Species composition and community structure of dung beetles (Coleoptera: Scarabaeidae: Scarabaeinae) compared among savanna and forest formations in the southwestern Brazilian Cerrado

Although dung beetles are important members of ecological communities and indicators of ecosystem quality, species diversity, and how it varies over space and habitat types, remains poorly understood in the Brazilian Cerrado. We compared dung beetle communities among plant formations in the Serra Azul State Park (SASP) in the state of Mato Grosso, Brazil. Sampling (by baited pitfall and flight-interception traps) was carried out in 2012 in the Park in four habitat types: two different savanna formations (typical and open) and two forest formations (seasonally deciduous and gallery). A total of 5,400 individuals collected comprised 57 species in 22 genera. Typical savanna had the greatest species richness and abundance, followed by open savanna and deciduous forest, while the gallery forest had the fewest species but high abundance. Tunnelers (one of three main nesting behavior guilds) showed the greatest richness and abundance (except in the gallery forest, where one dweller species was extremely abundant) in all plant formations. We found that species richness and abundance of the dung beetle community are influenced by differences among plant formations. Habitat heterogeneity in the different plant formations along with anthropic influences (fire, habitat fragmentation) are cited as important factors that explain guild and species richness and distribution patterns. These results emphasize the importance of protected areas, such as SASP, for the maintenance and conservation of species diversity in the Brazilian Cerrado.


INTRODUCTION
Dung beetles, found world-wide, are extremely diverse in the tropics (Halffter and Matthews 1966, Davis and Scholtz 2001, Scholtz et al. 2009). Often used as ecological indicators in diversity studies Forsyth 1998, Halffter andFavila 1993), dung beetles are taxonomically well known, easily sampled using low-cost and standardized protocols (Halffter andFavila 1993, Gardner et al. 2008), which permits comparisons of community structure at different times or locations (Hernández et al. 2019). Dung beetles also are important in ecosystem maintenance due to their influences on nutrient cycling, secondary seed dispersal, soil aeration, and parasite suppression (Halffter and Favila 1993, Nichols et al. 2007, 2008, Scholtz et al. 2009).
Dung beetles can be divided into three main nesting behavior guilds (hereafter, simply "guild") by the way they use the food resource: (1) rollers form balls of feces or carrion that they then roll and bury some distance from the source, (2) tunnelers bury Here, we compare dung beetle community structure and composition among four vegetation formations (typical savanna, open savanna, deciduous forest, and gallery forest) in Serra Azul State Park (hereafter, SASP), a protected park in the state of Mato Grosso, Brazil. Specifically, we test how taxonomic and guild composition and structure are associated with vegetation formation in the study area. Based on the geographical location of the SASP and the predominance of savanna formations in its limits, we hypothesize that these habitats show high richness and abundance, and that the composition and guild structure vary according to the particular habitat types in which they are found.

MATERIAL AND METHODS
We studied dung beetles in the SASP, in the state of Mato Grosso (15°51'S; 52°16'W), Brazil (Fig. 1). SASP was created in 1994 encompassing around 110 km 2 at 350-750 m elevation. The dry season (May to September) and wet season (October to April) are well-defined, with annual rainfall of 1447-1528 mm, and temperature varying from 21.8-28.1°C (Pirani et al. 2009). Following RAPELD (method adopted in long-term ecological research sites -PELD in Portuguese), plots were established in the state park. Collection was carried out in December 2012 on 10 plots, with a distance of 1000 m between plots of 250 m transects that maintain a constant elevation (Magnusson et al. 2005). The vegetation formation in each plot was previously classified following the protocol of the Biodiversity Research Program (with the Brazilian acronym PPBio, Programa de Pesquisa em Biodiversidade) adapted locally by the ComCerrado Network ( Fig. 1). Four plots were classified as typical savanna, another four as open savanna (both savannas had 356 to 1463 trees ha -1 ), and the other two plots were gallery forest (684 trees ha -1 ) and deciduous forest (586 trees ha -1 ). Vegetation height in both savannas varied from 1 to 13 m, while in both forests height varied from 2.2 to 25 m (Sousa et al. 2019).
