Animal seed dispersal recovery during passive restoration in a forested landscape

Abstract

Seed dispersal by animals is key for restoration of tropical forests because it maintains plant diversity and accelerates community turnover. Therefore, changes in seed dispersal during forest restoration can indicate the recovery of species interactions, and yet these changes are rarely considered in forest restoration planning. In this study, we examined shifts in the importance of different seed dispersal modes during passive restoration in a tropical chronosequence spanning more than 100 years, by modelling the proportion of trees dispersed by bats, small birds, large birds, flightless mammals and abiotic means as a function of forest age. Contrary to expectations, tree species dispersed by flightless mammals dominated after 20 years of regeneration, and tree richness and abundance dispersed by each mode mostly recovered to old growth levels between 40 and 70 years post-abandonment. Seed dispersal by small birds declined over time during regeneration, while bat dispersal played a minor role throughout all stages of succession. Results suggest that proximity to old growth forests, coupled with low hunting, explained the prevalence of seed dispersal by animals, especially by flightless mammals at this site. We suggest that aspects of seed dispersal should be monitored when restoring forest ecosystems to evaluate the reestablishment of species interactions.

This article is part of the theme issue ‘Understanding forest landscape restoration: reinforcing scientific foundations for the UN Decade on Ecosystem Restoration’.

1. Introduction

A current challenge in ecological restoration is the reestablishment of species interactions that accelerate forest regeneration and ecosystem functioning [1]. The recovery of forest cover in abandoned farmland or degraded landscapes is important, but understanding which ecological interactions reestablish during restoration and their speed of recovery can be equally important [1,2]. Ecological interactions are ultimately responsible for the maintenance of biodiversity, particularly in tropical ecosystems [3], and monitoring change in species interactions over time can help practitioners evaluate the functioning and integrity of restoration projects [2]. Accordingly, the UN Decade on Ecosystem Restoration highlights the necessity to understand and conserve the fauna and flora that assist ecological restoration to restore ecological processes and biodiversity [4].

Seed dispersal is an ecological interaction critical to the success of ecosystem restoration [5,6]. In tropical terrestrial ecosystems, where more than 65% of plant species are dispersed by animals [7,8], it is paramount to understand how different seed-dispersing animals assist restoration, and which landscape variables drive the abundance of seed sources and dispersers [6,9,10]. The dispersal modes of dominant tree species tend to vary as ecosystem succession progresses. Typically, small-seeded tree species dispersed by small birds, bats and wind colonize abandoned agricultural areas [11]. These species form the canopy that encourages colonization by larger birds and mammals that bring in larger-seeded and later-successional tree species [11]. However, landscape context impacts interactions among dispersers and seed sources, with a slow recovery of interactions in regenerating sites in a deforested matrix with few plant species [12], where seed movement usually depends on wind dispersal, and perhaps on bats and small birds [13,14]. By contrast, in sites surrounded by a heterogeneous matrix with large forest tracts, interactions among dispersers and plants recover more quickly [12] and seed movement is usually mediated by larger birds and flightless mammals, as well small birds and bats [15,16].

Therefore, to establish successful restoration pathways, we need to understand the factors (e.g. landscape forest cover) that drive differences in seed dispersal across regenerating sites and how these relate to site-specific variation in the speed and predictability of regeneration [17,18]. For example, re-establishment of seed dispersal networks that are typical of older forests supports regeneration at larger spatial scales, because dispersers connect habitats across landscapes [1,19,20]. Accordingly, to fully understand the drivers of forest restoration it is key to assess (1) how dispersal modes shift during regeneration, and (2) what factors within and across landscapes affect dispersal modes and accelerate regeneration. Describing unhindered paths of recovery of ecological interactions during passive restoration through use of exemplary scenarios (e.g. abundant seed sources and dispersers) can serve as a benchmark for restoration initiatives.

Typically, frameworks used to evaluate and update restoration actions using monitoring data (e.g. Adaptive Management Cycles) monitor physical (e.g. water flow) or biotic parameters (e.g. tree species composition) [21–23]. However, the reestablishment of ecological interactions, such as seed dispersal, is rarely considered when designing restoration projects or monitored during restoration [2,21,22] despite its importance for forest regeneration. A recent review of Neotropical restoration initiatives shows that only 3.3% of studies have monitored seed dispersal by animals [24]. When seed dispersal has been monitored, researchers have usually compared fewer than three modes of dispersal, typically biotic versus abiotic [25,26], and focused on seed arrival without differentiating among dispersers [13]. Although useful, such approaches do not account for ecological differences among dispersers that have important implications for plant turnover during restoration [14]. For example, some animal taxa disperse more shade-tolerant species than others, so the contribution to plant turnover varies across different groups of animal disperses [19,27,28].

