Large mammalian herbivores modulate plant growth form diversity in a tropical rainforest

Abstract

en

1. The world’s terrestrial biomes are broadly classified according to the dominant plant growth forms that define ecosystem structure and processes. Although the abundance and distribution of different plant growth forms can be strongly determined by factors such as climate and soil composition, large mammalian herbivores have a strong impact on plant communities, thus defaunation (the local or functional extinction of large animals) has the potential to alter the compositional structure of plant growth forms in natural ecosystems.
2. Tropical rainforests sustain a high diversity of growth forms, including trees, palms, lianas, shrubs, herbs and bamboos, all of which play important ecosystem functions. Here, we experimentally evaluate how large mammalian herbivores affect the dominance, diversity and coexistence of these major tropical forest plant growth forms, by monitoring communities of saplings on the understorey in 43 paired exclusion plots in a long-term replicated exclusion experiment in the understorey of the Atlantic forest of Brazil.
3. Over the course of 10 years large herbivore exclusion decreased diversity among growth forms, increased the absolute abundance of palms and trees (22% and 38% respectively) and increased the diversity of species within these two groups, to the detriment of other growth forms. Furthermore, all pairwise relationships between growth forms were positive on plots where herbivores had access, whereas several strong negative relationships emerged in plots where herbivores were excluded. This occurred despite strong background directional temporal trends affecting plant communities in both experimental treatments across the region.
4. Synthesis. Our work indicates that the defaunation alters growth form dominance by favouring palms and trees while eroding diversity among growth forms and coexistence on a temporal scale. Large herbivore mammals promote diversity among growth forms, preventing the hyper-dominance of trees and palms, yet without supressing the diversity of species within growth forms. We argue that large herbivore mammals affect growth forms through several non-mutually exclusive mechanisms, including herbivory, seed dispersal and physical disturbance, as well as differential effects linked to the morphological and physiological adaptations of growth forms. We conclude that defaunation might lead to profound impacts on important ecosystem functions underpinned by growth form diversity, and result in vertical and horizontal structural simplification of tropical rainforests.

Resumo

lt

1. Os biomas terrestres ao redor do mundo podem ser amplamente classificados de acordo com as formas de vida das plantas, as quais podem definir a estrutura e os processos ecossistêmicos. Embora a abundância e distribuição de diferentes formas de vida de plantas possam ser fortemente determinadas por fatores como clima e composição do solo, os grandes mamíferos herbívoros também têm um forte impacto nas comunidades vegetais, assim a defaunação (a extinção local ou funcional de grandes animais) tem o potencial de alterar a composição das formas de vida das plantas em ecossistemas naturais.
2. As florestas tropicais sustentam uma alta diversidade de formas de vida de plantas, incluindo árvores, palmeiras, lianas, arbustos, ervas e bambus, as quais desempenham importantes funções ecossistêmicas. Nesse trabalho, avaliamos experimentalmente como grandes mamíferos herbívoros afetam a dominância, diversidade e coexistência dessas principais formas de vida de plantas das florestas tropicais, monitorando comunidades de plântulas no sub-bosque em 43 pares de parcelas em um experimento de exclusão de longo prazo no sub-bosque da Mata Atlântica brasileira.
3. Nós encontramos que ao longo de 10 anos, a exclusão de grandes herbívoros diminuiu a diversidade entre as formas de vida das plantas, aumentando a abundância absoluta de palmeiras e árvores (22% e 38%, respectivamente) e aumentando a diversidade de espécies dentro desses dois grupos, em detrimento da outras formas de vida. Além disso, todas as interações pareadas entre as formas de vida foram positivas nas parcelas onde os herbívoros tiveram acesso, enquanto várias relações negativas fortes surgiram nas parcelas onde os herbívoros foram excluídos. Discutimos que isso ocorreu devido a vários fatores temporais atuando simultaneamente, que afetaram as comunidades de plantas em ambos os tratamentos experimentais em todas as regiões.
4. Nosso trabalho indica que numa escala temporal a defaunação altera a dominância das formas de vida das plantas, favorecendo palmeiras e árvores enquanto diminui a diversidade e a coexistência entre as os outros grupos de plantas. Os grandes mamíferos herbívoros tiveram um papel essencial em promover a diversidade entre esses grupos de plantas, evitando assim a hiperdominância de árvores e palmeiras sem suprimir a diversidade de espécies dentro das formas de vida. Nós discutimos que esses herbívoros afetam as formas de vida de plantas através de vários mecanismos, incluindo herbivoria, dispersão de sementes e distúrbios físicos, bem como através de mecanismos relacionados às adaptações morfológicas e fisiológicas desses grupos de plantas. Nós concluímos que a defaunação pode levar a impactos profundos e importantes nas funções ecossistêmicas mantidas pela diversidade de formas de vida de plantas, resultando numa simplificação estrutural vertical e horizontal das florestas tropicais.

1 INTRODUCTION

Through the course of its evolution, the plant kingdom has diversified, adopting a variety of life-history strategies and traits to adapt to different environments (Clarke et al., 2011). Such strategies have been crucial in allowing plants to thrive in almost every terrestrial ecosystem on Earth (Kreft & Jetz, 2007; Woodward & Williams, 1987). At the morphological level, the vast diversity in plant strategies and traits can be simplified in different plant growth form categories, each one defining broad strategies and structural adaptations common to all species within such growth form. In those terms, the growth form of any given plant species can be considered as the outcome of its adaptive evolution history resulting from its interaction with the biotic and abiotic environment over an evolutionary time-scale (Ewel & Bigelow, 1996). Beyond their adaptative value, growth forms have an important functional value, so that growth form composition influences both ecosystem structure and processes (Ewel & Bigelow, 1996).

Different biomes, such as tropical and temperate rainforests, grasslands, shrubland and tundra, are broadly classified based on the dominance of different plant growth forms, stressing the strong influence of these growth forms on ecosystem processes. For example, tropical rainforests are dominated by trees that stock large amounts of carbon and contribute to the litter layer with their leaves, branches and fruits, promoting feedbacks on decomposition, soil nutrient cycling and material flows (Ewel & Bigelow, 1996; Ghazoul & Sheil, 2010). In contrast, fallen fronds from palms with large volumes and slow decomposition rate can prevent plant recruitment in the understorey by physically damaging seedlings, suppressing the germination of many plants, and slowing down decomposition (Beck, 2017). Thus, plant growth forms dominance and diversity might influence ecosystem dynamics and function (Ewel & Bigelow, 1996).

