The Effect of Habitat Amount on Species Richness and Composition of Medium- and Large-Sized Mammals in the Cerrado Biome, Brazil

Abstract

Habitat fragmentation and reduction in the Cerrado are the primary threats to this biome’s biodiversity. The amount of habitat in the landscape has been proposed as the sole predictor variable for species richness in fragmented areas, potentially replacing the combined effects of fragment size and isolation (habitat amount hypothesis). This study aimed to test the influence of sampled fragment area, number of fragments, and habitat amount in local landscapes on the species richness of medium- and large-sized mammals in the Cerrado, southeastern Goiás, Brazil. The applicability of the habitat amount hypothesis to medium- and large-sized mammals in fragmented Cerrado habitats was thus evaluated. Medium- and large-sized mammal species were recorded from 2014 to 2018 in 14 Cerrado fragments in southeastern Goiás, Brazil. Using Landsat 7 and 8 satellite imagery from the year 2000 and the mammal sampling period, landscapes were delineated by creating buffers with a radius of 2 km from the central point of each sampled fragment. Through visual classification of these landscapes, the following variables were obtained: habitat amount in the landscape (HA), number of fragments (NP), and area of the sampled fragment (HF). The results indicate that the habitat amount in the past landscape (14 to 18 years before sampling) was the best predictor variable for the species richness and composition of medium- and large-sized mammals. The HA variable from the sampling period and the AREA variable from both periods also significantly influenced species composition. Therefore, considering the historical landscape context, the habitat amount hypothesis was applied to mammals in the Cerrado areas studied.

1. Introduction

Fragmentation and habitat loss are the main threats to biodiversity [1,2]. Fragmentation transforms larger areas into multiple, more minor remnants with different isolation levels, resulting in changes in habitat arrangement and connectivity. In contrast, habitat loss does not necessarily break the continuity of fragments but reduces habitat area and available resources for biota [3]. These landscape changes lead to alterations in local biodiversity, either at the population level—such as changes in abundance, reproduction, distribution of individuals, recruitment, and survival rate [3,4]—or at the community level, with alterations in species richness and composition [5,6].

The Cerrado is the second largest biome in Brazil in terms of territorial extension, covering approximately 2 million km², and stands out for both its high biodiversity and high rates of fragmentation and habitat loss, being considered a global biodiversity hotspot [7]. However, this biome has experienced a high alteration of its natural landscape for agricultural use, with the Cerrado already losing about 46% of its native vegetation cover and having only 19.8% of its area unaltered [7], with only about 9% of it under legal protection in the form of Conservation Units [8].

The Cerrado is home to more than 4800 species of plants and vertebrate animals [7]. The mammals inhabiting this biome number at least 268 species [9], which play a key role in ecosystem maintenance and balance, being considered keystone species in biodiversity conservation. Mammals influence the structure and functioning of ecosystems by affecting plant species composition and seed dispersal and regulating prey populations [10,11]. However, this mammal diversity is primarily threatened by habitat loss, fragmentation processes, and hunting pressures [12].

The theory of island biogeography [13] was initially used to explain species richness in fragmented sites, considering fragment size and isolation variables for local species richness. Later, the metapopulation dynamics theory [14,15] was proposed, offering more suitable models to increase our understanding of how habitat loss and fragmentation impact local species richness. The metapopulation theory, in conjunction with landscape ecology [16], has been a key approach in studies aiming to understand the effects of habitat loss and fragmentation on species richness, seeking to understand how landscape structure and spatial configuration affect local biota [6].

Several studies have shown a positive effect of fragment size on mammal species richness [5,17,18], indicating that the size of the fragment studied is an important variable for the mammalian fauna. However, Fahrig [19] proposed the habitat amount hypothesis to explain species richness in fragmented landscapes. This hypothesis predicts that species richness at sample sites of the same size should increase with increasing the total amount of habitat in the local landscape surrounding each fragment. It also predicts that the size and isolation of the patches containing the sample sites should have no additional effects on species richness at the sample sites if the amount of habitat in the surrounding landscape is the same [19]. Thus, the impact of the amount of habitat on species richness may be as strong or stronger than other predictors, such as fragment size and isolation or a combination of both [20].

