Habitat differences in seed‐dispersing vertebrates indicate dispersal limitation in tropical bracken‐dominated deforested areas

Shifting agriculture and anthropogenic fires are among the main causes of deforestation in the tropics. After fire and land abandonment, vast deforested areas are commonly dominated by the bracken fern Pteridium for long periods. Although forest regeneration in bracken‐dominated areas is mainly hindered by dispersal limitation, little is known about the role of seed‐disperser communities in slow forest succession. Our objective was to unravel the differences in the properties of the seed‐disperser communities between forests and bracken‐dominated areas to assess their role in dispersal limitation to foster ecological restoration. We compared species richness, diversity, abundance, and composition of seed‐dispersing birds, bats, and medium and large terrestrial mammals between both habitats in a montane forest of Bolivia. The species richness and diversity were similar for bats and higher for birds in bracken‐dominated areas than in the forest, but species composition was different between both habitats and groups. Although species composition was similar between both habitats for terrestrial mammals, the abundance was higher in the forest than in bracken‐dominated areas. Differences in species composition of seed‐dispersing birds and bats could be one of the main causes of dispersal limitation in forest regeneration in tropical bracken‐dominated areas. The few shared species between both habitats could explain the low seed rain of animal‐dispersed forest tree species in bracken‐dominated areas and the consequent hindered forest regeneration. Future studies relevant to natural forest regeneration should focus on analyzing the effects of animal‐attractants such as perches, artificial bat‐roosts, and seedling transplants on disperser communities, seed dispersal effectiveness, and forest regeneration.


| INTRODUC TI ON
Deforestation is one of the main causes of forest loss and biodiversity decline worldwide, negatively affecting species richness, community composition and biotic interactions (Curtis et al., 2018;Haddad et al., 2015;Marjakangas et al., 2020).Deforestation is threatening tropical montane forests, which are biodiversity hotspots of high conservation value (Beck & Richter, 2008;Karger et al., 2021).
Birds, large mammals, and bats are the most important seed dispersers in tropical forests (Howe & Smallwood, 1982;Medellin & Gaona, 1999).Therefore, these agents are essential to ensure the establishment of animal-dispersed trees in deforested habitats (Catterall, 2018;Howe, 2016;Muscarella & Fleming, 2007;Wunderle, 1997).This is particularly important for large-seeded plant species that are mainly dispersed by large frugivores, which are the first animals to be extirpated from disturbed habitats (Magioli et al., 2021;Markl et al., 2012).Deforestation threatens animal species and their ecological functions with cascading effects (Magioli et al., 2021).Plant-frugivore interactions are affected by deforestation in terms of quantity and diversity of visits, seed removal, and seed dispersal distance (Fontúrbel et al., 2015;Marjakangas et al., 2020;Markl et al., 2012).Furthermore, many seed dispersers avoid most deforested areas due to the lack of perch structures, food resources, hides, roosts, and vegetation complexity (Brady et al., 2011;Wunderle, 1997), consequently reducing the probability of seed arrival into deforested areas.These negative impacts are detrimental to forest recovery in deforested areas, especially if they are characterized by dispersal limitation.