In this study, we use a unique long-term dataset from a secondary forest chronosequence in central Panama that spans more than 100 years of regeneration [29] to describe how interactions between plants and dispersers recover during passive restoration (i.e. natural regeneration) in a well preserved landscape. While most forest regeneration and restoration projects across the tropics have described the first 50 years of regrowth [30,31], only a handful span more than a century of regeneration, providing rare longitudinal data to evaluate the role of seed dispersal by animals during restoration. We integrate these data with detailed descriptions of dispersal modes for more than 90% of tree and shrub species present at this site, enabling us to describe how five modes of dispersal change during regeneration. As far as we are aware, no previous study has assessed how seed dispersal by bats, small birds, large birds, flightless mammals and abiotic means change across such a long timeframe of regrowth.

Based on previous studies [9], we expected that the importance of seed dispersal by small birds, bats and abiotic means should show higher prevalence than large birds and flightless mammals and decline over time during regeneration. As forests gain structural complexity and continue ageing, seed dispersal by flightless mammals and large birds should become more important than other dispersal modes. As flightless mammals and large birds increase in abundance and disperse more shade-tolerant species, small-bodied dispersers and the plants they disperse should decrease in abundance [9]. In addition, the likelihood of a single plant species being dispersed by multiple dispersal modes (i.e. redundancy) should increase over time as flightless mammals and large birds become more important during regeneration [9].

2. Methods

We conducted this study in the Barro Colorado Nature Monument (BCNM), in central Panama. Mean annual precipitation at BCNM is approximately 2600 mm, with a dry season from mid-December to mid-April [32]. The vegetation is classified as tropical moist forest [33]. BCNM comprises Barro Colorado Island (BCI) and five adjacent peninsulas [34], which harbour a mix of old growth (OG) and secondary forest stands of different ages. These regenerating forests are located on areas that were used for cattle pasture, swidden farming or fruit production between the 1880s and the establishment of BCNM in 1979 [34,35]. However, the amount of time these areas were used for grazing and agriculture is unknown [35].

The forest chronosequence used in this study consists of five forest stands that were 20, 40, 70 and 100 year-old secondary forests, and old growth forests when first censused in 1994. Vegetation in each stand was sampled in two parallel 160 m × 10 m transects (3200 m² per stand), with at least 20 m between transects. These transects were recensused in 2001 and 2011. Stand ages in 1994 were estimated using historical records, aerial photographs and interviews with long-time residents [34]. Since abandonment, these stands have experienced minimal human disturbance. Old growth forest sites were not altered or cultivated in the 500 years prior to this study [36]. The descriptions of the forest history, soil properties and geographical locations of the stands that make up the chronosequence can be found elsewhere [34].

All stems ≥5 cm diameter at breast height (DBH) were identified and measured; we define these stems as trees. Saplings (exceeding 1 m in height and less than 5 cm DBH) were identified and measured in one half of the transect (160 m × 5 m, 1600 m² per stand). One of the youngest stands (Saino) has only one transect and is interrupted by creeks in two places. Therefore, at this site, trees were measured in only one transect (1600 m²), with saplings measured in half this area (800 m²). In 1994, each stem was counted but not tagged, and it was not noted if multiple stems were of the same individual. In 2001 and 2011, saplings and trees were tagged, measured, identified and all multiple stems per individual tree were measured and counted. We analysed saplings and trees separately to detect whether different life stages show similar patterns of change across dispersal modes, even though the richness and abundance of smaller size classes can be more variable during succession than larger size classes [5].

We assigned one or more dispersal mode to each species of sapling and tree using existing methods [37]: flightless mammals, large birds (of 300 g or more), small birds (less than 300 g), bats and abiotic (which includes wind, water and explosive dispersal). To assign dispersal modes for other species we used published data [38–45] and field observations (S. J. Wright 2022, personal observation; electronic supplementary material, table S1). Flightless mammals include dispersers such as monkeys, opossums, pacas, agoutis and spiny rats [3]. Because species were often assigned to more than one dispersal mode, we analysed changes in each dispersal mode separately, unless otherwise noted. For trees and saplings separately, we calculated the proportion of species assigned to each dispersal mode for each transect and census (1994, 2001 and 2011). We also calculated the proportion of individuals per dispersal mode, but only for 2001 and 2011 because trees were not individually tagged in 1994 so we were unable to differentiate whether an individual had more than one stem. To evaluate if species had more than one dispersal mode during regeneration (i.e. redundancy), we calculated the proportional number of modes per species (e.g. 0.2 = dispersed by one of the five modes) and took the average across the proportions of all species per plot, resulting in a redundancy index where higher values indicated greater redundancy.