Another factor that has been little evaluated in relation to habitat amount is the time needed for local extinctions to occur after habitat changes. Habitat fragmentation and reduced refuge areas may contribute to extinction debt [21]. This process occurs because populations can survive temporarily after a significant loss of habitat but then face a trajectory of population decline that leads them to local extinction due to the fragmentation and isolation resulting from these past changes [21]. Understanding the effects of fragmentation and past habitat amounts is crucial to analyzing how populations behave in different scenarios of fragmentation and isolation.

The present study was conducted to test the effect of fragment size, the number of fragments, and the habitat amount in local landscapes (habitat amount) on the species richness and composition of medium- and large-sized mammals in the southeastern region of Goiás, Brazil. Therefore, this study aims to test the habitat amount hypothesis (HAH) [19] in anthropogenically fragmented landscapes in the Cerrado, using medium- and large-sized mammals as a biological model. Furthermore, we will test whether species richness is more influenced by the past habitat amount in the landscape or by the habitat amount measured at the time of mammal sampling.

2. Material and Methods

2.1. Study Area and Mammal Data

The studied areas are located in the southeastern region of Goiás (Brazil), within the Brazilian Cerrado. The region’s climate is classified as AW (seasonal tropical), with an annual precipitation of approximately 1600 mm. It is characterized by two distinct seasons: a dry winter and a rainy summer, with average temperatures around 23 °C [22]. The study area is represented by a mosaic of forest and savanna physiognomies, consisting of semi-deciduous seasonal forests, considered enclaves of the Atlantic Forest, often associated with riparian and gallery forests, with Cerrado (Brazilian savanna) at the forest edges and patches of shrubby savanna [23,24]. The areas feature an anthropized rural matrix commonly found in the southeastern region of Goiás, dominated by pastures and agriculture [23].

Data on the occurrence of medium- and large-sized mammals in the studied fragments were collected between May 2014 and October 2018, according to the following procedure. Preliminary inventories of medium- and large-sized mammals were carried out in 14 remaining Cerrado fragments (Figure 1, Tables S1 and S3), which are located in the municipalities of Ipameri (n = 8), Catalão (n = 4), Urutaí (n = 1) and Campo Alegre (n = 1). The sampling effort was similar between the fragments studied. The occurrence of mammal species was sampled during four visits to each area, using direct (visual and vocal) and indirect (tracks, burrows, and other signs) methods to record species. To this end, the habitat area of the fragments was randomly traversed, and tracking was conducted on roads, trails, and their surroundings, as well as searches for traces on the banks of watercourses. Once an animal was sighted or a trace was found, a record was made in a field notebook, along with information about the location and type of environment sampled. In addition, to complement the species inventory, two camera traps (Bushnell brand, digital with 8-megapixel resolution) were installed in the fragments, in places of fauna passage, during the data collection period to obtain images of the mammals. Camera trap efforts totaled 588 traps-night (mean of the 42 traps-night per fragment). Richness was assessed by counting the number of medium- and large-sized mammal species recorded in each fragment.

Only medium- and large-sized mammal species [25] that use forest environments and denser Cerrado as an important source of resources, which are classified here as habitats for these species, were included in this study. Therefore, although they were recorded during the field surveys, species that have a preference for open habitats [e.g., the crab-eating fox Lycalopex vetulus (Lund, 1842), the maned wolf Chrysocyon brachyurus (Illiger, 1815) and the tapeti Sylvilagus sp.] and semi-aquatic species [e.g. the water opossum Chironectes minimus (Zimmermann, 1780), the otter Lontra longicaudis (Olfers, 1818), and the capybara Hydrochoerus hydrochaeris (Linnaeus, 1766)] were not included.

The list of mammal species inventoried in each studied fragment and included in the data analyses is found in Table S2.