Tropical rainforests hold the highest diversity of plants on the planet with a rich variety of life histories, growth forms and growth strategies, and the variation among these growth forms in abundance and richness can be useful to access different forest formations and successional stages among Neotropical forests (Alves et al., 2010; Andersen et al., 1997; Denslow, 1996; Kreft & Jetz, 2007). Trees, shrubs, lianas, palms, epiphytes and other plant growth forms coexist in a vertically and horizontally stratified and structurally complex ecosystem, where they engage in a myriad of biotic competitive, mutualistic and antagonistic interactions (Andersen et al., 1997; Ewel & Hiremath, 2005; Ghazoul & Sheil, 2010). Natural and anthropogenic disturbances, such as climate change, forest fragmentation and selective logging might favour some growth forms to the detriment of others (Dupuy & Chazdon, 1998; Feeley et al., 2020; Püttker et al., 2020; Wright et al., 2007). The local or functional extinction of wildlife, a phenomenon called defaunation (Dirzo et al., 2014; Püttker et al., 2020), might also affect the diversity and coexistence among plant growth forms yet its effects on growth forms composition and diversity have been poorly explored. These animals sustain functions that cannot be replaced by other faunal groups and are the most susceptible to any unsustainable human activity, with knock-on consequences on ecosystem dynamics and functioning (Dirzo et al., 2014; Gardner et al., 2019; Püttker et al., 2020).

There is growing evidence suggesting that large mammalian herbivores (LMHs) modulate plant species composition and community interactions in these ecosystems. LMHs consume large amounts of fruits and seeds in tropical forests, hence acting as seed predators and dispersal agents, although the net effects of the combination of these functions are still unresolved (Bodmer & Ward, 2006; Kurten & Carson, 2015; Villar et al., 2020; Williams & Brodie, 2020). LMH also consume seedlings and damage plant tissues and physical structures through browsing, trampling and by overturning large amounts of soil (Coley & Barone, 1996; Villar, Paz, et al., 2021). Furthermore, LMHs prevent the proliferation of smaller mammalian herbivores (Galetti, Guevara, Neves, et al., 2015) and affect nutrient cycling in tropical forests (Villar, Paz, et al., 2021). Thus, LMH affects plant communities in tropical rainforests through many direct and indirect pathways, all of which can have a strong impact on plant recruitment, dominance, diversity and composition, most likely benefiting some growth form adaptations to the detriment of others (Kurten et al., 2015; Luskin et al., 2019; Malhi et al., 2016; Villar et al., 2020; Villar & Medici, 2021).

Studies addressing defaunation effects on tropical rainforest growth form composition and diversity are few, mainly focused on the balance between trees and lianas, and sometimes contradictory. For instance, defaunation has been shown to promote lianas within the seedling banks in a Central American rainforest (Wright et al., 2007). This contrasts with the findings of another study from Southeast Asia, where the experimental exclusion of artificially abundant large herbivore densities has led to a temporal decline in lianas to the benefit of trees (Luskin et al., 2019). Such contradictory results highlight the need for a broader evidence base that addresses the impacts of LMH on growth form composition, diversity and coexistence. This is also important from a functional perspective. For example, lianas have lower carbon storage capacity when compared to trees (van der Sande et al., 2013), and therefore an increase in lianas cannot compensate for the deficits from a concomitant decrease in other carbon-rich growth forms. Yet, by an excessive focus on the dichotomy of lianas versus trees, the limited number of studies available neglect the potentially important effects of defaunation on other dominant tropical growth forms, such as palms, and others at the understories including shrubs, herbs and bamboos. Evidence suggests that the spatial distribution and demography of palms, for example, might be affected by defaunation (Fragoso et al., 2003; Galetti et al., 2006; Portela & Dirzo, 2020; Valverde et al., 2020), but less is known about other ‘lesser’ growth forms.

In most tropical Atlantic forests of South America, the losses of medium- and large-sized mammals by hunting and forest deforestation have simplified most mammal communities to a few herbivores species with <30 kg in weight (Bogoni et al., 2020; Gonçalves et al., 2018; Souza et al., 2019). Even on most locations where species with >30 kg are present, population sizes are so small that these might play an inefficient functional role as seed consumers and dispersers, with a knock on effects on plant dominance (Villar et al., 2020) or more generally, plant–vertebrate interactions through direct and indirect pathways (Galetti et al., 2021; McConkey & O’Farrill, 2015). This may lead to a forest structural simplification by favouring the hyper-dominance of a few plant growth form groups over others (Luskin et al., 2019; Malhi et al., 2016; Wright et al., 2007).

Here we address how defaunation impacts upon plant growth form dominance, diversity and coexistence in tropical rainforests focusing on the early ontogenetic stages of plants (saplings). Growth forms inspected include trees, palms, lianas, bamboos, shrubs and herbs. We used a long-term multi-site replicated large mammalian herbivore exclusion experiment to investigate if defaunation in tropical rainforests affects (i) absolute and relative abundances of different plant growth forms, (ii) pairwise relationships between coexisting growth forms and (iii) diversity among and within growth forms. Our naïve hypothesis is that lianas will become the dominant growth form in defaunated rainforests since they benefit at different degrees from alternative seed dispersal strategies (wind or small birds), low seed predation rates and low mechanical damage that occurs in the absence of LMH (Wright et al., 2007). We also expect that trees and palms will be favoured by defaunation since in tropical forests these are the growth forms most consumed by large herbivores (Beck et al., 2013; Theimer et al., 2011). Consequently, we expect that by increased dominance of a few growth forms defaunation will negatively affect growth form diversity and coexistence (Theimer et al., 2011; Villar, Paz, et al., 2021).