Each 250 m transect was divided into 5 sampling points 50 m apart (the span of 200 m was within the 250 m). At each point, three pitfall traps were established at the angles of a triangle (each leg 3 m in length) centered on the point. Pitfalls were 19 cm diameter, 11 cm deep round plastic containers placed in the ground with the rim at the level of the soil surface. A roof to prevent pitfalls from filling with rainwater was built over each. A smaller recipient was placed at an edge of each pitfall to contain approximately 20 g of human feces used as bait. Beetle captures in the three pitfalls were combined to make each point a replicate, for a total of five sampling points per transect in ten transects. Two flight interception traps (FIT) were placed in each of the two savanna types, and one in each forest type, for a total of six traps. To preserve captured insects, approximately 250 ml of saline-detergent solution was placed in each pitfall and flight interception trap. After traps were open for 48 hours, trap contents were placed in 70% alcohol and transported to the laboratory at  Grosso (CEMT) in Cuiabá, where the material was then deposited.
Community comparisons between plant formations included only captures from pitfalls, while captures from all traps were used in community-wide summaries of guilds and total species richness. We compared species diversity among the plant formations using sample-size-based rarefaction and extrapolation curves (Magurran and McGill 2011), using the iNEXT function in the package iNEXT (Hsieh et al. 2020) only for Hill numbers of q = 0 (species richness) with the maximum reference sample size (Chao et al. 2014). We compared the proportion of individuals and species in each of the guilds among plant formations using the Chi-square test. In addition, we plotted the standardized abundance rank by log 10(n+1) in the four habitats, which allowed us to compare species richness and abundances among the habitat types.
As the distance between samples within each plot is short and dung beetles are volant and search for resources, spatially close samples may show greater similarity in species composition than distant samples, causing spatial autocorrelation. To test for spatial autocorrelation of the dung beetle community we used Mantel correlogram . The data was first detrended so that second-order stationarity must be met and the number of the distance class was calculated using Sturge's rule (Borcard et al. 2018). The standardized Mantel statistic has the same formula as the Pearson correlation coefficient, but it is computed between the values in distance or similarity matrices X and Y (Borcard and Legendre 2012). The condition of normality was relaxed because significance was tested using permutation. The statistics were tested for significance using 999 permutations and plotted against distance classes to form the Mantel correlogram. A positive (or negative) correlation (rM) indicates that for the given distance class, the multivariate similarity among sites is greater than (or less than) that expected by chance (Borcard and Legendre 2012).
In the case of important spatial autocorrelation, we used Distance-Based Moran's Eigenvector Maps (dbMEM) as a means of quantifying the spatial effect in spatial distribution of the dung beetles community. We used Euclidean distances and the function "dbmem" in the adespatial package (Dray et al. 2020). This generated nine dbMEM that may represent the effect of dispersion as well as unmeasured variables that are also spatially structured (Peres-Neto and Legendre 2010, Fernandes et al. 2014). The dbMEMs used in the non-parametric multivariate analysis of variance (PERMANOVA) were selected using the forward selection routine (forward.sel) in the adespatial package (Dray et al. 2020).
To compare the community structure among plant formations [F] and control for the effects of spatial structure [S], and their interactions [F|S] (if any), we used PERMANOVA (Anderson 2001). A dissimilarity matrix was calculated using Bray-Curtis distances with the Hellinger transformation Gallagher 2001, Legendre and. We compared dung beetle species among plant formations using the pairwise.perm.manova function in the package RVAideMemoire (Hervé 2020). Additionally, we used PERMDISP (analysis of multivariate homogeneity of group dispersions) to identify location or dispersion effects (Anderson and Walsh 2013) using the function "betadisper" in the vegan package (Oksanen et al. 2020). We graphically illustrate the distributions of dung beetle assemblages in the different plant formations using Principal Coordinates Analysis (PCoA) . As with PERMANOVA, species compositions were used to estimate dissimilarity matrices between formations using Bray-Curtis distances after Hellinger transformation. In addition, dbMEM were also represented by vectors in the PCoA that illustrate the relationship between spatial attributes and communities of dung beetles in these plant formations. Vectors were derived with the function "envfit" of the vegan package (Oksanen et al. 2020). All analyses were carried out using R (R Core Team 2020).