2.2. Landscape Metrics

The satellite images from two distinct periods were analyzed to obtain data on the studied landscapes: (1) images from the year 2000; and (2) images from the years in which mammal surveys were conducted in the field, hereafter referred to as the sampling period (years 2014, 2016, and 2018). The choice of images from the year 2000 was aimed at evaluating the influence of past landscape alterations on the richness and composition of medium- and large-sized mammals, given that mammal species richness may exhibit a time lag in response to habitat alterations [6].

Landsat 7 satellite images from the year 2000 and Landsat 8 images for the sampling period (years 2014, 2016, and 2018) were obtained from the USGS (United States Geological Survey) repository [26]. Satellite images from July, corresponding to the dry season with low cloud cover, were selected.

Initially, to differentiate the various landscape features (e.g., agricultural areas, forest remnants, water bodies), an RGB (red, green, blue) composition—“false color”—was created using three distinct bands from the Landsat 7 and 8 images, generating a composite color image. Subsequently, buffers with a radius of 2000 meters were created from the center of each sampled fragment [27] to perform a visual classification of land use within the landscape. Two landscape classes were adopted: (1) habitat (forests and denser Cerrado areas) and (2) non-habitat (areas with altered original vegetation, lakes, rivers, grasslands, and urban areas). The land-use classification shapefile was saved in raster format for further analysis of landscape and fragment metrics. Each buffer covers an area of the 1257 hectares, which was considered the landscape unit under study. This scale was chosen because previous studies indicated that it is suitable for studies on medium- and large-sized mammals [6,28].

The R Studio program [29] analyzed the raster images using the landscapemetrics package [30] to generate fragment and landscape metrics. For this study, three metrics were selected for statistical analyses: class area (CA)—representing the sum of the areas of all habitat patches in the landscape, expressed in hectares (ha); number of patches (NP)—counting the number of habitat patches in the landscape; and patch size (AREA)—the area of the sampled fragment, in hectares (ha) (Table S3).

2.3. Statistical Analyses

The data on landscape metrics (HA and NP), fragment size (AREA), and mammal species richness were subjected to principal component analysis (PCA), using correlation, to explore the relationship between the obtained metrics and species richness. Based on this result, the NP metric was excluded from further analyses, as no relationship with species richness was observed.

The data on the amount of habitat in the landscape (HA) and the size of the sampled fragments (AREA) in 2000 and during the sampling period (2014, 2016, and 2018) were compared using the non-parametric Wilcoxon test. This analysis was performed to test for significant changes in the HA and AREA metrics between the studied periods.

Principal coordinate analysis (PCoA) was performed using the Jaccard dissimilarity index between all pairs of fragments to represent the species composition of the studied fragments. The first axis of the PCoA was used to represent the species composition of medium- and large-sized mammals in the sampled locations [31].

The influence of the HA and AREA metrics from 2000 and the sampling period on the richness and composition of medium- and large-sized mammal species was evaluated using a structural equation model, as performed by Melo et al. [31]. Additionally, simple linear regression was used to test the influence of the HA metric from 2000 on the richness and composition of medium- and large-sized mammal species, as this metric proved to be the most important in the previous analysis. Therefore, the habitat amount data from 2000 were also used to classify the percentage of habitat cover in the studied landscapes into three groups: less than 15%, 15–30%, and greater than 30% cover.

All statistical analyses were performed in R Studio [29] using the vegan [32] and lavaan [33] packages. The ggplot2 [34], factoextra [35], and semPlot [36] packages were used to visualize the results.

3. Results

The species richness of medium- and large-sized mammals observed at the sampled sites totaled 23 species that use forest environments and denser Cerrado areas as their primary habitat. Among the fragments, species richness ranged from 10 to 19 species (Table S2).

The studied landscapes showed no significant changes in habitat amount (W = 97; p = 0.982) or sampled fragment size (W = 111; p = 0.571) between the year 2000 and the field sampling period (2014, 2016 and 2018). On average, the HA metric increased slightly from 308 ha in 2000 to 317 ha during the sampling period, while the mean size of the sampled fragments changed marginally from 127 ha in 2000 to 126 ha during the sampling period. However, despite the absence of significant differences in the studied metrics, 8 areas showed an increase in HA values, while 6 experienced a reduction. Similarly, 6 fragments exhibited a rise in AREA, while 8 showed a decrease in this metric (Figure 2).

