO N FLI C T O F I NTE R E S T
The authors declare no potential conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in this manuscript have been uploaded to the Dryad Repository and are accessible using the following DOI: https://doi. org/10.5061/dryad.6hdr7 sr4v.