The greatest number of species (47 spp., 82.45% of the total) and greatest abundance (1962 individuals, 36.33% of the total) were from typical savanna. Open savanna followed with 39 species (68.42%) and 1804 individuals (33.40%). Deciduous forest had 21 species (36.84%) and 256 individuals (4.74% of the total), while gallery forest had 18 species (31.57%) and 1378 individuals (25.51%), of which 1032 (74.89%) were of a single species, Eurysternus caribaeus (Herbst, 1789) (Table 1). Sample-size-based rarefaction and extrapolation curves showed that overall species richness (q = 0) was similar between the two savanna formations and the grand total from all plant formations (Fig. 2). The two forest formations were the least diverse, but the overlapping confidence intervals of the deciduous forest indicates no difference with the open savanna (Fig. 2). The richness estimated for the typical savanna was 59 dung beetle species, followed by open savanna with 49 species, deciduous forest with 27, and gallery forest with 11 (Fig. 2). Tunneling dung beetles were the most species-rich (29 species, 50.87% of the total), followed by rollers (12 spp., 21.05%), and dwellers (7 spp., 12.28%). Tunneling beetles were also the most abundant, with 2365 individuals (43.79% of the total), followed by dwellers (1842, 34.11%), and rollers (822, 15.22%). Nine species (371 individuals) could not be placed in any particular guild ( Table 1). The proportion of species per guild was independent of plant formation (tunnelersχ 2 = 0.84, p = 0.83; rollersχ 2 = 0.81, p = 0.84; dwellersχ 2 = 0.79, p = 0.85), with the tunnelers having the most species in all formations, followed by rollers and dwellers (Fig. 3). However, relative abundance varied by habitat (tunnelersχ 2 = 32.69, p < 0.001; rollersχ 2 = 40.46, p < 0.001; dwellersχ 2 = 62.171, p < 0.001), with the tunnelers having the most individuals in almost all formations, except in the gallery forest, where dwellers were the most abundant, and both forests had fewer rollers (Fig. 3).  While in the gallery forest, where one dweller species (E. caribeus) comprised 74.89% of the individuals collected, the two savanna formations were more similar and had no strongly numerically dominant species (Fig. 4)   1840 and Canthon aff. simulans were equally abundant in both savanna formations. In the forested sites, aside from the previously mentioned E. caribaeus, Dichotomius aff. carbonarius was relatively abundant in both. However, in the deciduous forest, Onthophagus aff. hirculus was somewhat more abundant than D. aff. carbonarius, which was followed by E. nigrovirens; the latter was common in the savannas (Fig. 4). The Mantel Correlogram indicated that spatial autocorrelation existed. At the first distance class (i.e., 29.06 to 471.63 m), the correlation was strongest and positive (r = 0.37, p = 0.001), and weaker and negative (r = -0.20, p = 0.001) at the second distance class (i.e., 471.63 to 914.63 m). Thus, spatial autocorrelation was found at less than 914 m (Fig. 5). All subsequent correlations were |r| < 0.19, indicating that any additional spatial autocorrelation was relatively unimportant (see supplementary material for more details, Table S1).

DISCUSSION
Dung beetle communities are more species-rich and species are more abundant in the savanna as compared to forest formations in the SASP. In addition, species composition and guild structure varied between the two formations. Dung beetles are diverse in the Brazilian Cerrado (mostly savanna, but within which are patches of gallery and dry-deciduous forests, much like those we examined here), where savanna formations tend to be more diverse and species-rich than forests (Milhomem et al. 2003, Endres et al. 2007, Silva et al. 2010. The tribes Coprini, Deltochilini, Ateuchini, and Phanaeini were the most speciose (Costa et al. 2009, Daniel et al. 2014, and they also tend to be more common and diverse in the savanna formations. We collected a total of 57 species in a region where more than 50 are often collected (Milhomem et al. 2003, Almeida and Louzada 2009, Daniel et al. 2014), but our interpolation and extrapolation curves suggest that more than 70 species may eventually be found in the Cerrado as a whole (Fig. 2).