To determine the effect of forest age on the relative importance of each dispersal mode separately, we used generalized linear mixed models (GLMMs) with a beta distribution (since our proportion data do not equal 0 or 1). We modelled the proportion of species or proportion of individuals that could be dispersed by that mode as a function of continuous forest age (year of abandonment plus years between censuses), with forest stand as a random effect (two transects per stand). We used another set of GLMMs to determine the relative importance of all dispersal modes simultaneously against a baseline (abiotic) and tested whether changes in the proportions of each mode were significantly different against forest age. To do so, we tested for an interaction between forest age and dispersal mode. Given that we cannot accurately judge the age of old growth forest sites, we used GLMMs to test whether the proportion of species and individuals for saplings and trees in the old growth sites (baseline) significantly differed from categorical age classes of secondary forests (20, 40, 70, 100 years) for each dispersal mode. Finally, we evaluated whether the dispersal redundancy index increased with forest age and used the same GLMM described above. We fitted models using the function ‘glmmTBM’ from the package glmmTBM in R v. 4.2.1 [46].

3. Results

Across our three censuses (1994, 2001, 2011), we found 319, 244, 236 species of saplings, and 227, 203 and 210 species of trees, respectively. Most species of saplings and trees were dispersed by both small birds and flightless mammals, or by both large birds and flightless mammals (figure 1a, electronic supplementary material, figure S1). By contrast, few species were uniquely dispersed by large birds, flightless mammals or bats. Species that were abiotically dispersed were almost never dispersed by animals (figure 1a, electronic supplementary material, figure S1). The number of shared species among dispersal modes was similar across forest age classes (electronic supplementary material, figures S2–S3) except for the number of sapling species dispersed by both flightless mammals and large birds, which decreased from 62 in young forests to 42 in old growth sites (electronic supplementary material, figures S2–S3).

Figure 1.

Shared number of 388 species of saplings across (a) dispersal modes, and (b) forest age, on passively restored forests at the Barro Colorado Nature Monument, Panama. Intersections without values indicate no species shared among modes. Dispersal redundancy is the proportional average number of dispersal modes per species of saplings per plot. The modelled trendline and 95% confidence intervals (CIs) were derived from a GLMM with a Michaelis-Menten saturating function, and site as a random effect (z = 3.605, p < 0.01).

The effect of forest age varied among dispersal modes (figure 1b, electronic supplementary material, table S2). For flightless mammals, the proportion of sapling species and the proportion of individual trees increased significantly with forest age (electronic supplementary material, table S2). For large birds, the proportion of species and individual saplings, and the proportion of individuals trees, also increased significantly with forest age (electronic supplementary material, table S2). By contrast, the number of sapling individuals dispersed by small birds and the number of tree individuals dispersed abiotically significantly decreased with forest age (figure 1b, electronic supplementary material, table S2). We did not detect a significant change in the number of species or individuals dispersed by bats (figure 1b, electronic supplementary material, table S2). Redundancy of dispersal modes across species of saplings significantly increased with forest age (figure 2), but the change in magnitude was small; from 0.309 at 20-year-old forests to 0.385 at 100-year-old forests. Likewise, redundancy for saplings at the youngest stands compared to old growth forests was significantly lower, but the change in magnitude was small (electronic supplementary material, figure S4). For trees, redundancy across dispersal modes did not differ significantly across forest age but was significantly lower in 20- and 70-year-old stands versus old growth stands. Again, changes in magnitude were small, from 0.367 at 20-year-old forests to 0.385 at 100-year-old forests (electronic supplementary material, figures S5–S6).

Figure 2.

Proportion of individual trees for each dispersal mode (±SE) across forest age in 2001 at the BCNM, Panama. Rankings among modes did not change in subsequence census (data not shown). Asterisks represent significant differences (p < 0.05) from old growth forests (OG).

Interestingly, dispersal by flightless mammals was always more important than other dispersal modes in all forest ages (figure 1b and figure 3, electronic supplementary material, table S2). The importance of plant species dispersed by flightless mammals, large birds and small birds was significantly greater than abiotic dispersal, both in terms of proportion of individuals and proportion of species (figure 3, electronic supplementary material, table S3). The relative contribution of these modes varied with forest age—with flightless mammals and large birds increasing, and small birds decreasing (figure 3, electronic supplementary material, table S4). The proportion of individuals and species of saplings and trees dispersed by flightless mammals in secondary forests, as well as proportion of species dispersed by large birds at these sites, were often significantly lower compared to old growth forests (electronic supplementary material, table S4). The proportion of sapling species dispersed abiotically in the 40- and 70-year-old secondary forests was significantly lower than old growth forests (electronic supplementary material, table S4).