2 MATERIALS AND METHODS

2.1 Study sites

The study is part of the DEFAU-BIOTA, a long-term experiment on large mammalian herbivores (>10 kg) exclusion in the Atlantic forest of Brazil (https://souzayuri.shinyapps.io/biota/). The experiment is based on four old-growth dense ombrophily evergreen moist forest regions in the Atlantic forest in Southeast Brazil that are within the same protected area Serra do Mar State Park (Brazilian permit number COTEC No. 43.104/2007). Despite the geographical distance, these areas share similar vegetation composition with large trees and a diverse understorey and, even with a past history of degradation, they are considered as late-successional or mature forests sharing also a similar background assemblage of mammals species, including herbivores, carnivores and seed dispersers (Alves et al., 2010; Brocardo et al., 2012; Galetti et al., 2017; Rocha-Mendes et al., 2015; Souza et al., 2019; Villar et al., 2020). Two of them, Itamambuca and Vargem Grande field station (ITA and VGM respectively), located at Santa Virgínia, are in the largest continuous reserve area throughout the Atlantic forest (332,000 ha). The other two sites are Carlos Botelho State Park (CBO, 37,633 ha) and Ilha do Cardoso State Park (CAR, 13,500 ha) (Figure S1).

The altitude in these areas varies from sea level to 975 m a.s.l., with the annual temperature average around 20–24°C and precipitation around 1500–4000 mm. In these sites, tapirs Tapirus terrestris and white-lipped peccaries Tayassu pecari are the largest ground-dwelling LMHs (Galetti et al., 2017). Agoutis Dasyprocta spp., pacas Cuniculus paca, brocket deer Mazama spp., and collared peccaries Pecari tajacu, among other ground-dwelling frugivorous species, are also present at the sites, as well as arboreal primates (e.g. howler monkey Alouatta guariba, woolly spider monkey Brachyteles arachnoides, capuchin monkeys Sapajus nigritus), although the effects of these are not the focus of the study. For more information about respective areas and sample design see Supplemental Material and Villar et al. (2020).

2.2 Experimental plots and plant sampling

In 2009, 15 exclusion and open control paired plots, separated by 2–7 m, were established in each of the four areas, totalling 60 exclusion and 60 control plots sampled semi-annually, twice yearly. Each pair plot was located at least 200 m from the closest pair. The exclusion plots were fenced using a 1.6 m high metal fence, with 5 cm wire mesh, barring the entrance of all medium- and large-sized ground-dwelling mammals but allowing access to small rodents and marsupials. The control plots were open, only demarcating their areas with a line and plastic piles. Due to the natural falling of trees by natural uncontrolled causes, some plots were damaged and lost. In this work, we used data from the remaining 86 individual plots (43 pairs) that survived to October 2019. CBO was the first sampled area, starting in 2009, and the other three areas started in 2010; thus, we standardized our dataset to the minor common number sampled between the areas, 108 months and 19 visits.

Both exclusion and control plots were 5 × 3 m (15 m²) and each plot was divided by eight 1 m² subplots. We sampled only three 1 m² subplots from each plot and used the other ones for fieldwork manipulations and access, avoiding human damage or disturbance on the main sampled plots. We focused on sampling all the plant early-stage saplings inside the three 1 m² subplots (defined as young plants ranging between 0.1 and 1 m in vertical height), since early-stage dynamics are critical for the dynamics and composition of tropical forests (Green et al., 2014). The horizontal instead of vertical stem length was considered for creeping liana species that were not climbing. We considered in our experiment the pre-existent individuals that were already present before the exclusion and all the newly recruited plants in each sampling visit that was in agreement with our height criteria. Every individual was tagged using an aluminium tag containing a unique number as a registered code, and these tags were tied at the base of the plant by a wire with enough space to avoid the plant constriction by its growth. In this way, we could track the fate of all the old and new individuals in each visit and along with the entire duration of the experiment.

For each sapling, we identified the family and species and classified it as one of the following six growth forms: lianas, palms, trees, herbs, shrubs or bamboos (Table 1; Figure S2). No palms with liana habit (e.g. Desmoncus spp.) occurred in our sites. All the individuals were identified in the fieldwork by checking the family, genus and/or species attribute or taking photos for posterior confirmation. Over the 10 years we sampled a total of 8729 sapling individuals and identified 169 plant species (Table S1). A total of 1629 individuals were removed from our analyses; these could not be identified by taxonomy or their growth form because saplings did not have floral structures or fruits during this stage that could facilitate their identification, or yet if they were too uncommon to be included in our analyses (e.g. ferns). Since some plant families or species have more than one growth form during their lifetime, we used the sapling growth form description from (Souza & Lorenzi, 2012) and the ‘Flora do Brasil’ (http://floradobrasil.jbrj.gov.br/) for classifying these individuals.

TABLE 1. Growth forms general description and main biological features in the Atlantic forest, Brazil. The references on this description can be accessed on the supplemental material (Supplementary references). Photo credits: Rafael Souza

