Body mass determines the role of mammal species in a frugivore-large fruit interaction network in a Neotropical savanna

Abstract

Frugivorous mammals play an important role in maintaining biodiversity and are considered one of the main dispersers of large seeds. In this study, we describe the structure of the interaction network between non-flying mammals and seven plant species with large fruits in a megadiverse savanna-forest mosaic in the Brazilian Cerrado. We also evaluated the individual contribution of each species to the organization of the interaction network and tested whether body mass determined the mammals’ role in the network. To record frugivory events of mammals with arboreal and terrestrial habits, camera traps were installed at ground and canopy levels. We identified 18 mammal species interacting with seven plant species in 515 frugivory events. Our observations highlight an interaction network with a modular and non-nested topology and the important role of large mammals in the network structure, which reflects the importance of the group in potential seed dispersal. The extinction of large frugivorous mammals can cause several damages to ecosystem services in the Brazilian Cerrado through changes in network structure, especially threatening the survival of plant species with large fruits.

Introduction

Animal-plant interactions are important for maintaining biodiversity and ecosystem functions (Jordano et al. 2011). Among vertebrates, mammals receive special attention given their important role in pollination and seed dispersal, which provides gene flow and helps in the proliferation of many plant species (Jordano et al. 2007, Golin et al. 2011). Interactions between animals and plants are influenced by biological attributes, such as fruit biomass, that shape species interactions (Fuzessy et al. 2018). At the same time, several biological characteristics of the species may constitute morphological or spatiotemporal barriers that limit some types of pairwise interactions, which are described as the ‘forbidden links’ hypothesis (Jordano et al. 2003). Therefore, large fruits tend to be dispersed by large mammals, as they could not be dispersed by other animal groups due to the morphological restrictions inherent to small dispersers (Jordano et al. 2007, Lim et al. 2020).

Interaction networks have been used as tools to understand the complexity of ecological interactions and the influence of morphological traits and have been considered a powerful methodology that maps interactions, characterizes the functional roles of species within communities, the diversity of relationships established between frugivores and plants, and the importance of the species involved (Bascompte & Jordano 2007). Network analyses include interactions between species as an additional layer in ecological evaluations, resulting in great advances in understanding the establishment of biological communities in different ecosystems (Delmas et al. 2019). Recently, this approach has provided important information on the ecological role of vertebrates in seed dispersal (Vidal et al. 2013), in addition to revealing the drastic reduction in the abundance of the dispersers and changes in the functional roles of species (Galetti et al. 2013, Carreira et al. 2020).

The use of the interaction network approach can help in conservation and restoration strategies, ensuring the best functioning of ecosystems (Harvey et al. 2017, Raimundo et al. 2018) and avoiding species extinctions (Carreira et al. 2020). In this sense, the investigation of legally protected areas that can serve to define conservation strategies emerges as part of environmental planning for the maintenance of biodiversity and ecological processes in threatened ecosystems. These protected areas have remnants of natural vegetation that are important for the maintenance of biodiversity due to the existing complex structure (Sukma et al. 2019, Magioli et al. 2021a, Magioli et al. 2021b) and serve as stepping stones for connection with smaller fragments (Wintle et al. 2019). The low degree of human intervention in these areas allows for the occurrence of several groups of animals and plants, including a high diversity of mammals, which play an important role in seed dispersal, especially of plants that produce large fruits, that, due to morphological restrictions, have their seeds dispersed only by a specific set of mammals (Bello et al. 2015, Magioli et al. 2021a).

Herein, we aim to describe the structure of the interaction network between frugivorous non-flying mammals and seven plant species that produce large fruits, in a protected savanna-forest mosaic in the Brazilian Cerrado, and to evaluate the individual contribution of the species involved. Our hypothesis is that the interaction network will present a modularity and non-nested structure, due to the existing biodiversity in the area (Santos-Filho & Silva 2002, Santos-Filho et al. 2012) and the presence of morphological barriers that restrict consumption of the large fruits evaluated, which can form a set of species closely linked within modules (Almeida-Neto et al. 2008, Donatti et al. 2011). We also expect a positive relationship between mammalian biomass and its relevance in the context of mutualistic interactions, since larger and heavier species tend to have less morphological restrictions concerning the ingestion of large fruits, besides needing a greater energy demand, which drives the consumption of fruits (Donatti et al. 2011, Galetti & Dirzo 2013).