Bracken occurs in all continents except Antarctica (Marrs & Watt, 2006) and is considered an invasive species in most parts of the world (Amouzgar et al., 2022;Hartig & Beck, 2003;Marrs et al., 2000).It dominates vegetation after fire or post-agricultural land abandonment and is very difficult to control (Alday et al., 2013;Pakeman et al., 2005;Suazo-Ortuño et al., 2015).Despite its dominance for long periods, the role of bracken in tropical forest regeneration is still controversial.Some authors claim that bracken inhibits forest regeneration because the high density of its fronds and litter shade other plants and act as physical barriers for seeds reaching the ground and seedling growth (Carvalho et al., 2022;Levy-Tacher et al., 2015;Marrs et al., 2000;Paz et al., 2022), that is, habitat limitation (sensu Münzbergová & Herben, 2005).Other authors suggest that habitat is hardly limited, as tree seedling survival and growth rates are similar in bracken-dominated areas and forest (Gallegos et al., 2016).Bracken presence reduces photosynthetic and water stress in adult plants (Paz et al., 2022).Tree seedling performance in bracken-dominated areas is higher than after bracken removal, both for naturally regenerating tree seedlings and after seed addition (Gallegos et al., 2015(Gallegos et al., , 2016;;López et al., 2023).Similar patterns were found in South America and Africa (Palomeque et al., 2017;Silva Matos & Belinato, 2010;Ssali et al., 2017Ssali et al., , 2018)).Therefore, Gallegos et al. (2016) suggest that tropical bracken-dominated areas are limited by seed dispersal rather than habitat (sensu Münzbergová & Herben, 2005).This means that tree species cannot establish in bracken-dominated areas because their seeds do not arrive, rather than they cannot grow there.Dispersal limitation was associated with lower species richness and density in seed rain, soil seed bank and tree seedlings in bracken-dominated areas, compared to forests (Gallegos et al., 2016;Günter et al., 2007;Palomeque et al., 2017;Saavedra et al., 2015;Ssali et al., 2017Ssali et al., , 2018)).Consequently, if the main limiting factor for forest regeneration in bracken-dominated areas is dispersal limitation and the main seed dispersers of most forest tree species are animals, the study of the seed disperser community is essential to understand the main filters of forest regeneration in such widely distributed habitats.This knowledge is essential for proposing restoration strategies to accelerate forest regeneration.
habitats could explain the low seed rain of animal-dispersed forest tree species in bracken-dominated areas and the consequent hindered forest regeneration.Future studies relevant to natural forest regeneration should focus on analyzing the effects of animal-attractants such as perches, artificial bat-roosts, and seedling transplants on disperser communities, seed dispersal effectiveness, and forest regeneration.
Abstract in Spanish is available with online material.
bats, birds, Bolivia, camera traps, frugivory, Pteridium, restoration, terrestrial mammals Dispersal limitation could be related to lower species richness or abundance of seed-dispersers in the deforested area compared to the forest (Figure 1b,c), or to differences in species composition of seed dispersers in the two habitats (Figure 1d).Differences in the community properties of seed-dispersers can be detrimental for forest regeneration in tropical disturbed areas, because of cascading effects on seed rain and, consequently, seedling establishment.To date, few studies evaluated seed disperser communities in tropical bracken-dominated areas (Maya-Elizarrarás & Schondube, 2015) and few studies analyzed the whole seed disperser community in deforested areas and adjacent forests at the same spatial and temporal scales (Catterall, 2018;Escribano-Avila et al., 2014;Santos et al., 1999).Therefore, our main objective is to unravel the differences in the community properties of seed dispersers between forests and bracken-dominated areas to assess their role in dispersal limitation (Figure 1).We aim to answer the following questions: What are the differences in species richness, diversity, and abundance of seed-dispersing birds, bats and terrestrial mammals between a tropical montane forest and the surrounding bracken-dominated areas?Does the species composition of seed dispersers differ between the two habitats?Are the dominant species in the forest also common in deforested areas?To our knowledge, this is the first study to compare seed-dispersing communities from different vertebrate groups between bracken-dominated areas and forests in the tropics.
F I G U R E 1 Scheme of likely scenarios of differences in the seed disperser community properties between the forest (left) and the bracken-dominated area (right) that could be related to dispersal limitation.The likely scenarios are ordered from lower (a) to higher (d) expected effects on dispersal limitation.

Forest Bracken
(a)

| Study area
The study was carried out between 2020 and 2022 on the eastern slope of the Andes, in the vicinity of Chulumani village, Sud Yungas, La Paz, Bolivia (16°24′ S, 67°31′ W).The study area is located in the tropical montane humid forest between 1850 and 2450 m a.s.l.
(Figure S1).The mean annual temperature is 20.5°C and the mean annual precipitation is 1400 mm, with a peak between December and February (Molina-Carpio et al., 2019).The landscape is heavily deforested by frequent human-induced fires and the extension of coca (Erythroxylum coca) plantations for traditional use (Killeen et al., 2008).Today, there are only two relicts of mature continuous forest (Lippok et al., 2014) of about 1000 and 3500 ha in the area.