Trees		Trees have wood and bark material that provides most of the mechanical support, guiding them to a vertical distribution with individuals of tall species and different statures (Jeronimidis, 1980; Gonzalez de Tanago, 2018). This vertical composition may reflect the relative illumination at different heights above the forest floor (Turner, 2004). The horizontal variation in light availability due to irregularity in canopy structure (presence of gaps, different heights, etc.) could also result in the subtle environmental variations providing many niches variation for trees, and also other growth forms, as the epiphytes (Kohyama, 1993). Trees fruits and seeds are consumed and dispersed abiotically and biotically by many animal groups (Parolin et al., 2013; Sinha & Davidar, 1992)
Palms		Tropical palms may be tall enough to be emergent and to form a part of the canopy, or they may be understorey species of short stature adapted to shady conditions (Johnson, 2011). Some palm species may require from a minimum of 2 years to 40 years or more to reach maturity, and begin to flower and produce fruits (Johnson, 2011). Palms are dominant in most tropical ecosystems, and are often a key functional component, forming complex assemblages, comprising coexisting growth forms and occupying all layers of the forest (Balslev et al., 2016). Palms also produces a huge amount of fruits, and zoochory is the common mode of dispersal (Macía, 2004; Zona & Henderson, 1989)
Lianas		Lianas are recognized by their unique ability to use other plants to climb to the forest canopy, producing large numbers of leaves that cover their hosts, generally trees, palms and shrubs, thus competing for light resources (Schnitzer et al., 2014). By their tangles, lianas provide critical inter-crown pathways in the canopy and in the understorey, providing structure as refuges, nests and connections for nonvolant arboreal and terrestrial animals (Yanoviak, 2014). Lianas also produce leaves, flowers and fruits which may be critical for the survival of many animal species, especially during the dry season when lianas grow faster and produce more leaves and fruits than trees and palms (Arroyo-Rodríguez et al., 2014)
Shrubs		Shrubs can be distinguished as small woody plants either with multiple stems or with branching very close to the ground (Turner, 2004). Woodiness must be counted as an advantage in providing reduced palatability to herbivores. The short and medium stature of shrubs reduces transpiration costs under drought conditions, and susceptibility to being damaged from winds and storms (Stutz, 1989). The success and arrangements of shrubs also depend on their interaction with animals as pollinators, frugivores and herbivores, since many species of shrubs have attractive, succulent or nutritive fruits (McArthur, 1989)
Herbs		Herbs are very abundant in tropical forests since this forest has high heterogeneity, opportunities and niche diversification, allowing many herbs life-history strategies to have success in an environment with low light conditions (Vieira et al., 2015). Herbs spend their entire life cycle in the forest understorey continually subject to mammalian impacts, including trampling and herbivory (Royo & Carson, 2005). Their population structure is temporally variable due to variation in factors such as nutrients, weather and natural disturbances (Goodwillie & Jolls, 1996; Fröborg & Eriksson, 1997; Whigham, 2004). Herb resources as pollen, leaves and fruits are important for the food web of many herbivores, small frugivores vertebrates and invertebrates (Schemske et al., 1978; Whigham, 2004)
Bamboos		Bamboos present great plasticity in their physiological traits, being adapted to live in disturbed forests and human-modified environments (Clark & Oliveira, 2008). Below-ground, bamboos produce extensive rhizome networks that allow for clonal growth and rapid resprouting response after disturbances (Griscom & Ashton, 2006). Life cycle length is highly variable among species, and some bamboos species have a gregarious monocarpic life cycle in which an entire population flower, produces seeds, and then subsequently dies (Franklin, 2004; Fadrique et al., 2020). Their structure can provide critical habitat for specialized invertebrates, birds and mammals (Dunnum & Salazar-Bravo, 2004; Rother et al., 2013)

2.3 Statistical analyses

To investigate the effects of the exclusion of LMH on plant growth form abundances and diversity we fitted several generalized linear mixed model (GLMM) statistical modelling approaches. As the focus in many of our analyses was to test for net differences in the temporal trajectories caused by experimental treatment effects rather than describing the best shape of each individual trajectory, we did not attempt to fit nonlinear functions to those trajectories. First, we tested for differences between experimental treatments in the absolute abundance, summing the individuals sampled in each period for every growth form separately. We expected that experimental exclusion of LMH would lead to a divergence in the trajectories of the time series of abundances of open and closed plots for every growth form. Thus, the absolute abundance of every growth form was modelled as the function of the interaction between treatment and time, the number of months since the onset of the experiment (e.g. 0–108 in twice-a-year steps), as predictors (fixed effects). This approach has the advantage of accounting for pre-experimental differences in intercepts between treatments, while allowing for corrections for any background long-term trend across both treatments (see, e.g. Figure 1e). To account for the nested nature of our experimental design and ensure pairwise comparisons between treatments, we included site, plot and survey occasion as random effects in all GLMMs. For absolute abundances, we used GLMMs with a Poisson distribution (log link), or a negative binomial distribution when the distribution showed zero inflation.

Second, since LMH modifies the environment in many ways, defaunation may directly or indirectly affect the competitive abilities of some growth forms relative to others. To test this hypothesis, we modelled differences between treatments in the relative abundances, dividing the absolute abundance of every growth form by the community abundance (e.g. sum of all other growth forms), following a similar statistical approach to the one described above, except that in this case a binomial distribution (logit link) was used (zero-inflated models were fitted when necessary). In addition, we tested if the experimental exclusion of LMH affected pairwise abundance correlations between growth forms so as to elucidate shifts in the relative competitive and coexistence ability of growth form pairs between treatments. Thus, we modelled the absolute abundance of every growth form as a function of the log abundances of all other growth forms (after adding the value of one), one pairwise correlation at a time. We used a Poisson distribution, or a negative binomial distribution when the distribution showed zero inflation and the same random effects as for all other models.

Third, we tested how LMH exclusion affected diversity among and within growth forms (i.e. diversity of growth forms, and diversity of species within every growth form respectively). Our hypotheses pivot around LMH effects on dominance; thus, diversity indexes emphasizing evenness in the community (such as the inverse Simpson index [InvSimp], an index considering both community richness and proportions but emphasizing rare species or growth forms are most adequate to test such hypotheses (Chao et al., 2014). However, other indexes allocate no or intermediate weighting to dominance and proportions (species richness and the Shannon diversity index respectively; Chao et al., 2014), so that by comparing results from those alternative indices allows to discern whether shifts in diversity are due to changes in the number of species or dominance by a few species. Thus, to examine diversity among growth forms first we calculated the InvSimp by pooling the abundances of all individuals within every growth form irrespective of the species; subsequently, we examined the effect of treatment on the diversity among growth forms by modelling InvSimp as a function of the interaction between treatment and time as predictors (fixed effects) and fitted the same random effects as for other models. The InvSimp within every given growth form was calculated for each growth form considering the species richness and the absolute abundance of every species classified into this growth form. Subsequently, we modelled the InvSimp into each (within) growth forms with identical predictors and random effects as above. A similar procedure was applied to the other two indices for both diversity among and within growth forms. In all cases, indices were modelled using a Gaussian distribution.

We report estimates and statistical significance for single terms in the models and results from likelihood ratio tests of the treatment × time interactions. Time (months) was log-transformed in all models as exploratory analyses indicated that this was appropriate (we added the value of 1 before log-transformation). All analyses were performed in r (R Core Team, 2017), using packages ‘lme4’ and ‘glmmTMB’ for GLMMs (Bates et al., 2014; Brooks et al., 2017; R Core Team, 2017), ‘stats’ for likelihood ratio tests (Fox & Weisberg, 2018) and ‘hillR’ to extract InvSimp values (Li, 2018). See Supplemental Material for model structures.

3 RESULTS

During 108 months (from July 2009 to October 2019), we monitored the annual survival of 7100 plants: 2118 trees (29.83%), 2018 palms (28.42%), 1480 herbs (20.84%), 749 lianas (10.54%), 407 shrubs (5.73%) and 328 bamboos (4.61%) in both open and LMH exclusion plots in four Atlantic rainforests of Brazil.