Materials and methods

Study area

The study was carried out at the Serra das Araras Ecological Station (hereafter SAEE), a Federal Integral Protection area that occupies an area of 28.700 hectares, between the municipalities of Porto Estrela and Cáceres in the State of Mato Grosso, Brazil (15°38’32.0” S 57°11’27.3” W) (Brasil 2016). The predominant climate is of the semi-humid hot tropical type, classified as megathermal Aw with two seasons: dry, which extends from May to October, and rainy from November to April, with annual precipitation around 1.500 mm and maximum average temperature of 30º C and minimum 20º C (Alvares et al. 2013). From a biogeographic perspective, the SAEE is located in an ecotone area with high biodiversity, inserted in the Cerrado and in contact with two other Brazilian biomes, the Amazon Forest and Pantanal wetland (Vitorino et al. 2018). Inside the SAEE, the samples were taken in a vegetation mosaic of semi-deciduous seasonal forest, savanna woodland (cerrado sensu stricto) and gallery forest.

Data collection

In the SAEE, we collected data between September 2019 and September 2020, totalling 13 months of consecutive field expeditions. We used the photographic trapping methodology to record frugivorous mammal-fruit interactions because this method is efficient for sampling interactions between mammals and plants, and it is a minimally invasive methodology (Bogoni et al. 2018). To record information about the interactions, camera traps were installed about 50 cm above the ground and attached to the trunks of the trees (Raíces et al. 2017, Carreira et al. 2020). To record the frugivorous species with arboreal habits and their ecological interactions established there (Zhu et al. 2021, Moore et al. 2021), a wooden structure was created, in the centre of which the camera was fixed and hoisted so that it could be placed on the branches with the greatest abundance of fruits (Figure S1). The cameras remained active for 24 hours each day and were configured to record 10-second videos after motion detection, with intervals of five seconds between videos.

The criteria for selecting the trees to be sampled were to be in the fruiting period of the focal plant species (Table S2), to have fleshy and/or attractive fruits for frugivorous species with a size greater than 40 mm (Table S2), as described by Kuhlmann (2018), and presenting a minimum distance of 200 metres from the other individuals sampled. Thus, the cameras were installed focusing on seven species, viz., Hymenaea courbaril L. (Fabaceae), Genipa americana L. (Rubiaceae), Pouteria ramiflora (Mart.) Radlk. (Sapotaceae), Cordiera macrophylla (K.Schum.) Kuntze (Rubiaceae), Dipteryx alata Vogel (Fabaceae), Diospyros hispida ADC. (Ebenaceae), and Attalea speciosa Mart. ex Spreng. (Arecaceae).

We used 79 camera traps operating an average of 30 days on each individual of the seven plant species, in which we carried out a sampling effort of 28,344 hours of monitoring with the camera traps on the canopy and 68,288 hours on the ground, totalling 97,632 hours of sampling. The sampling effort conducted for each species was based on the number of individuals present in the area, as well as on species phenology (i.e., availability of fruits over time), so that the most representative species were sampled for a longer time (Table 1).

Table 1. Number of camera traps installed and sampling effort in hours per fruit plant species in the Serra das Araras Ecological Station.

We defined an interaction event (i.e., frequency) every time a mammal ingested or carried a fruit with its seed. We consider as independent records all interactions that were separated from each other by an interval equal to or greater than 30 seconds (sensu Carreira et al. 2020). The taxonomic classification for plants and mammals followed Brazilian Flora (Brazil Flora 2020) and List of Mammals of Brazil (Abreu-Jr et al. 2020), respectively. Non-flying mammals were separated into three categories: small mammals, with a body mass of up to one kilogram, medium-sized mammals, with body mass between one and seven kilograms (Chiarello 2000), and large mammals with a body mass greater than seven kilograms (Emmons & Feer 1997). Information on the body mass of mammal species was obtained from Wilman et al. (2014).