| Birds and bats
To capture birds and bats, we installed understory mist nets at the edges of the two forest fragments at eight sites 1-12 km apart (Figure S1).Each site was sampled four times between March 2020 and June 2021, covering all seasons during the sampling period.We placed eight (9 × 2.5 m) mist nets 0.5 m from the ground, at each site, four in the forest and four in the bracken-dominated area, each 20 m inward or outward from the forest border.Mist nets were open for 3 days at each site, from 0630 to 1830 h to capture birds, and from 1830 to 2300 h to capture bats, with a total capture effort of 4067 net-hours for birds and 2576 net-hours for bats.After capture, birds and bats were kept for 20-30 min in cloth bags to collect regurgitates and fecal samples.The captured individuals were then identified, measured, tagged, and released.All species that had intact seeds in their feces were included as seed dispersers in the analyses.
Other captured species were classified as seed dispersers only if they frequently disperse seeds according to the literature (Billerman et al., 2022;Herzog et al., 2016;Stotz et al., 1996).

| Terrestrial mammals
To survey terrestrial mammals, we established 100 m-long transects perpendicular to the forest edge at five sites separated by at least 2 km (Figure S1).In the first sampling period from October 2020 to September 2021, we installed two Reconyx Hyperfire2 TM camera traps with infrared sensors at each transect, one in the forest interior, 50 m from the forest edge, and one in the bracken-dominated area, 50 m from the forest edge.In the second sampling period, from February to June 2022, we installed four camera traps at each transect: two in each habitat, 50 and 20 m inward and outward from the forest border.The cameras were placed near animal paths, without bait, tied to a pole or tree 50-80 cm above the ground.The cameras were programmed to take pictures 24 h a day, with highsensitivity and 5-s interval between consecutive pictures (10 per trigger).Cameras were checked every 3 months, when batteries and memory cards were replaced.The photographs were handled using NAIRA III free software (Pulido et al., 2018), and the species were classified manually by three experts.We considered independent events when consecutive photographs of individuals of the same species were taken more than 24 h apart at the same camera station to avoid overestimation of unmarked animals (Jax et al., 2015).In the second sampling period, data from cameras at 20 and 50 m inwards and outwards from the forest edge were pooled together before analysis.The total sampling effort was 4558 days.We detected 7255 photographs with terrestrial mammals (including small, medium, and large mammals).We classified as seed dispersers all registered medium and large terrestrial mammals (>800 g, hereafter terrestrial mammals) that had at least 20% of their diet consisting of fruits/seeds (Marjakangas et al., 2020) according to the Elton Traits database (Wilman et al., 2014).We removed all the other mammals from the analyses.

| Data analysis
We performed separate Generalized Mixed Effects Models to compare species richness and abundance of seed-dispersing birds, bats, and terrestrial mammals in the forest and bracken-dominated areas with the package lme4 (Bates et al., 2015).All models included habitat (forest vs. bracken) as the fixed independent variable and site as the random effect, to control for the study design.For each animal group, we included species richness or abundance as the response variables, considering a Poisson distribution of errors or negative binomial in cases of overdispersion.To analyze diversity, we built Linear Mixed Effects Models (LMM) with a similar structure, including the inverse Simpson's index as the response variable (Magurran, 2004).
For terrestrial mammals, we calculated the camera trapping rates or relative abundance index (RAI) per camera trap (to consider the variation among cameras, sites and habitats) as in Mandujano (2022).
We built a LMM including the RAI as the response variable.In every case, we calculated p values from log-likelihood ratio tests (LRT) comparing each model with a null model containing site as random effect, with the package MuMIn (Barton, 2022).For each model, we also calculated the marginal coefficient of determination R 2 to obtain the variance explained by the habitat.We also quantified the sample completeness of the surveys and compared species richness between habitats with individual-based rarefaction curves with the package iNEXT.4steps(Chao et al., 2020).