Most species captured in all habitat types were tunnelers, followed by rollers and dwellers, a result that appears to be the general pattern for dung beetles in the Neotropical region (Halffter et al. 1992, Louzada and Lopes 1997, da Silva and Di Mare 2012, Campos and Hernández 2013, Silva et al. 2015. In addition to being the most diverse group, tunnelers are also the most abundant dung beetles in all habitat types. Although in the same guild, tunneler habits are not identical; different species have different patterns of tunneling, beneath or alongside the feces, and this variation perhaps allows more species to use the same resource in different ways, thereby avoiding competition, a point that may explain the greater abundance of this group (Daniel et al. 2014, Silva et al. 2015. Both rollers and dwellers are also more abundant and species-rich in the savanna formations, with the exception of the dweller E. caribaeus, which was super-abundant in the gallery forest. This is surprising because the more humid environment of the forest is expected to generate microclimates that favor these guilds because feces should dry more slowly (Hanski and Cambefort 1991). It is noteworthy that the interaction between tree density, canopy height, and microclimate is complex. Thus, because the range of tree density in the savanna formations may be greater than in the two forest types (while canopy height is always greater in the forests), tree density in savannas may sometimes generate the microclimate that favors rollers and dwellers. Testing the fine details that determine habitat quality for dung beetles was outside the scope of this study, but future ecological studies should measure microhabitats at the point of capture to better understand these dynamics in the SASP.
No single species was dominant in the savanna formations where two or more species tended to be similarly abundant, in contrast to the forests, especially gallery forest. Dominance has been found in some dung beetle assemblages in both savannas and forests (Almeida and Louzada 2009). Varying patterns of dominance may be explained in part by resource abundance (or scarcity), where species may be co-dominant (Kadiri et al. 1997), or due to habitat modification (Halffter et al. 1992, Halffter and Arellano 2002, Nichols et al. 2007, or perhaps even time of year when collections took place. Future study should examine these possibilities, especially with respect to the surprisingly abundant E. caribaeus in the gallery forest. Some spatially structured processes may also influence local species compositions in dung beetle communities (Louzada et al. 2010). Thus, for understanding dung beetle assemblages in the SASP, dispersal, habitat heterogeneity, and human-caused modifications may all be important spatial processes. In the Brazilian Cerrado, habitat heterogeneity is due to the variety of plant formations (Silva et al. 2010). These authors compared dung beetle communities in forest and savanna in Chapada dos Parecis, Mato Grosso, showing that less heterogeneity in savanna habitats ("campos sujos") caused greater species richness than in gallery forest (with greater heterogeneity), a point that could explain our results. Alternatively, the greater heterogeneity of habitats (related to a greater number of plant species and greater plant stratification) may promote greater richness and abundance of dung beetles (Almeida and Louzada 2009). In the SASP, studies demonstrated that many species that are often very abundant are found in typical savanna habitats, and that plant cover increases with rainfall (Pirani et al. 2009, Santos et al. 2017, when we carried out this study. Thus, it is possible that the variety of species and their abundance in the typical savanna is due to the variety of habitats available as a consequence of structural complexity.
Although the SASP is a protected area for conservation, anthropic pressures, such as forest fires, can influence habitat structure and availability, thereby influencing the dung beetle community. For example, forest fires occurred in 2002, 2005, and 2007, when the entire Park burned ). In the savanna, the consequences of forest fires could be less severe to plant communities than those found in forests . Because fires influence indirectly dung beetle communities through habitat modification (Louzada et al. 1996, de Andrade et al. 2014, we predict that fires will have a smaller impact on dung beetle communities in savannas than in forests because of the smaller effect fires have on savannas.
Considering the history of SASP occupation and its surroundings, deforestation is cited as one of the most important anthropic pressures on the Park (FEMA 2000, Santos et al. 2006, causing fragmentation and loss of habitat. The influences on dung beetle communities of the size of the habitat fragment, its isolation, and the surrounding matrix have been mentioned in several studies (Estrada and Coates-Estrada 2002, Milhomem et al. 2003, Nichols et al. 2007, Filgueiras et al. 2011). As SASP is within a predominant savannah matrix (FEMA 2000, we infer that forests were more strongly influenced by isolation and habitat loss, thus reducing the richness and abundance of beetle species in these formations, especially the gallery forests that may present communities most affected by these factors (Milhomem et al. 2003, Silva et al. 2010. We consider that this regional landscape mosaic, formed by combinations of forests and savannas of the Brazilian Cerrado, has been an important determinant of the dung beetle communities found in the SASP. For these reasons, regional diversity is very high. We believe that additional measures are necessary for protecting the SASP. These measures should take into account the complex matrix of habitat types and species that occur in each of these habitats (Almeida and Louzada 2009). Additionally, we recommend that conservation efforts should be directed towards reducing anthropic pressures within the SASP. These actions would favor stability of the distinct habitats. Such stability would in turn be beneficial for the dispersal of dung beetles among habitat types in the area.