3.1 Effects of large herbivore exclusion on the absolute abundance of plant growth forms

We found that the absolute abundances of trees and palms were favoured by LMH exclusion [trees: χ² (1, N [number of plots × sample period] = 1634) = 73.25, p < 0.001; palms: χ² (1, N = 1634) = 206.82, p < 0.001]. Tree abundances did not increase over time in the open plots, but by the 108th month, their abundances on closed plots were on average 38% higher than at the onset of the experiment (Figure 1a; Table S2). Exclusion also favoured palm abundances, but in this case, the difference between treatments was due to a decline over time on open plots but not on closed ones (Figure 1b; Table S2). The trajectory of palms on closed plots changed over the course of the experiment, with an initial demographic explosion (43% increase by the 36th month) followed by a net decrease roughly parallel to the open treatment. Contrary to trees and palms, the absolute abundance of lianas increased over time, but the increase was larger on open plots (74%) [χ² (1, N = 1634) = 5.57, p = 0.01, Figure 1c and Table S2]. Neither shrubs [χ² (1, N = 1634) = 1.14, p = 0.28, Figure 1d] nor bamboos [χ² (1, N = 1634) = 2.38, p = 0.12, Figure 1f] experienced any net temporal increase or divergence in the trajectory between treatments. Herbs increased in both treatments over the course of the experiment, and although there was a tendency for a larger increase in open plots, differences between treatments were not significant [χ² (1, N = 1634) = 3.35, P = 0.067, Figure 1e].

FIGURE 1

Mean absolute abundances of plant growth forms over the course of the 108-month experiment in the 43-paired plots in the Atlantic forest, Brazil. The Y-axes represent the absolute number of saplings for each plot area, 3 m². Dark green = open plots, dark purple = closed plots, shaded area = standard error. The ‘*’ represents the significative results from our models

3.2 Dominance: Effects of large herbivore exclusion on plant growth form relative abundances

In order to assess how defaunation affects growth form dominance, we examined the response of relative abundances of different plant growth forms to experimental exclusion over the course of the experiment. Neither trees nor bamboos experienced a net increase in relative abundances across time in any of the treatments [χ² (1, N = 1634) = 1.78, p = 0.18 and χ² (1, N = 1634) = 0.03, p = 0.84, respectively, Figure 2a, Table S2]. Relative abundances of palms were favoured by experimental mammal exclusion [χ² (1, N = 1634) = 34.20, p < 0.001, Figure 2a, Table S2]. The relative abundance of palms decreased on both treatments, but in open plots this decline was steeper than in closed ones. Conversely, the relative abundance of lianas increased in the open plots, but slightly decreased in the closed treatment [χ² (1, N = 1634) = 34.6, p < 0.001, Figure 2a, Table S2]. The relative abundance of shrubs decreased on closed plots but did not change on controls [χ² (1, N = 1634) = 4.88, p < 0.02, Figure 2a, Table S2], while herbs increased in open plots more than on closed plots [χ² (1, N = 1634) = 15.63, p < 0.001, Figure 2, Table S2].

FIGURE 2

The relative abundance of growth forms and pairwise relationships between growth forms in the LMH exclusion experiment at the Atlantic forest (Brazil). (a) Model estimates (mean and SE) of temporal changes in the relative abundance of every growth forms in every treatment. (b) An illustrative example of an open plot (left and closed plots (right). In (c) we show the strength (line thickness) and sign (colour, green = positive; pink = negative) of pairwise correlations between growth forms (only statistically significant correlations [p < 0.05] are displayed). And in (d) an illustrative representation of the community structure expected in the presence of LMH (left) and under a defaunated scenario (right), according to the proportion of each growth form by the end of the experiment. The ‘*’ represents the significative results from our models

3.3 Coexistence: Effects of large herbivore exclusion on pairwise abundance relationships between growth forms

In order to assess how defaunation affects the coexistence among growth forms, we examined how the pairwise relationships in absolute abundances between growth forms responded to experimental exclusion (see Section 2 for details). All possible growth forms pairwise abundance correlations on open plots had positive or neutral effects (Figure 2c; Table S4). In contrast, there were six pairwise negative growth form correlations in the closed treatment. In addition, mammal exclusion significantly shifted the slopes of four pairwise correlations towards less positive values, but only one pairwise correlation (herbs vs. bamboos) towards more positive values (Table S4). There were four non-significant pairwise correlations in the open plots and three on closed plots. Close examination of results suggested that pairwise correlations involving trees were of lower magnitude, either positive or negative, regardless of treatment. Correlations involving lianas were less strong on closed plots, yet only one (lianas-shrubs) was negative. Correlations involving palms, bamboos and herbs (in increasing order) shifted towards more negative magnitudes with mammal exclusion. All pairwise correlations with shrubs on the closed treatment were negative (but neutral with trees).

3.4 Diversity: Effects of large herbivore exclusion on species evenness within and among growth forms

In order to assess how defaunation affects dominance among and within growth forms, we examined, respectively, how growth form and species InvSimp within every growth form category (growth form and species evenness respectively) responded to experimental exclusion. Experimental exclusion of large mammals had a strong impact on the evenness of growth forms found in a given plant community. The InvSimp results shown that evenness among plant growth forms increased substantially on open plots over the course of the experiment (mean [SE] = 0.041 (0.016); χ² (1, N = 19) = 6.87, p = 0.008), but not on closed plots (mean [SE] = −0.044 (0.022); χ² (1, N = 19) = 3.90, p = 0.048), so that the gap in the diversity of growth forms increased with time (Figure 3a). Growth form richness was not affected by large mammal exclusion (open: mean [SE] = 0.010 (0.014); χ² (1, N = 19) = 0.52, p = 0.46; closed: mean [SE] = 0.003 (0.021); Χ² (1, N = 19) = 0.03, p = 0.85, Figure 3c), neither was Shannon diversity (open: mean [SE] = 0.008 (0.005); χ² (1, N = 19) = 2.86, p = 0.09; closed: mean [SE] = −0.005 (0.007); χ² (1, N = 19) = 0.63, p = 0.42, Figure 3d). Yet, differences in the trajectories of this index between both treatments showed an intermediate behaviour to those of the other two indices, strongly supporting that LMH lead to a shift in dominance among growth forms rather than growth form richness.