To assess the effect of habitat on community composition, we performed permutational multivariate analyses of variance (PERMANOVA) using the adonis2 function from the vegan package (Oksanen et al., 2022), setting the locality as strata to consider the block design, and the Jaccard dissimilarity method, and 999 permutations to calculate p values.To visualize differences in community composition, we performed nonmetric multidimensional scaling (NMDS) with the Jaccard dissimilarity index for birds and bats, and the Bray-Curtis index for terrestrial mammals, using the metaMDS function from the package vegan (Oksanen et al., 2022).Finally, we plotted rank-abundance curves to visualize the distribution of species abundance in forest and bracken-dominated communities.Due to the high degradation of our study area, we compared the species that were detected during our surveys (Table S1) with those included on a reference list that was built from the findings of previous studies in the area and its surroundings (Tables S2-S4).All analyses were conducted in R version 4.2.1 (R Core Team, 2022).
In the camera traps, we recorded nine species of medium and large terrestrial mammals (370 independent events): nine species (226 individuals) in the forest and eight species (144 individuals) in the bracken-dominated areas.
The completeness analysis for seed-dispersing birds revealed that in the forest we sampled 26% of the total species assemblage, with the detected species covering 86% of the assemblage's individuals and 100% of the most abundant species (q = 0, 1, 2 in Figure S2a).In the bracken-dominated habitat, we sampled 69% of the total bird species assemblage, with the detected species covering 87% of the assemblage's individuals and 99% of the most abundant species.For bats, in the forest, we sampled 51% of the total species assemblage, with the detected species covering 88% of the assemblage's individuals, and 99% of the most abundant species (q = 0, 1, 2 in Figure S2c).In the brackendominated habitat, we sampled 79% of the total bat species assemblage, with the detected species covering 88% of the assemblage's individuals, and 96% of the most abundant species.For terrestrial mammals, in the forest, we sampled 100% of the species (q = 0, 1, 2 in Figure S2e) while in the bracken-dominated habitat, we sampled 94% of the total species assemblage, with the detected species comprising 99% of the assemblage's individuals, and 99% of the most abundant species.
We found significant differences in species composition between forest and bracken-dominated areas for birds (PERMANOVA, R 2 = 0.23, F = 3.91, p = .008,Figures 1d and 3a) and bats (PERMANOVA, R 2 = 0.18, F = 2.58, p = .03,Figures 1d and 3b), but not for terrestrial mammals (PERMANOVA, R 2 = 0.05, F = 0.375, p = .75,Figure 3c).For terrestrial mammals, two of the three most common species in the forest were also among the three most common species in the bracken-dominated area (Mazama americana and Dasyprocta variegata, Figure 4c).The second most common species in the forest (Cuniculus paca) was scarce in the bracken-dominated area.The third most common species in the bracken-dominated area (Eira barbara) was the sixth most common species in the forest.

| DISCUSS ION
We compared the communities of seed-dispersing birds, bats, and terrestrial mammals in an Andean tropical montane forest and surrounding bracken-dominated areas to assess their contribution to dispersal limitation.We found that species richness, diversity, and abundance were similar for bats and higher for birds in the brackendominated areas than in the forest, but species composition was different between habitat types for both groups.Terrestrial mammals had a similar species composition in both habitats, but a higher abundance in the forest than in the bracken-dominated area.The high turnover in the species composition of seed dispersers could be one of the main causes of dispersal limitation in bracken-dominated areas, because few species are shared between both habitats, which reduces the seed rain of animal-dispersed forest tree species in deforested areas, hindering forest regeneration.Therefore, shared species are a keystone for forest recovery.