Figure 3.

Change in the proportion of species and individuals for saplings and trees per dispersal mode as predicted by GLMMs during forest regeneration from passively restored sites at the BCNM, Panama. Dashed lines represent the 95% CIs.

4. Discussion

Consistent with our predictions, the importance of species dispersed by flightless mammals and large birds increased with forest age, whereas species dispersed by small birds and abiotic means tended to decrease as forests regenerate. However, and in contrast to findings in other tropical post-agricultural landscapes [11,47,48], dispersal by flightless mammals remained the most prevalent dispersal mode across all forest ages (from 20-year-old to old growth forests). The prevalence of species dispersed by flightless mammals contrasts with other studies, where this group is usually present in low numbers across regenerating forests due to habitat loss, fragmentation or hunting [16,49,50]. We also found that many species of saplings and trees were dispersed by multiple taxa, and redundancy was high across the entire chronosequence. We consider our results an ‘ideal scenario’ for recovery because secondary forests within the BCNM have regenerated unhindered, due to its protected status and the presence of a biodiverse forested matrix.

Several landscape factors and site-specific characteristics across BCNM can help explain why seed dispersal by flightless mammals was prominent in our study. First, our regenerating sites have been surrounded by a forest matrix that includes large tracts of old-growth forests since abandonment, and have been protected for decades from land use change [34]. Second, when BCNM was created in 1979, hunting was prohibited [35], increasing the abundance of the flightless mammal community around the abandoned sites. Indeed, compared to agricultural matrices in Panama, BCNM has a diverse community of terrestrial mammals [51], and low hunting has significantly increased the abundance of large-seeded plants, which are usually dispersed by the mammals that hunters prefer [52]. For example, dispersal of large-seeded palms is higher in areas where the abundance of terrestrial and arboreal mammals is high thanks to low hunting pressure [52]. We lack data for the first two decades of forest regeneration, so we cannot assert the importance of seed dispersal by flighless mammals and other dispersers during that time period. However, given the well-preserved nature of our study site and the low levels of disturbance since abandonment, species dispersed by flightless mammals may have already been prevalent shortly after abandonment. Regardless, larger dispersers move across large areas, thus connecting different habitats across a landscape [53], and a diverse community of flightless mammals, plus the presence of mature trees, promotes regeneration at large spatial scales [54,55]. Therefore, promoting the recovery of dispersers of large-seeded plants, especially, can have knock-on effects for the wider landscape [54,55].

As predicted, the relative importance of large birds relative to small birds increased over time. During the first years of succession, small birds can disperse a disproportionately large number of rare species compared to the diversity of fruits available in the landscape [56]. With time, large birds usually become more important during regeneration [28]. In fact, a study quantifying bird communities at BCNM found that large birds increase in abundance in older forests, and most larger, canopy species have frugivorous or omnivorous diets [57]. Therefore, turnover in the bird community may explain why more tree species and individuals are dispersed by large birds in older secondary forests compared to small birds (figures 1 and 3). Finally, and contrary to our predictions, bat dispersal was not important throughout the sampling period. Dispersal by bats is key during the first years of succession [9,14,16,38], so perhaps we missed their contribution to the plant community at the onset of regeneration given that the youngest forests were already 20 years old when our sampling began. Studies assessing seed dispersal by animals in the first 20 years of regeneration indicate that bats and small birds are the most important dispersers [11,14,16,38,56,58].

Our unexpected observation that redundancy was high across all years probably stems from the fact that plants do not rely on a single species or a single group of frugivores for seed dispersal because it could be maladaptive [3]. Moreover, animals can have broad diets, and change their feeding preferences according to fruit availability, which in turn is seasonal and spatially distributed [3,59]. Therefore, the high number of plant species shared by flightless mammals and small birds, and by flightless mammals and large birds (figure 1, electronic supplementary material, figures S1–S3), indicates that fast seed deposition and successful plant establishment in post-agricultural landscapes are contingent on the richness and abundance of many dispersers and many seed sources [9]. When flightless mammals are in low abundance or absent, smaller dispersers such as small birds and bats can kickstart regeneration [14,16,56,60]. But ultimately, movement of large-seeded plants will depend primarily on larger disperses like monkeys, opossums, pacas, agoutis, guans and toucans, among others [28,48,52,54,61]. Finally, we cannot rule out that redundancy could have been lower in the first two decades of restoration compared to what we found at 20-year-old sites. Nevertheless, in a well-preserved system like BCNM, redundancy may be high from the onset of succession. Further studies should focus on how seed dispersal redundancy changes during the first decades of regrowth.