FIGURE 3

Species diversity among and within growth forms in the LMH exclusion experiment at the Atlantic forest (Brazil). (a) The temporal trajectory of the mean inverse Simpson index among growth forms in the 43-paired plots (shaded areas show standard errors). (b) Model estimates (mean and SE) of temporal changes in the inverse Simpson index within every growth form in every treatment. (c) Richness among growth forms over the course of the experiment, and (d) the Shannon diversity index among growth forms. In both figures dark green = open plots and dark purple = closed plots. The ‘*’ represents the significative results from our models

The temporal trajectories and the impact of experimental exclusion on evenness (InvSimp) within trees, lianas, shrubs, herbs and bamboos tend to follow the same qualitative trends as the ones described for absolute abundances (Figure 3b; Table S2). However, none of the growth forms showed a statistically significant evenness decrease over the course of the experiment on any of the treatments. Evenness within trees strongly increased over time on closed plots, but not on control plots, so that the gap between the open and closed treatments increased over time [χ² (1, N = 1634) = 25.26, p < 0.001, Figure 3b and Table S2]. Within palms, evenness also experienced an increase on closed plots, but not on open plots [χ² (1, N = 1634) = 4.31, p = 0.03, Table S2]. In contrast, lianas and herbs showed a significant and positive evenness change on both treatments [χ² (1, N = 1634) = 0.34, p = 0.55, and χ² (1, N = 1634) = 0.11, p = 0.73, respectively, Table S2]. Neither shrubs nor bamboos experienced any temporal change in evenness on any of the treatments [χ² (1, N = 1634) = 1.78, p = 0.18, and χ² (1, N = 1634) = 0.30, p = 0.57, respectively, Table S2]. In general, results for species richness and Shannon diversity within growth forms followed similar patterns as the InvSimp results, with some differences in the significance levels (but not sign) of exclusion effects (Table S3).

4 DISCUSSION

Our results indicate that the defaunation shifts the dominance, impacts upon the coexistence and erodes the diversity of plant growth forms in tropical rainforests. Over the course of the 10-year experiment, the exclusion of large mammals favoured palms and trees, to the detriment of other growth forms, especially lianas, by decreasing their absolute and relative abundances, rejecting our hypothesis. This occurred despite strong background temporal trends affecting sapling communities in both treatments in our experiment. We argue that this shift in dominance and reduction in diversity is probably a product of the interaction of several mechanisms operating simultaneously, including trophic (such as herbivory, seed predation and dispersal) and non-trophic (trampling) processes.

4.1 Mechanisms for LMH impacts on plant growth form communities

LMH exclusion increased the total abundance of trees and palms and reduced the total abundance of lianas. One intuitive mechanism that might explain this result is the decline of seed predation on LMH exclusion plots. In the Neotropics, large frugivores preferably feed on large fruits and seeds which are more common among trees than in other growth forms (Wright et al., 2007). As a consequence, seed predation by LMH is thought to strongly reduce tree recruitment and population densities (Beck et al., 2013; Kurten et al., 2015; Kurten & Carson, 2015; Theimer et al., 2011). Indeed, while diversity and abundance of small mammal herbivores (rodents) increase in defaunated tropical forests (Galetti, Guevara, Neves, et al., 2015), exclosure experiments show that small mammal herbivory cannot fully compensate for the magnitude of LMH seed predation and herbivory (Kurten, 2013; Villar et al., 2020; Villar, Rocha-Mendez, et al., 2021; Williams et al., 2021; Wright et al., 2007). Furthermore, large seeds are not heavily consumed by small mammals (Galetti, Guevara, Galbiati, et al., 2015). Thus, defaunation of LMH may release those tree species with large seeds from LMH predation and subsequently increase their recruitment and persistence in the community (Beck et al., 2013; Kurten et al., 2015).

Additionally, trees with large seeds might benefit from a higher competitive advantage at the sapling stage, which might contribute to dominating sapling communities in the absence of LMH (Wright et al., 2007). A similar mechanism might explain the increase in absolute and relative abundances of palms after the exclusion of LMHs. Like trees, many palms produce large seeds with relatively high germination rates and vigour (Beck et al., 2013; Galetti, Guevara, Galbiati, et al., 2015; Pizo et al., 2006). Palms have higher fecundity and are hyper-dominant in various Neotropical regions including the Atlantic and the Amazon forest (Staggemeier et al., 2017; ter Steege et al., 2013). As consequence, LMHs are strongly attracted to palm stands (Akkawi et al., 2020; Beck, 2017; Bodmer, 1990), spatially structuring plant communities and ecosystem processes across palm density gradients (Villar, Paz, et al., 2021; Villar, Rocha-Mendez, et al., 2021). Furthermore, in defaunated areas where LMHs are absent, most palm seeds accumulate beneath reproductive adults (Martínez-Ramos et al., 2016; Valverde et al., 2020; Wright et al., 2000).

Results also reject the hypothesis that seed dispersal by LMH might favour trees and palms, and suggest that any positive effects of seed dispersal by LMH on tree and palm recruitment might be outweighed by the negative effects of seed predation and trampling by those. In the presence of LMH, lianas and other growth forms might have a competitive net dispersal advantage over trees and palms since the predation of seeds of such growth forms might be proportionally lower, and recruitment much higher. While many trees and palms have seeds that are consumed, predated and dispersed by LMH, lianas for example have a large number of species with abundant small wind-dispersed or bird-dispersed seeds (Harrison et al., 2013; Michel et al., 2015; Terborgh et al., 2001; Wright et al., 2007). Indeed, the most abundant lianas in the open plots were two wind-dispersed (Schnella microstachya and Serjania communis) and one bird-dispersed species (Paullinia seminuda), suggesting that LMHs have a net negative effect in the recruitment of species whose seeds they consume.