| Birds
Species richness and diversity of seed-dispersing birds were higher in the bracken-dominated area than in the forest, and bird abundance showed a similar trend.These results may reflect F I G U R E 2 Differences in biodiversity between forest and bracken-dominated areas for seed-dispersing birds (left column), bats (medium column) and medium and large terrestrial mammals (right column) in terms of (a-c) species richness, (d-f) inverse Simpson index, (g-h) abundance, and (i) relative abundance index.Asterisks indicate significant differences between both habitats (p < .05).Whittaker, 2007;Saavedra et al., 2014).This is consistent with the intermediate disturbance hypothesis as a result of the interplay of disturbance and competition/colonization that promote species coexistence (Liao et al., 2022).In contrast to our results, Maya-Elizarrarás and Schondube (2015) found lower species richness but similar abundance of birds in bracken-dominated areas than in forests in Mexico.We may have obtained different results because our study sites were older (5-10 years old) than theirs (<3years old), and they used a different method (i.e., point-counts for 2 months).Our lower species richness and abundance of birds in the forest may also be due to a lower survey completeness since the mist-nets only capture a subset of the community, especially small understory birds (Thiel et al., 2021).However, because small birds usually have a higher proportion of interactions in dispersal networks than other species (Béllo Carvalho et al., 2023;Carreira et al., 2020), our results may be representative of seed dispersers involved in interactions between both habitats.Furthermore, although bird species richness is commonly used to assess forest recovery, this indicator is not useful for discerning variations in F I G U R E 4 Rank-abundance curves for seed disperser communities of (a) birds, (b) bats, and (c) terrestrial mammals between forest and bracken-dominated areas.The species are ranked from highest to lowest abundance.The three most abundant species in each habitat are labeled with their abbreviation (first three letters of the genus and the species) and indicated by different colors.The same colors in both habitats denote the same species.Full species abbreviations are shown in Table S1.

Bracken Forest
community composition, which is important when analyzing the role of seed dispersers for forest recovery (Catterall et al., 2012;Dunn, 2004).
Species composition of seed-dispersing birds was different between the two habitats, consistent with other studies (Maya-Elizarrarás & Schondube, 2015;O'Dea & Whittaker, 2007), which could be the result of species replacement or rewiring (Marjakangas et al., 2020).Differences in species composition in deforested areas may be detrimental for forest recovery because a low number of species using both habitats reduces the likelihood that forest animaldispersed seeds will arrive in deforested areas.In fact, only nine species from the forest were also registered in the bracken-dominated area, suggesting that forest regeneration in disturbed habitats is highly dependent on these shared species (Rehm et al., 2017).These compositional dissimilarities could be explained by differences in habitat preferences, particularly for forest specialists, which are more sensitive to habitat disturbance (Colles et al., 2009;de Souza Leite et al., 2022;O'Dea & Whittaker, 2007).Future studies should analyze the interaction networks and functional traits from the seed dispersers and plant species to understand how species composition affects seed dispersal dynamics.
In our study, M. striaticollis was the only species of seed-dispersing birds common in both habitats.This species tolerates intermediate disturbance (Montaño-Centellas & Garitano-Zavala, 2015) and can feed on fruits and insects (Greeney et al., 2006).Interestingly, this was one of the species with more seed-dispersal interactions at the forest edge (Saavedra et al., 2014) and in bracken-dominated habitats in our study area, where it dispersed the highest number of forest dependent plant species (C.Mayta in progress).Generalists such as M. striaticollis can be critical to the dispersal and establishment of woody plants, increasing plant diversity in deforested areas (Carlo & Morales, 2016).Further data suggest that M. striaticollis could be a keystone species in our area because it is abundant in both habitats, disperses several forest species, and its seeds have high germination rates (Rojas, 2022).Some bird species registered only in the bracken-dominated area can also disperse seeds from earlysuccessional species.However, since early-successional species are light-demanding and bracken is an important source of shade, earlysuccessional species rarely colonize these areas (López et al., 2023).
Therefore, shade-tolerant forest species, which are mainly dispersed by animals will have more chances to establish in bracken-dominated areas and promote forest regeneration.In addition to the understory birds that we captured in mist nets, we observed large frugivorous birds, such as Penelope montagnii and Aulacorrynchus coeruleicinctis (C.Mayta submitted), in the forest and also sometimes perching in scatter burned trunks in bracken-dominated habitats.Both species move large distances through different habitats and are capable of dispersing several forest tree species (Reid et al., 2021;Saavedra et al., 2014), including large-seeded species with higher probabilities of establishment in bracken-dominated areas (Gallegos et al., 2015;Ssali et al., 2019) because of their higher germination rates and shadetolerance (Bazzaz & Pickett, 1980;Daws et al., 2007).Depending on perch abundance and frequency of flights, guans and toucans may play an important role in forest regeneration in bracken-dominated areas.