The principal component analysis (PCA) (Figure 3) shows that the first two axes accounted for 84.9% of the total data variance, with axes 1 and 2 explaining 59.4% and 25.5%, respectively. The vectors representing the amount of habitat in the landscapes (HA sampling period and HA year 2000), the habitat area of the sampled fragments (AREA sampling period and AREA year 2000), and mammal species richness were more strongly associated with axis 1. Conversely, the vectors representing the number of fragments in the landscape (NP sampling period and NP year 2000) were more closely associated with axis 2. Therefore, it can be observed that sites with more significant amounts of habitat in both the landscapes and the sampled fragments also exhibited higher mammal species richness (Figure 3).

The mammal species composition was highly variable across the studied sites, and the PCoA summarized 50% of this variation in its first two axes, with axis 1 accounting for 32% and axis 2 for 18% of the community variation (Figure 4). The samples were scattered along the PCoA, revealing a gradient in sample ordination based on habitat amount in the landscape (HA), with some species (e.g., Dasyprocta azarae, Puma concolor, Mazama rufa, and Dycotyles tajacu) being associated with landscapes containing greater habitat amounts (Figure 4 and Figure 5).

Landscape metrics influenced the diversity of medium- and large-sized mammals in the studied region. Structural equation models (SEM) explained 74.5% and 90.6% of species richness and composition variation, respectively (Figure 6). The variable habitat amount in the landscape for the year 2000 (HA—year 2000) was the only metric that significantly influenced mammal species richness (β = 0.438; p = 0.018) (Figure 6A). However, for species composition, all tested metrics significantly influenced community structure (Figure 6B), including HA—year 2000 (β = −1.128; p < 0.001), HA—sampling period (β = 0.713; p < 0.001), AREA—year 2000 (β = 0.499; p = 0.002), and AREA—sampling period (β = −0.526; p = 0.004). It is also evident that their past conditions positively influence the current (sampling period) fragment and landscape metrics (Figure 6).

The metric habitat amount (HA) in the year 2000 was the variable that most influenced mammal species richness and composition (Figure 6) and was thus further analyzed using simple linear regression (Figure 7). This analysis revealed a gradual increase in species richness with the increase in the habitat amount in the landscape (Figure 7A). It also showed a gradual shift in species composition in response to habitat amount in the landscape (Figure 7B), where landscapes with similar habitat amounts exhibited close values for species richness and composition (PCoA axis 1).

4. Discussion

Our results demonstrate that metrics indicative of the available habitat are important variables for the richness and composition of medium- and large-sized mammal species in Cerrado fragments. The predictor variables evaluated in the structural equation modeling showed a more significant influence on species composition (90.6%) than on species richness (74.5%) of mammals (Figure 6). Melo et al. [31] observed a similar result in a study with small mammals in Cerrado areas.

Habitat availability was shown to be an important variable for the mammalian fauna. The trend of increasing richness of medium- and large-sized mammal species with growing size of sampled fragments has been evidenced in other studies [5,17]. However, Fahrig [19], when presenting the habitat amount hypothesis, states that the species richness found in a fragment is linked to the amount of habitat in the local landscape; that is, the number of species should increase according to the total amount of habitat in the local landscape. On the other hand, as habitat loss advances, species richness in the landscape decreases along with the area of the remaining habitat, regardless of the individual size of the remaining fragments [19].

In this study, the past amount (year 2000) of habitat in the landscape (habitat amount—HA) was the best predictor for both the richness and composition of medium- and large-sized mammal species. In addition, the variables HA—sampling period and AREA of the year 2000 and sampling period also had a significant influence on the composition of mammal species, and for the current period, the variable HA had a more significant impact on species composition than the variable AREA (Figure 6). Species richness was higher in landscapes with a greater past amount of habitat, and landscapes with similar amounts of habitat had more similar species richness and composition (Figure 5 and Figure 7). Other studies have also shown a positive influence of habitat amount on the richness of medium- and large-sized mammal species in the Cerrado [6], in the Amazon [37], and in the Atlantic Forest [38]. For small mammals, the amount of habitat in the landscape was also a variable that significantly influenced both the richness [39] and the composition of species in Cerrado areas [31].