The mechanical impact of LMH on sapling tissues through trampling, leaf and sapling consumption might also contribute to increasing growth form diversity. While adult trees are resistant to mechanical disturbance, trees at the sapling stage are susceptible to physical disturbance of medium- and large-sized mammals by trampling and browsing (Rosin et al., 2017). In a tropical forest of South-East Asia, Luskin et al. (2019) found that the experimental exclusion of the wild boar Sus scrofa increased tree populations by 86% while liana abundance, through recruitment and survival, increased by 86% in the presence of these. This study suggested that, even at early ontogenic stages, lianas can better tolerate the physical damage caused by wild boar. Interestingly, the wild boar and the most abundant species of LMH in our experiment (the white-lipped peccary Tayassu peccari) are both Suiformes, which are well known for their trampling, rooting and soil turnover behaviour. Thus, coherence in results from both studies and dominant faunal composition supports the hypothesis of an important role of mechanical disturbances in favouring lianas over trees. Morphological adaptations (e.g. thin and flexible stems) may allow lianas to partially compensate for the damage and mortality caused by mechanical disturbance (Andersen et al., 1997; Beck et al., 2013; Kilgore et al., 2010; Luskin et al., 2019; Wright, 2002). Furthermore, several studies in tropical forests demonstrate that lianas can invest more in-depth roots than in above-ground growth during the sapling stage, increasing their ability to survive against physical damages and competition (Martínez-Izquierdo et al., 2016; Restom & Nepstad, 2004; Schnitzer, 2005; Swaine & Grace, 2007). The ability to recover from disturbances, such as trampling or leaf consumption, might also explain why bamboos were not affected by the experimental exclusion of LMH (Buckingham et al., 2011; Calderón-Sanou et al., 2019; Terborgh et al., 2008). Bamboos possibly have low attractiveness for herbivores, and their solid biomechanical structures, as well as vegetative regeneration and propagation, might additionally contribute to regeneration from LMH disturbance (Rother et al., 2013).

A frequently overlooked mechanism that might also contribute to our results is the changes in soil physical properties derived from the impact of LMH trampling and the large amounts of soil overturned by LMH. Changes in palm abundance were mainly driven by the hyper-dominant palm Euterpe edulis, which represents 95% of the palms in our plots. Saplings of this palm have a superficial root system that is incapable of reaching the water in the deep wet soil layer during the dry season, so they are weak competitors for soil water availability (Matos et al., 1999). LMHs have been shown to increase soil evapotranspiration on grassland ecosystems, shifting plant communities towards more drought-tolerant species (Veldhuis et al., 2014). If this mechanism also operates in tropical forests (as suggested by Villar, Paz, et al., 2021), we expect that LMH might shift sapling communities towards more drought-tolerant growth forms, to the detriment of palms. Shrubs, whose relative abundance was favoured by LMH presence, have an efficient below-ground root system that allows them to uptake water and nutrients in systems with variable soil water availability (Brown et al., 1998). Herbs, also favoured by LMH presence, while being less competitive than shrubs in their below-ground ability to capture water under stress conditions, have a superior ability to tolerate variable microclimates and above-ground stress conditions, such as drought periods or light limitation (Brown et al., 1998; Whigham, 2004). Furthermore, while tree saplings might be more resilient to drought, palm saplings, which benefited the most from exclusion, are the most drought sensitive among the growth forms studied (van der Sande et al., 2013). Additionally, below-ground tissue of sapling growth forms as lianas, shrubs and herbs could be protected from above-ground herbivory instead of palms that in our experiment present a fragile superficial root system (Matos et al., 1999; Rifai et al., 2019). Thus, LMH impact on below-ground and soil physical conditions may contribute to increased growth form diversity.

In addition to these mechanisms, it is likely that the ongoing climate change and increasing drought conditions (Colombo & Joly, 2010; Rifai et al., 2019), might have also played an important role in the population increment of some growth forms. For instance, during the course of our experiment, lianas and herbs increased in both treatments, while neither shrubs nor bamboos experienced any temporal change in abundances at any of the treatments. Lianas are more abundant in seasonally dry tropical forests and their abundance increases relative to trees with the increasing length and intensity of the dry season (DeWalt et al., 2010; Schnitzer, 2005). Early ontogenic stages of lianas have some physiological adaptations that improve deeper water uptake, water flux and stomatal control, allowing them to perform well under drought conditions and leading to higher growth and survival rates than other growth forms on dry periods (Cai et al., 2009; Schnitzer, 2005; Swaine & Grace, 2007; Umaña et al., 2019). Thus, an increase in lianas across both treatments might be suggestive of a long-term increase in drought in our study region. This might also explain trends in the herbs, whose ability to tolerate drought periods might have contributed to their increase in both treatments. Conversely, as a result of LMH exclusion palms experienced a demographic boom during the first 3 years of study, but soon afterwards they joined the steady long-term decline evident on open plots, decreasing their absolute abundance in 10 years by 22% in exclusion and 35% in LMH presence. This decline is also consistent with the hypothesis of palms sensitivity to climate change as a long-term increase in drought conditions over the course of our experiment (Colombo & Joly, 2010; Sales et al., 2021) (Figure S3), which seemed to affect the balance between different growth forms across the study sites. Furthermore, shrubs, most likely the growth form least sensitive to below-ground drought, experienced no net changes in absolute abundance throughout our experiment. Thus, we suspect that these patterns indicate the increasing influence of climate change in the background long-term regional trends affecting our experiment.

4.2 The impact of LMH on growth form diversity and coexistence

We argue that defaunation might also alter coexistence dynamics between and among growth forms. The increase in palm relative abundance on closed plots occurred with a concomitant increase in palm species evenness driven by a decrease in the dominance by E. edulis and an increase in the abundance of the other palm species, including Geonoma elegans, G. gamiova, G. pauciflora, G. schottiana, Astrocaryum aculeatissimum and Attalea dubia. Tree species evenness also increased substantially on closed plots over the course of the experiment, with an increase in abundance of several species: 22% of the tree species in closed plots experienced an increase in abundance, while in open plots only 13% of the tree species experienced an increase. Exclusion also doubled the abundance of the medium and large fruiting tree species Eugenia expansa and Garcinia gardneriana during the experiment period. In open plots, G. gardneriana was more abundant in the first sampling but experienced a significant decrease in the following months, while in the closed plots its abundance increased up to 83% higher than in open plots by the end of the experiment. Thus, LMH exclusion led to an increase in evenness within both growth forms that were also the most favoured by this treatment.