Considering the long history of studies assessing seed dispersal in the tropics [9,59], we propose a list of parameters that could be incorporated into forest restoration planning and monitoring to help researchers and practitioners evaluate the speed of ecological recovery during tropical restoration. Low-tier parameters require minimal expert knowledge at linking animal dispersers and their food plants and are relatively cheap and easy to measure (table 1). These could include the number of propagules arriving to restored sites, and censuses of larger frugivores at restored sites [13,48,54,57]. Intermediate tier parameters could start to differentiate which dispersers are more important and could assess seed removal from focal trees, or evidence of landscape connectivity by tracking seed movement to restoration sites (table 1) [7,16,25,48,58,62–65]. Higher-tier parameters, which require more detailed ecological knowledge and potentially more time investment and funding, could indicate the capacity of the frugivore community to disperse many plant species, or track properties of seed dispersal networks [16,20,60,63]. For example, tracking interactions among dispersers and rare plant species can elucidate the role that animals play in the recovery of plant populations during restoration [20,60]. Increasing the number of restoration sites that study seed dispersal will enable us to determine critical parameters required to describe the movement of frugivores, seeds and ultimately the recovery of ecological functioning.

Table 1.

List of parameters that can evaluate the recovery of seed dispersal interactions among animals and plants during forest restoration in tropical forests. Expert knowledge and ease of implementation (e.g. costs) depend on the local context where restoration initiatives take place. Please see §4 on how different parameters can help researchers and practitioners evaluate the speed of ecological recovery during restoration.

tier	parameter	expert knowledge	ease of implementation	references
low	total seed number in seed rain	low	low–medium	[13,48,54]
	abundance of frugivores essential for the dispersal of late-successional species (i.e. large birds or flightless mammals)	low	low	[28,57]
intermediate	total number of seeds from native species dispersed by different modes	medium	medium–high	[25,62]
	diversity and species richness of the seed rain	medium	medium–high	[14,58,63]
	number/proportion of seedlings/saplings/trees dispersed by various modes	medium	medium	[7,16,64]
	frugivories of specific plant species or animal disperser	medium	medium	[48,65]
High	changes in different dispersal modes during restoration through association of tree community composition and animal diets.	medium–high	low	([16,24,63], our study)
	redundancy among dispersers	low	high	(our study)
	network analyses that relate specific plant species with specific dispersers	high	high	[20,60]

In conclusion, our uniquely long-term dataset shows that dispersal by flightless mammals and large birds can surpass the importance of other dispersal modes as forests regenerate. The importance of flightless mammals and large birds at our site is probably due to the presence of large tracts of preserved forests contiguous to young secondary forest, low hunting, and high functional redundancy across dispersers. Therefore, our study highlights two constraints for restoration: hunting of flightless mammals, which are key dispersers of many species including late-successional large-seeded plants, and proximity to large tracts of forests, which provide habitat for large dispersers and guaranteed seed sources for small and large dispersers. Our study can serve as a benchmark for an ‘ideal scenario’ where high landscape connectivity accelerates the recovery of interactions among seed dispersers and plants, particularly those responsible for the dispersal large-seeded species. Assessing how different animal taxa contribute to plant community turnover during passive restoration in tropical forests can be useful for ensuring the success of adaptive ecological restoration.

Data accessibility

The data are provided in electronic supplementary material [66].

Authors' contributions

S.E.-V.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, writing—original draft, writing—review and editing; P.R.S.: conceptualization, methodology, supervision, writing—review and editing; O.L.: data curation, funding acquisition, methodology, supervision, writing—review and editing; S.J.D.: data curation, funding acquisition, project administration, supervision; L.S.C.: conceptualization, formal analysis, funding acquisition, methodology, supervision, writing—original draft, writing—review and editing; D.H.D.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

The establishment of the forest plots was funded by the NSF (grant no. DEB-9208031) awarded to Julie S. Denslow. Following forest censuses were funded by Secretaría Nacional de Ciencia, Tecnología e Innovación (SENACYT), Panama, International Collaboration Grant COL10-052 awarded to D.H.D. and O.L. S.E.-V. was supported by the Cullman Fellowship from the School of the Environment at Yale University and the New York Botanical Garden.