In this study, we registered 37 bird species, corresponding to only 49% of the frugivorous/omnivorous species previously registered in the study region (Rubey & Hennessey, 2005;Saavedra et al., 2014; M. Villegas unpublished data, Table S2).However, the completeness analysis revealed that we registered more than 86% of the abundant species in both habitats.Abundant species are particularly relevant, because they might collectively disperse more seeds.Since mist-nets only capture understory small birds, we could not capture species like Tersina viridis, Tangara spp., and Psarocolius spp.which commonly move in the canopy, or other large species such as P. montagnii, Ortalis guttata, Zentrygon frenata, and Patagioenas spp.Point counts could be more effective to register a larger set of bird species than mist-nets (Mulvaney & Cherry, 2020).Nevertheless, mist-nets were useful for our study to assess birds' diet and to detect the abundant species.We also only captured birds in the bracken-dominated area and the forest edge, and not in the forest interior.Several species from the reference list such as Lipaugus uropygialis, Pharomachrus antisianus, and Odontophorus balliviani are typical for the forest interior and avoid forest edges, explaining their absence from our dataset.
However, since our aim was to examine the species that could potentially disperse seeds into bracken-dominated areas, those species that avoid the forest edge probably also avoid deforested areas, and will not disperse seeds into bracken-dominated areas.Contrarily, we found some species that were not previously registered in the

| Bats
We found similar species richness, abundance, and diversity of seeddispersing bats in forest and bracken-dominated habitats.Other studies also found similar species richness and abundance of frugivorous bats between disturbed and undisturbed areas (de Oliveira et al., 2017;Laurindo et al., 2019).These patterns could be related to their tolerance to habitat disturbance and to the high affinity of many frugivorous bats for early successional plant species, like Piper, Solanum, and Cecropia (Muscarella & Fleming, 2007) that are common close to the forest edge.
The species composition of bats was different between the two habitats, agreeing with other studies (Chambers et al., 2016;de Oliveira et al., 2017;Valle et al., 2021).These patterns could reflect the longer time that species composition takes to recover after disturbance (Dunn, 2004) or species-specific differences related to food resources and home ranges (Chambers et al., 2016).
The most common bat in the forest was C. perspicillata and was also present in bracken-dominated areas but in lower abundance.
Since Piper is the main food resource for this species (Muscarella & Fleming, 2007), this pattern could be related to the lower availability of Piper in bracken-dominated areas (Lippok et al., 2013).
Since C. perspicillata can change its feeding habitat depending on food resources (Trevelin et al., 2013), planting Piper in brackendominated areas could increase its abundance, and consequently seed dispersal.
In this study, we detected 86% of the frugivore/omnivore bat species previously reported in the study region, in the Cotapata National Park located 35 km from our study area (Montaño-Centellas et al., 2015;Moya et al., 2024;SERNAP, 2005).This high percentage was not reflected in the completeness analyses, which suggested that we sampled only 51% of the total species assemblage in the forest and 79% in the bracken-dominated area.This could be related to the high amount of singletons that we found during sampling, which probably overestimated the species richness of the study region.The high number of species that we registered only once could suggest a low sampling effort, or could be a result of the high proportion of rare bat species in certain tropical areas (Meyer et al., 2015).The completeness analysis also suggests that we sampled at least 88% of the abundant species.In this study, we registered Anoura caudifer, Acinia peruana and Uroderma magnirostrum, which were not detected previously in the region.
We did not register Artibeus planirostris and Platyrrhinus umbratus, which were found in the Cotapata National park.We probably were not able to detect these species because they normally move in the canopy, or because these species are rare at these elevations (Meyer et al., 2010).Their absence from our study site could be also related to their sensitivity to habitat degradation.Longer-term monitoring would be necessary to be able to detect a higher proportion of the bat assemblage (Meyer et al., 2015).

