Introduction
Owing to common ancestry, close relative species tend to explore similar resources (i.e. phylogenetic niche conservatism), which can constrain their coexistence (Losos Reference Losos2008). Resource partitioning is thought to be a critical mechanism underlying the coexistence of species that feed on the same types of resources (MacArthur and Levins Reference MacArthur and Levins1967), which may be especially important for groups that underwent rapid radiation. This is the case of the genus Sporophila, which is composed of about 40 species that often coexist and share food resources in their breeding sites (Di Giacomo Reference Di Giacomo2005; Silva Reference Silva1999). However, how the coexistence of such recently evolved and morphologically similar species takes place and the extent of resource sharing within this group remains poorly investigated. Importantly, understanding which resources the threatened species of this group rely on is relevant to identify, protect, and/or restore habitats that can effectively sustain viable populations.
Seedeaters are some of the most iconic Neotropical grassland-dependent species, most of them being medium-distance migrants who feed primarily on seeds of grasses (Azpiroz et al. Reference Azpiroz, Isacch, Dias, Di Giacomo, Fontana and Palarea2012; Serafini et al. Reference Serafini, Martins-Ferreira, Bencke, Fontana, Dias and Repenning2013; Silva Reference Silva1999). Although multiple congeneric species co-occur in flocks during migration (Silva Reference Silva1999) and occupy the same breeding areas (Di Giacomo Reference Di Giacomo2005), little is known on whether and to what extent seedeater species overlap in the food resources used (Ilha and Ragusa-Netto Reference Ilha and Ragusa-Netto2023). Furthermore, knowledge of the plants consumed by seedeaters is largely based on opportunistic observations. In fact, to date, there is no comprehensive inventory of the plants used by seedeaters in breeding sites occupied by multiple species.
Several seedeater species are threatened with extinction owing to increasing rates of habitat loss especially due to the conversion of grasslands into crop fields (Azpiroz et al. Reference Azpiroz, Isacch, Dias, Di Giacomo, Fontana and Palarea2012; Di Giacomo Reference Di Giacomo2005; Serafini et al. Reference Serafini, Martins-Ferreira, Bencke, Fontana, Dias and Repenning2013). In this scenario, knowledge of which plants these seedeaters consume and whether and how species share resources may be helpful to inform habitat protection initiatives and grassland management. Here we surveyed plant species consumed by three seedeater species in Brazilian upland grasslands over four breeding seasons and tested the extent of resource partitioning among these species. All three species studied are threatened with extinction at local, regional. and global levels, namely Tropeiro Seedeater Sporophila beltoni: globally “Vulnerable” (VU); Brazil = “VU”; Rio Grande do Sul (RS state) = “Endangered” (EN); Santa Catarina (SC state) = “Critically Endangered” (CT); Black-bellied Seedeater Sporophila melanogaster: globally “Near Threatened (NT); Brazil = “VU”; RS = “EN”; SC = “VU”; Tawny-bellied Seedeater Sporophila hypoxantha: Brazil = “VU”; RS = “VU”; SC = “VU” (BirdLife International 2017, 2019; CONSEMA/SC 2011; MMA 2022; SEMA/RS 2014).
Methods
Data were collected during four consecutive breeding seasons (October–March) from 2007 to 2011 in upland grasslands in the municipalities of Bom Jesus (area 1: 28°40’S, 50°28’W, with the occurrence of S. beltoni and S. hypoxantha; area 2: 28°35’S, 50°24’W, with S. melanogaster; area 3: 28°29’S, 50°43’W, with S. beltoni and S. hypoxantha), and Vacaria (area 4: 28°8’S, 50°54’W, with S. beltoni), in Rio Grande do Sul state, and Lages and São Joaquim (area 5: 28°18’S, 50°17’W, with S. beltoni, S. hypoxantha, and S. melanogaster), in Santa Catarina state, southern Brazil. In total, the areas studied covered ~2,160 ha. During the breeding season, we carried out slow ad libitum walks within breeding sites in each location to search for nests in the grassland. Searches took place throughout the day when individuals detected feeding on seeds were observed at a distance until they flew away. The observer then approached the spot and identified the plant consumed. Plants unidentified in the field were collected and later identified by specialists (see Acknowledgements).
We used seed consumption data to build quantitative interaction matrices, where each row represents a bird species i, each column represents a consumed plant species j, and intersections aij represent the number of interactions observed between two species. Each feeding event observed was considered as an independent interaction regardless of the number of seeds consumed.
We tested the structure of this interaction matrix for nestedness and modularity. A nested structure would occur if the more specialist bird species feed on subsets of the resources used by the more generalist bird species, while only the more generalist bird species would consume seeds of the plants rarely consumed. We quantified nestedness using the WNODF metric, which quantifies the non-overlap and decreasing fill of quantitative matrices (Almeida-Neto and Ulrich Reference Almeida-Neto and Ulrich2011). Modular structure would occur if the identity and frequency of plants consumed differ substantially among bird species. We quantified modularity using the metric Q and the DIRPLPAwb+ algorithm, which searches for the optimal division of the matrix into modules (Beckett Reference Beckett2016). Modules represent subsets of birds and plants interacting more among themselves than with other species in the community. To test the significance of nestedness and modularity, we used 1,000 randomisations of the vaznull null model. This null model reshuffles interactions within random matrices while preserving the same dimensions (i.e. number of species in each trophic level) and connectance (i.e. number of cells filled) as the observed matrix as well as keeps nearly constant marginal totals (i.e. the sum of interactions on each row and column). We considered a metric to be statistically significant if the observed values fell above the 95% confidence interval (CI) expected by the null model. All analyses were run in the R-package “bipartite”.