Data analysis

For the analysis of interaction networks, we created a matrix weighted by the frequency of interactions collected in the field. All subsequent analyses were performed in the R software (R Development Core Team 2019). The completeness of our sampling was obtained by dividing the total number of observed links (which quantifies the pairing between species) by the estimated number via Chao 1 (Chao 1984), using the iNEXT R-package (Hsieh et al. 2020). We used the bipartite R-package (Dormann et al. 2020) to assess the network metrics: species richness of both mammals and plants (network size); number of interactions; number of links; nestedness, using NODF (Nestedness metric based on Overlap and Decreasing Fill, Almeida-Neto et al. 2008) and wNODF (Weighted Nestedness metric based on Overlap and Decreasing Fill (Almeida-Neto & Ulrich 2011), that describes the pattern of interaction in which specialist species interact with a subset of generalist species; and modularity (Q _w ), which identifies the presence of subsets of species that tend to interact more often with each other than with species from other subsets (Olesen et al. 2007). Specifically, for NODF, wNODF, and Q _w , we evaluated the level of significance by comparing the results obtained with those of 1,000 random networks generated according to null models using the vaznull function, which maintains the same patterns of connectance and total marginals in relation to the observed matrix (Vázquez et al. 2007, Dormann et al. 2020).

To determine the role of species in the network, we calculated the metrics Species Strength (SS), which quantifies the importance of each species based on the sum of the dependencies of their respective partners, and Closeness Centrality (CC), which measures the proximity of a species to all others, indicating the capacity of a species to act as a hub and increase the cohesion of the network (Martín González et al. 2010, Delmas et al. 2019). Also, we verified the role of species in the modular structure, calculating the standardized pattern and connectivity of species between modules (c-score) and within their respective module (z-score). In this approach, species can be classified as peripheral, when they present low values of c- and z-score; connector, with high c-score and low z-score; module hub, with high z-score and low c-score and; network hub, with high c- and z-score values (Olesen et al. 2007). We determined the cut-off values of c- and z-score from 100 null matrices, following Dormann & Strauss (2014), which in our case was c- _critical = 0.71 and z- _critical = 1.86. Next, we performed a non-parametric Wilcoxon test to assess whether there was variation in the metrics at the species level (Species Strength, Closeness Centrality, c- and z-score) concerning different trophic levels (mammals and plants).

Additionally, we used a principal component analysis (PCA) to synthesize the role of mammals according to the metrics Species Strength, Closeness Centrality, and c- and z-score. The first principal component (PC1) explained 67% of the variation of the metrics and was used as the descriptor of the contributions exerted by the species. Thus, the higher the value of PC1, the greater the relevance of the species in the structuring of the network. Finally, we tested the mammals’ body mass (logarithmized for better fit) as a possible predictor to determine the role of these species in the network, using a linear model (LM).

Results

We identified 18 mammal species, grouped according to their body size into small (n = 5), medium (n = 9) and large (n = 4), interacting with seven plant species. Moreover, we recorded 515 frugivory events distributed into 48 links. The number of estimated links was 76, resulting in a sampling completeness of 63%. The interaction network evaluated was not significantly nested (NODF = 54.95, p > 0.05; wNODF 23.45, p > 0.05) but significantly modular (Qw = 0.43, p < 0.05). The mammal species that consumed fruits the most were Tapirus terrestris, Dicotyles tajacu, and Cuniculus paca, with 130, 69, and 66 frugivory events, respectively. Among the plant species with the largest number of interactions, Diospyros hispida (n = 164), Pouteria ramiflora (n = 129), and Dipteryx alata (n = 60) stand out (Figure 1). When evaluating the Species Strength metric, among the mammals Tapirus terrestris (2.15), Cuniculus paca (1.17), and Cerdocyon thous (0.76) had the highest ones, while among the plants the highest values were of Dipteryx alata (5.10), Cordiera macrophylla (3.45), and Diospyros hispida (3.32). Regarding the Closeness Centrality measure, the species that had the highest values were Tapirus terrestris (0.06), Mazama sp. (0.06), and Cerdocyon thous (0.06) among the mammals, and Dipteryx alata (0.14), Cordiera macrophylla (0.14), and Diospyros hispida (0.14) among the plants (Tables 2 and 3). By checking the role of species in the modular structure of the network, the mammals Tapirus terrestris (2.18), Cerdocyon thous (1.78) and Dicotyles tajacu (1.78) and the plants Cordiera macrophylla (0.90), Pouteria ramiflora (0.70), and Genipa americana (0.16) showed higher z-score values, while the mammals Tayassu pecari (0.67), Dicotyles tajacu (0.66), and Proechimys longicaudatus (0.64) and the plants Pouteria ramiflora (0.55), Hymenaea courbaril (0.38), and Diospyros hispida (0.36) showed higher c-scores. Tapirus terrestris was the only species in the network classified as a module hub, while the others were peripheral (Figure 2).