| Terrestrial mammals
We found similar species richness, diversity, and composition of seed-dispersing terrestrial mammals in both habitats.These patterns could be related to the high vegetation cover in bracken-dominated areas, which is positively related to mammal species richness (Brady et al., 2011), as well as the short distance to the forest remnants, the high mobility of these animals and the presence of different food resources (Estrada et al., 1994).However, we found lower relative abundance of terrestrial mammals in bracken-dominated areas than in the forest, probably because of the higher availability of resources, hides, and roosts in the forest (Brady et al., 2011).
M. americana was the most common species in the brackendominated habitat and the third most common in the forest interior.
The high abundance of this species could be related to their large body size and greater foraging range, being able to cross the forest edges and exploit other habitats (Lees & Peres, 2009).D. variegata was the most common species found in the forest and the second most common in the areas dominated by bracken.Dasyprocta species are important seed dispersers, common in forests and disturbed areas (Jax et al., 2015).Therefore, D. variegata could be a keystone species, because it travels relatively long distances, frequently eats fruits, and could disperse forest tree species to bracken-dominated areas.Many of the terrestrial mammal species that we recorded as seed dispersers can also be classified as herbivores, omnivores, or even seed predators, and their disappearance can have cascading effects on forest regeneration (Kurten, 2013).Furthermore, since most tree seeds in the forest of our study areas are smaller than 1 cm, there is little risk that they will be preyed upon by most recorded species (Norconk et al., 1998).
We detected 82% of the species of seed-disperser terrestrial mammals registered in Cotapata National Park located 35 km from the study area.The completeness analysis revealed that we sampled 100% of the species.The absence of the two missing species could be related to the rarity of Cuniculus taczanowskii, and to the high gaming pressure for Tayassu tajacu in the past decades in our study area.However, as the completeness analysis suggested, we were able to detect the abundant species in both habitats.
Furthermore, the presence of Tremarctos ornatus and different felids highlight the conservation value of the forest remnants in our study area.
Although several studies analyze seed-dispersal networks (Castaño et al., 2018;Chaves et al., 2022;Saavedra et al., 2014), additional analysis of the seed dispersal community has the advantage of assessing the source of seed limitation.Seed dispersal is thought to be limited in deforested areas because of lower number of animals dispersing seeds.However, in this study, we show that the main cause of seed dispersal limitation is the scarcity of shared species between both habitats.This pattern could explain the low seed rain and the low seedling establishment of forest tree species previously reported in tropical bracken-dominated areas.Brackendominated areas are covered by bracken fronds, creating shade and a potential physical barrier, which gives shade-tolerant species advantages to recruit, survive and grow (Gallegos et al., 2016;Ssali et al., 2019).Given that forest shade-tolerant species tend to have larger seeds, and these seeds are mainly dispersed by animals, promoting the arrival of their seed dispersers will probably increase seed rain (Saavedra et al., 2015) and seedling establishment (C. Mayta, submitted).
Based on our findings, we suggest that increasing visitation by seed-dispersing animals capable of moving across both habitat types could help reduce dispersal limitation in bracken-dominated areas.
This could be done by installing animal attractants in the form of artificial and natural perch structures that increase species richness and abundance of bird-dispersed seeds (Saavedra et al., 2015) and seedlings in bracken-dominated areas (C.Mayta, submitted).We also suggest installing artificial bat roosts (Kelm et al., 2008), which increased seed rain of bat-dispersed species in bracken-dominated areas in a parallel study (C.Mayta, submitted).In addition, transplanting nursery-raised seedlings of several animal-dispersed species, including Clusia, Miconia, Hieronyma, Piper, Vismia, Cecropia, and Ficus, into bracken-dominated areas could foster their establishment and rapid growth.These transplants could accelerate the availability of food resources and shelter for seed dispersers in the medium and long term, and increase species richness and abundance of animaldispersed tree seeds (Camargo et al., 2020;Galindo-González et al., 2000) and seedlings (Martinez-Garza & Howe, 2003;Reid et al., 2021;Rivas-Alonso et al., 2021).Future research should focus on evaluating the effect of these restoration measures on seed disperser communities, seed dispersal effectiveness, and forest regeneration.