Results
The three seedeater species fed on seeds of 62 plant species in total, summing up 692 feeding events observed throughout the four breeding seasons. Sporophila hypoxantha fed on 32 plant species (257 foraging events), while S. melanogaster fed on 27 plants (187 events), and S. beltoni on 33 plant species (248 events). All species belonged to the family Poaceae, except for a Cyperaceae and an Asteraceae species.
The plant–seedeater interaction network was not nested (WNODFobs = 24.77; 95% CI WNODFnull_vaznul = 30.81–38.18), but it was modular (Qobs = 0.49; 95% CI Qnull_vaznull = 0.12–0.22), presenting three modules (Figure 1). Each module corresponds to one bird species and the plants on which it was the most frequent or exclusive consumer.
Figure 1. Plant-seedeater network in Brazilian upland grasslands, presented a modular structure indicating low overlap in food resource use among bird species. Bar thickness indicates the number of feeding events recorded between a bird and a plant species. Each colour (orange, black, and yellow) indicates a distinct module. Grey links indicate interactions between species that belong to distinct modules.
Only seven plant species (11.3% of the total) were consumed by all seedeaters (Figures 1 and 2). All three birds had high numbers of exclusive resources (S. melanogaster 15 species, S. beltoni 13 species, and S. hypoxantha 11 species). Seedeater species presented overall low overlap in the resources used, with the highest overlap being between S. beltoni and S. hypoxantha (17 species), and the lowest between S. beltoni and S. melanogaster (9 species).
Figure 2. Venn diagram showing the partitioning of seeds of plant species (number of species shared) among three seedeater species in Brazilian upland grasslands.
The plants consumed most frequently by each bird species were distinct. For each bird species, over half of the foraging events observed included only three plant species. Specifically, Setaria parviflora, Sorghastrum pellitum, and Paspalum plicatulum correspond to 57.2% (147 foraging observations) of the seeds consumed by S. hypoxantha; Paspalum exaltatum, Carex brasiliensis, and Andropogon lateralis correspond to 61.5% (115 events) of the seeds consumed by S. melanogaster; Piptochaetium montevidense, P. stipoides, and Paspalum guenoarum correspond to 50.8% (126 events) of the seeds consumed by S. beltoni (Table 1).
Table 1. Plant species whose seeds were consumed by each of the three seedeater species on Brazilian upland grasslands during four breeding seasons (2007–2011). Cells are filled with the number of foraging events observed
Discussion
Our findings show the existence of considerable resource partitioning among the three sympatric seedeater species studied. Although all three species fed almost exclusively on seeds of Poaceae, the main plant species consumed differed substantially among seedeater species, which resulted in the emergence of a modular structure in the consumer–resource network as well as its lack of nestedness.
Multiple non-exclusive factors may explain the low overlap of resources consumed by the three species studied here. The most plausible explanation is that each bird species occupies different habitats and, therefore, interacts with distinct subsets of plants that are more common in their preferred microhabitats (e.g. S. hypoxantha and S. beltoni). In fact, while S. melanogaster regularly occupies wet grasslands (bogs) in higher grounds, S. hypoxantha and S. beltoni are more associated with drier grasslands in inner valleys in our study sites (Franz and Fontana Reference Franz and Fontana2013; Repenning and Fontana Reference Repenning and Fontana2016, Reference Repenning and Fontana2019; Rovedder and Fontana Reference Rovedder and Fontana2012).
In addition, bill size may also contribute to the observed pattern of resource partitioning. Sporophila beltoni has a larger bill than the other seedeaters studied (Porzio et al. Reference Porzio, Repenning and Fontana2019) and, thus, may consume larger seeds that are inaccessible or less efficiently consumed by the small-billed seedeaters. This species shares the same foraging habitats with S. hypoxantha, but with low overlap in consumed resources such as, for instance, the large seeds of Piptochaetium stipoides which are frequently consumed by S. beltoni but not by the small-billed species. As in mutualistic networks involving birds and flowers or fruits, morphological matching between bills and plant structures are known to mediate resource partitioning, leading to the emergence of modular interaction networks (Maruyama et al. Reference Maruyama, Vizentin-Bugoni, Oliveira, Oliveira and Dalsgaard2014; Sonne et al. Reference Sonne, Vizentin-Bugoni, Maruyama, Araujo, Chávez-González and Coelho2020). However, explicit tests on whether the sizes of seeds consumed by distinct seedeater species differ are still lacking and we encourage further studies that measure and compare such traits directly.