Simultaneously, evenness among growth forms increased on open plots over the course of our experiment, but neither growth form richness or Shannon diversity did, although patterns in Shannon diversity were intermediate (Figure 3). Since growth form richness, the Shannon and the inverse Simpson indices reflect a gradient where the proportions of growth forms have increased weight in the value of the index, our results indicate that the effects of LMH are linked to changes in dominance. Growth forms that dominate the community dictate the functional processes and dynamics shaped by the community. For example, in forests where palms dominate, they play a keystone role as foundation species, shaping plant recruitment, richness, beta diversity, nutrient cycling and primary productivity through their interaction with LMHs (Villar et al., 2020; Villar, Paz, et al., 2021; Villar, Rocha-Mendez, et al., 2021). Thus, shifts in growth form dominance have strong consequences for ecosystem structure and function. Many studies demonstrate that the positive effects of LMH on plant species diversity are linked to shifts in dominance (Jia et al., 2018; Koerner et al., 2018; Mortensen et al., 2017; Villar & Medici, 2021), and our study indicates for the first time that this might also apply to growth form diversity in tropical forests. We suggest that the patterns in our study indicate suppression of dominance and a trade-off between diversity within growth forms versus among growth forms within the boundaries imposed by community size and competition for resources. In other words, LMH disturbances might favour a variety of alternative plant growth form strategies, subsequently preventing the many species within otherwise dominant growth forms to take over the resources available for the whole plant community.

Examination of pairwise abundance relationships also supports this view. Results clearly show that negative pairwise abundance correlations only occurred on closed plots, suggesting that the conditions for coexistence between different growth forms might be more constrained when defaunation occurs. Within such pairwise abundance relationships, trees were the least affected by the defaunation, while for other growth forms negative relationships of different magnitudes emerged, suggesting that defaunation triggers antagonistic dynamics among these ‘lesser’ growth forms. In this scenario, shrubs, for which all pairwise relationships were negative on closed plots but positive on open plots, clearly underperformed compared to other growth forms as a result of experimental defaunation. In contrast, palms appeared to benefit from such competitive dynamics. It is important to bear in mind that our approach does not allow us discerning direct pairwise spatiotemporal competitive dynamics or the conditions for coexistence, yet it facilitates detecting shifts in the relative competitive abilities of different growth forms between open and closed plots. It is remarkable, although, that all pairwise abundance relationships on open plots had a positive or neutral sign, suggesting that disturbances from LMH might, at least partially, override competitive dynamics between growth forms.

4.3 LMH impacts on growth form communities in the face of global environmental change

In general, our results are consistent with the fact that the functions carried out by LMH may be critical to preserving the structure and diversity of tropical forests. Many grassland ecosystems world-wide are experiencing woody plant encroachment with the increasing dominance of woody plants such as trees and shrubs (Briske, 2017). This is alarming since this process has many unwanted outcomes on, for example, soil composition, soil structure and climate regulation (Briske, 2017; Staal et al., 2020). In these ecosystems, LMH effects and interaction among plants are considered prominent factors that prevent woody plant encroachment (Staal et al., 2020) so that LMH defaunation benefits woody species (Levick & Rogers, 2008; Wigley et al., 2014). Accordingly, our results show that LMH prevents the proliferation of trees and palms in tropical forests and might modulate the competitive dynamics between growth forms.

Our experimental results, and in agreement with (Luskin et al., 2019), show that the temporal effects of LMH appear to benefit the early ontogeny stages of herbs, shrubs and especially lianas to the detriment of larger woody growth forms, such as trees and palms. Our analyses also show that LMH increases growth form diversity and modulate growth form dominance. Studies from defaunated forests suggest that removal of LMH and large arboreal vertebrates (large birds and primates) might benefit lianas (Kurten, 2013; Wright et al., 2007). This suggests that other functional groups of large herbivores affected by defaunation (such as arboreal primates and large birds), might play a complementary functional role in preventing lianas to proliferate, hence also promoting growth form diversity.

Long-term changes in environmental stability, like precipitation or temperature, have been shown to affect plant composition and communities on the global scale (Feeley et al., 2020; Sala et al., 2000). Large mammals might have also an important role in buffering tropical forests against such changes. For example, recent work from Villar and Medici (2021) suggests that LMH slow down the rapid temporal decline of plant diversity in diversity hotspots by buffering against community compositional change and shifts in dominance. Our results show that the long-term shifts in growth form composition, possibly strongly influenced by climate change in the region, do not increase growth form diversity on closed plots but only on open plots. Lianas and herbs, whose absolute abundances increased on both treatments along over the experiment, also experienced a concomitant increase in species evenness, suggesting that a higher long-term gain in recruitment will also benefit diversity within these growth forms. Furthermore, large mammals might curtail the dominance of trees and palms but, as our results show, this does not lead to a net decline in their diversity over the course of the experiment. Such results strongly suggest that LMH might be a useful management tool to improve diversity (within and among) plant growth forms, and hence the dynamics and ecosystem functions of tropical rainforests subjected to climate change. While these considerations need further inspection, we conclude that, by decreasing growth form diversity, defaunation of large ground-dwelling mammals will likely reduce diversity in the understorey which might lead to the vertical structural simplification of tropical rainforests and erode the alternative functions carried out by different plant growth forms.

ACKNOWLEDGEMENTS

We thank the Fundação de Amparo a Pesquisa do Estado de São Paulo (FAPESP) for funding this study throughout the 10 years of the experiment, and for continuing to have excellent research support service. FAPESP Grants: BIOTA/FAPESP grants 2014/01986-0 and 2013/50424-1; graduate fellowship to YS (2019/05538-5); Postdoctoral Fellowship to NV (2015/11521-7 and travelling grant 2018/20599-8). MG received a fellowship from CNPq (300970/2015-3). We thank the Instituto Florestal and Fundação Florestal do Estado de São Paulo for allowing and supporting our research in their nature reserves, and Rafael Souza for assistance in the field. We also thank all the LABIC members, who always help us with constructive discussions. This paper is dedicated to the 620,000 Brazilians dead from SARS-Covid-19 during President Jair Bolsonaro administration.

CONFLICT OF INTEREST

We ensure that none of the authors declared a conflict of interest.

AUTHOR CONTRIBUTIONS

Y.S., M.G., N.V. and V.Z. designed the research, conceived the manuscript and contributed towards the final version of the manuscript; V.Z. and S.N. collected most of the data; Y.S. and N.V. analysed the data, drafted the manuscript and contributed equally to this work.

Open Research

PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1111/1365-2745.13846.

DATA AVAILABILITY STATEMENT

Data available from the Zenodo Digital Repository https://doi.org/10.5281/zenodo.5842614 (Souza et al., 2022).