Figure 1. Interaction network between frugivorous mammals and plants that produce large fruits, in a Neotropical savanna in the Serra das Araras Ecological Station, Brazil. Modules are highlighted in the network by different colours, and the grey lines represent the interactions established between species of distinct modules.

Table 2. Network metrics evaluated at the species level (Species Strength, Closeness Centrality, c and z-score) of an interaction network between frugivorous mammals and plants that produce large fruits in a Neotropical savanna in the Serra das Araras Ecological Station, Brazil. Mammal body mass was obtained from Wilman et al. (2014). The species were separated into small mammals (weighing up to 1 gg), medium-sized mammals (1 to 7 kg) (Chiarello 2000), and large mammals (more than 7 kg) (Emmons & Feer 1997).

Table 3. Network metrics evaluated at the species level (Species Strength and Closeness Centrality) for the group of plants of an interaction between frugivorous mammals and plants that produce large fruits, in a Neotropical savanna in the Serra das Araras Ecological Station, Brazil.

Figure 2. Species’ role in the modular structure of an interaction network between frugivorous mammals and plants that produce large fruits in a Neotropical savanna at the Serra das Araras Ecological Station, Brazil.

Using the Wilcoxon non-parametric test, we verified that there was significant variation between the two trophic levels for Closeness Centrality (W = 0; p < 0.001) and Species Strength (W = 8; p < 0.001), with plants assuming more central positions and having greater strength of interactions (Figure 3), but we did not observe significant variation for c and z-scores. We did not observe variation between the different trophic levels for c and z-scores (p > 0.05). In addition, we also identified that the body mass of mammalian species acts as an important predictor of the role that these species play in the network, with mammals with higher biomass being the most relevant (R² _adj = 0.40; p < 0.01) (Figure 4).

Figure 3. Significant variation in the species roles of an interaction network between frugivorous mammals and plants that produce large fruits, in a Neotropical savanna in the Serra das Araras Ecological Station, Brazil.

Figure 4. Body mass as a predictor of the role of mammals in the network structure. PC1 Index reflects the species-level metrics Species strength, Closeness centrality, and c- and z-scores.

Discussion

Our findings evidenced a modular but non-nested interaction network, with a high number of frugivorous mammals acting as potential dispersers for several plant species that produce large seeds. Furthermore, we found that disperser biomass was a good predictor of the role that these mammals play in the network. In the evaluated interaction network, the largest (heaviest) mammals were also the most important in the modular structure, assuming a central position and with a high value in the species strength.

We observed significant modularity values of the interaction network of mammals with large fruits in the Cerrado, as previously observed in the Atlantic Forest and Pantanal (Donatti et al. 2011, Carreira et al. 2020), and a non-nested pattern, which was in accordance with our expectations. However, environmental variables such as phenology and plant abundance observed may have influenced the structuring of the interaction network, such as grouping into modules (Vázquez et al. 2007, Encinas-viso et al. 2012, Machado-de-Souza et al. 2019). Regarding modularity, still in line with what was observed for other megadiverse areas, we identified a system in which a set of species tends to interact more with each other than with species from other sets (Olesen et al. 2007), which results in a more robust and resilient system in the presence of possible indirect and direct impacts (Carreira et al. 2020). In non-modular networks, as observed, for example, by Queiroz et al. (2021) and Naniwadekar et al. (2019), environmental impacts are felt more intensely and can result in a cascade effect if a species disappears from the system, compromising the network of interactions (Olesen et al. 2007). Neotropical interaction networks tend to be less nested (Dugger et al. 2019), as we have observed, which may indicate that large-fruited plants attract different subsets of frugivorous species so that interactions are not necessarily occurring with more generalist mammals (Almeida-Neto et al. 2008, Crestani et al. 2019, Naniwadekar et al. 2019).