An alternative explanation is that foraging sites are defined by the choice of breeding sites, which is known to be defined by vegetation structure and floristic composition (Franz and Fontana Reference Franz and Fontana2013; Repenning and Fontana Reference Repenning and Fontana2019; Rovedder and Fontana Reference Rovedder and Fontana2012). If birds forage near the nests to minimise energy costs associated with travel distances, the differences observed in the plants consumed may be an indirect product of breeding site choices. These sites may also have distinct microclimatic conditions and soil characteristics that define plant composition and may ultimately influence the resources consumed.
Another potential explanation for the high resource partitioning observed could be that seedeaters compete for seeds, with some species dominating the most profitable resources and displacing competitor species to other plant species. However, while seedeaters are known to exhibit agonistic behaviour towards congeneric species in breeding sites (Turbek et al. Reference Turbek, Browne, Di Giacomo, Kopuchian, Hochachka and Estalles2021), this potential explanation is less plausible since grass seeds are highly abundant in the studied sites and thus, is unlikely to represent a limiting resource.
Our findings have important implications for the conservation of these threatened seedeaters. Since the three species rely predominantly on distinct subsets of plant species, protected areas need to encompass a vegetation mosaic that includes the distinct preferred plant communities to benefit all three species. Furthermore, by identifying plants frequently used by these seedeaters, it is possible to design management strategies that promote their availability during the breeding period which may potentially boost adult survival and reproductive success. As 66.9% (463 out of the 692) of the foraging events observed included only nine plant species (namely, Setaria parviflora, Sorghastrum pellitum, Paspalum plicatulum, P. exaltatum, P. guenoarum, Carex brasiliensis, Andropogon lateralis, Piptochaetium montevidense, and P. stipoides), grassland restoration initiatives that favour such plants, for example, by adjusting livestock load or producing (or harvesting) and sowing their seeds, may benefit these seedeaters in their breeding areas. Furthermore, two out of the nine plant species mentioned are cool-season C3 grasses (Piptochaetium spp.), which are significantly affected by grassland fires carried out at the end of the austral winter (August/September) when these plants are fruiting (Boldrini Reference Boldrini1997). Compliance with the legislation that regulates this management practice (CONAMA Resolution 423/2010) by, specifically, not burning the field after August and skipping at least three years without burning, therefore, represents a measure that would favour resource availability for the threatened Sporophila from the Brazilian upland grasslands.
Recently, grassland restoration has gained momentum in South America and several strategies have been suggested depending on the level of degradation (Andrade et al. Reference Andrade, Koch, Boldrini, Vélez-Martin, Hasenack and Hermann2015; Thomas et al. Reference Thomas, Overbeck, Dutra-Silva, Porto, Rolim and Minervini-Silva2023). In the case of degradation by overgrazing when plant species composition has not been dramatically changed, it may be sufficient to implement management initiatives that reduce grazing such as reducing livestock overload or temporary fencing (from early spring to mid-summer) the areas where such plants occur to facilitate plant reproduction and increase seed availability. Permanent fencing is not recommended as vegetation tends to encroach and disfavour seedeaters, but it may be an option if combined with mowing or controlled grazing or fire, as it has been shown to increase grassland bird diversity (Silva and Fontana Reference Silva and Fontana2020). In the case of grassland restoration following land conversion, the plants that seedeaters feed on may not be capable of spontaneous re-colonisation from the regional species pool, especially in highly degraded landscapes where sources of seeds may be scarce or absent, or even grassland self-recovery may not occur or achieve alternative stable states (Andrade et al. Reference Andrade, Koch, Boldrini, Vélez-Martin, Hasenack and Hermann2015). In this case, outplanting or sowing seeds may be necessary. However, as Paspalum notatum is the only species sold in the market among the plants consumed by these seedeaters (Rolim et al. Reference Rolim, Rosenfield and Overbeck2022), harvesting seeds from natural areas may be necessary for effectively restoring the vegetation. This suggestion assumes that physical and chemical features of the soil have not passed a threshold where the establishment of grasslands becomes impossible without, for example, recovery of soil features and control of invasive species (Andrade et al. Reference Andrade, Koch, Boldrini, Vélez-Martin, Hasenack and Hermann2015). In any case, we stress that the most effective initiative to conserve these seedeaters is to protect the mosaic of unconverted habitat where these species breed (Franz and Fontana Reference Franz and Fontana2013; Repenning and Fontana Reference Repenning and Fontana2019; Rovedder and Fontana Reference Rovedder and Fontana2012) and their food plants are abundant.
Acknowledgements
We thank Rafael Trevisan (UFRGS) for plant identification and taxonomic update; Mariana L. Gonçalves, Gabriel de La Torre, and many colleagues for help during fieldwork; landowners for their hospitality and permission to work on their properties. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES; Finance Code 001 to JVB; PNPD process 88882.314826/2019-01 to IF), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq; CSF process 303318/2013-9, 308700/2022-8), Neotropical Grassland Conservancy, Fundação Grupo Boticário de Proteção à Natureza, Idea Wild Small Equipment Grants, and supported by IGRÉ – Associação Sócio Ambientalista.