Medium- and large-sized mammals encompass a wide variety of species with different habitat uses and requirements [11], requiring large areas for survival and having low reproductive rates. These characteristics make this biological group vulnerable to landscape changes, mainly the larger species [6]. The reduction in habitat amount limits the availability of resources, reduces colonization rates, alters reproductive success, imposes a restriction on maximum population size, and, in an extreme condition, exposes populations to an increased risk of local extinction [15], altering species richness and composition. In addition, habitat reduction can make mammals, especially larger ones, more susceptible to hunting by humans and natural predators [12], increasing the risk of local extinctions.

Our results corroborate the habitat amount hypothesis proposed by Fahrig [19]. However, for medium- and large-sized mammals, the past amount (the year 2000) of habitat in the landscape was more relevant than the current amount of habitat in the landscape, evidencing a response delay (extinction debt) for the richness and composition of species of this biological group after disturbance in the environment. Extinction debt can be assumed when the past characteristics of a given landscape better explain the current species richness than the current characteristics of that landscape [21,27], as observed in this study (Figure 6).

It is known that several changes in biodiversity can be observed shortly after changes in the structure of the landscape. However, some species only decline and disappear after a long period [21]. In landscapes with little change in habitat, many species can persist for a long time without adverse impacts. In contrast, local extinctions are expected to occur rapidly in landscapes where habitat loss has been very intense [40]. The observed response delay in the present study can be attributed to these dynamics. As little change in the landscape was observed between the periods evaluated, the richness and composition of mammal species are still strongly influenced by the past structure of the landscape. In this sense, Rocha et al. [41] did not observe a significant change in the richness of medium- and large-sized mammal species in the short term (2–3 years) after a reduction of 18% in the amount of habitat in a Cerrado landscape. However, changes are expected in the long term.

Using the amount of habitat as the only landscape variable to explain species richness is still controversial. Haddad et al. [42], based on the results of controlled experiments conducted with plants and micro-arthropods, refuted the habitat amount hypothesis and indicated that long-term studies are needed to understand the effects of fragmentation, including other landscape variables. Torrenta and Villard [43], studying birds in forest fragments in Canada, concluded that the amount of habitat in the local landscape needs to be a sufficient predictor of species richness. For mammals, existing studies [6,31,37,38,39] indicate that the amount of habitat in the local landscape is the primary predictor variable for species richness.

The high dispersal capacity of mammals and the need for large home ranges, especially for larger species, means these animals move through the non-habitat matrix and use other habitat fragments available in the local landscape. In addition, for many species (such as large felines), more than the area of a single fragment may be needed to support their populations, and they need to use the nearby habitats available in the local landscape. Therefore, the amount of habitat in the landscape has stood out as the most important predictor variable for the richness of this biological group.

Our results indicate that the amount of habitat in the past landscape (14 to 18 years before sampling) was the best predictor for the richness and composition of species of medium- and large-sized mammals, reinforcing that studies on fragmentation should consider the extinction debt and the long-term effect, as mentioned by Haddad et al. [42]. Therefore, the habitat amount hypothesis applied to medium- and large-sized mammals in the Cerrado areas studied but considering the past situation of the landscape.

5. Conclusions

Our results indicate that metrics representing the amount of available habitat in the sampled fragment (AREA) or the landscape (HA) are positively related to the richness of medium-sized mammal species. However, no influence of the number of fragments in the landscape (NP) on mammal species richness was observed.

The amount of habitat in the landscape (HA) in the year 2000 was the most important variable for mammal species richness, indicating that the habitat amount hypothesis (Fahrig 2013) applies to medium- and large-sized mammals in the studied Cerrado areas, but considering the past landscape (14 to 18 years before sampling).

Both current and past metrics of habitat amount in the landscape (HA) and sampled fragment size (AREA) significantly influenced mammal species composition. However, HA proved to be a stronger predictor than AREA.