Using species-level metrics, we highlight three important fruit trees for maintaining the structure of the interaction network: D. alata, C. macrophylla, and P. ramiflora. In addition, we showed that plants have greater strength of interaction and act as connectors in the system. In this sense, this group increases the connectivity and cohesion of the network, suggesting that it is the species that act to maintain the existing biodiversity and the dynamics of the ecosystem, thus avoiding extinctions (Cagua et al. 2019, Ramos-Robles et al. 2018). These large-seeded plant species are categorized as attractive to fauna due to their high nutritional value (Kuhlmann 2018) and make potential contributions to carbon storage (Bello et al. 2015). Also, they are important for the maintenance of the mammal community, including rare species for the region where the study was carried out, such as the Kinkajou (Potos flavus), which was recorded interacting with D. alata, this being the first record with documented evidence for the species in the SAEE (see Figure 1).

Our results demonstrate the importance of medium and large mammals in the evaluated interaction network with large fruits. Among the species, we can highlight three T. terrestris, C. thous and T. pecari, two of which are categorized as Vulnerable to Extinction at the national and international level (IUCN 2022) and considered important in other studies of ecological interactions (Donatti et al. 2011, Vidal et al. 2013, Bogoni et al. 2018). Specifically, the tapir (Tapirus terrestris), identified as module hubs, as explained by Donatti et al. (2011), interacts with a high diversity of plant species and stands out for the quality of seed dispersal over long distances (O’Farrill et al. 2013, Jordano et al. 2007, Fuzessy et al. 2018). These results demonstrate the need for identifying the ecological functions performed by the species, such as frugivory and seed dispersal, given that extinctions can cause cascading effects in the system and compromise these ecosystem services (O’Farrill et al. 2013, Vidal et al. 2013, Godínez-Alvarez et al. 2020). Regarding C. thous, this species was also mentioned as relevant in ecological interaction studies, especially in degraded ecosystems, due to its tolerance to environmental changes and ability to disperse seeds over long distances (Bogoni et al. 2018).

Besides verifying the important species in the evaluated interaction network, we also showed that mammal biomass is an ecological determinant of the role that these species play in the network, with a positive and significant relationship with the metrics used in this study. These results support our hypothesis that large mammals provide a major contribution to network structure and fruit removal. Other studies observed similar results, as highlighted by Donatti et al. (2011) in a system evaluated in the Pantanal, in which large frugivores interacted with many plants with fruits of different sizes. Therefore, large mammals are essential elements in the structure of frugivory networks (Palacio et al. 2016) and play a fundamental role in the processes of seed dispersal and recruitment (Donatti et al. 2011, Fuzessy et al. 2018), including those of plants that produce large seeds, as shown in our study.

The disappearance of large mammals is one of the current problems of the Anthropocene and is the result of fragmentation, habitat loss, and hunting (Dirzo et al. 2014, Ripple et al. 2015). Defaunated ecosystems, where the large fauna is extinct, present showchanges in ecological processes (Young et al. 2016, Lim et al. 2020) and in functional roles (Carreira et al. 2020). This favours mesocarnivores and generalists at small scales (Ripple et al. 2015), increases seed predation (Galetti et al. 2015, Lacher et al. 2019), and changes the carbon and nitrogen cycles (Bello et al. 2015, Villar et al. 2020).

Body mass and species richness are related to ecosystem services and ecological function. Thus, the presence of large mammals in large remnants indicates the importance of preserving these areas (Magioli et al. 2021a, Magioli et al. 2021b). Therefore, our results emphasize the importance of conserving areas of Cerrado, to preserve species and promote the stability of ecological interactions (Ferreira et al. 2020). Researches carried out in this context provide data that enable the implementation of efficient measures to keep communities viable and prevent extinctions (Carreira et al. 2020), ensuring better ecological functioning (Harvey et al. 2017, Raimundo et al. 2018).

Conclusions

In summary, we highlight the importance of a protected environment in the savannas of the Neotropical region for the maintenance of interactions between species of the fauna and flora. In general, the seven observed plant species strongly contribute to the structure of the interaction network with non-flying mammals. We emphasize the importance of large mammals in this process, especially Tapirus terrestris, which is a threatened species. The disappearance of large mammals can harm the structure of interaction networks, mainly compromising the maintenance of plant species that have large fruits and, consequently, other ecosystem services. Thus, our results demonstrate the importance of the large fauna, indicating that the absence or loss of large frugivores will have negative consequences on ecological dynamics.