Assessing community properties of the entire seed dispersing community is critical to understanding constraints on forest regeneration in deforested areas.Here, we found that while species richness of seed-dispersing bats was similar and that of birds was higher in bracken-dominated areas than in the forest, the species composition of both groups was different in both habitats.These results could explain the low seed rain and seedling establishment of forest tree species previously reported in tropical bracken-dominated areas.Our results highlight the importance of including animalrelated aspects in restoration projects.

CO N FLI C T O F I NTER E S T S TATEM ENT
The corresponding author confirms on behalf of all authors that there have been no involvements that might raise the question of bias in the work reported or in the conclusions, implications, or opinions stated.
Of the three most common species of birds recorded in the forest, only one was also among the most common birds recorded in the bracken-dominated area (Mionectes striaticollis, Figure4a).Out of the remaining two, one species (Chiroxiphia boliviana) was rare in the bracken-dominated area, while the other (Atlapetes rufinucha) was only the fifth most common species in the bracken-dominated area.Besides M. striaticollis, the other two species of common birds in the bracken-dominated area (Elaenia albiceps and E. obscura) were absent in the forest.Two of the three most common bat species detected in the forest were also among the most common bats in the bracken-dominated area (Carollia perspicillata and Sturnira tildae, Figure4b), while the other common bat species in the forest (Carollia brevicauda) was only the fifth most common species in the bracken-dominated area.The most common bat species in the bracken-dominated area (Sturnira oporaphilum) was scarce in the forest.
Comparison of seed disperser communities of (a) birds, (b) bats, and (c) terrestrial mammals, between forest and bracken-dominated areas using non-metric multidimensional scaling ordinations (NMDS).Species are denoted in gray, forest sites in green, and bracken-dominated sites in red.Ellipses differentiate the communities with their names on the centroids.Species abbreviations are shown in TableS1. of open-land generalist birds in early and midsuccessional forests in the tropics(Catterall et al., 2012;Pizo & dos Santos, 2011;Shankar Raman et al., 1998), as well as the ability of several bird species to survive in disturbed habitats, especially generalist frugivores (de SouzaLeite et al., 2022;Gomes et al., 2008;O'Dea & Whittaker, 2007).Higher bird diversity has been reported in areas of moderate disturbance, such as forest edges (O'Dea & study area, including Spinus spp.which are more common in drier ecosystems, Elaenia spp.which have preference for edges and open habitats, and Cyclaris guianensis and Leptopogon superciliaris which are typical for forest edges.Given the high deforestation rates in the study area, leaving only two large forest patches, several species that are more sensitive to forest degradation could be locally extinct, potentially explaining the absence of Ampelion rufaxilla, Myadestes ralloides, Creurgops dentata and Phylloscartes ophthalmicus. , JR, GMA conceptualized the research and methodology; CM, MV, VR, ME and SCG conducted the investigation and data curation; SCG, GMA, JR, MEV, NB-H, KN conducted the formal analysis; SCG and IH were involved in project administration, funding acquisition, and wrote the original draft; SCG, IH, CM, MV, NBH and KN wrote, reviewed and edited the manuscript.ACK N OWLED G M ENTSWe thank Noel Taboada, Jairo Dominguez, Miguel Montenegro, Cristian Vargas and Cecilia L. López for field assistance.We are grateful to the private land owners and local communities for facilitating access to their lands.We thank Diana Bowler for language polishing and two anonymous reviewers for their comments.This study was financed by the German Research Foundation DFG (grant He 3041/23-1) and by TWAS-Sida (grant 19-107 RG/BIO/LA_I-FR3240310157). All captured animals were manipulated according the Bolivian regulations (permit: CAR/MMAYA/VMABCCGDF/ DGBAP/MEG 0012/2020).Open Access funding enabled and organized by Projekt DEAL.