Introduction
In drylands, anthropogenic disturbances such as logging, fire, agricultural expansion, and livestock grazing practices can result in loss of habitat and connectivity, with consequent biodiversity losses (Jones, Reference Jones2000; Asner et al., Reference Asner, Borghi and Ojeda2003; Millennium Ecosystem Assessment, 2005; Reynolds et al., Reference Reynolds, Stafford Smith, Lambin, Turner, Mortimore, Batterbury, Downing, Dowlatabadi, Fernández, Herrick, Huber-Sannwald, Jiang, Leemans, Lynam, Maestre, Ayarza and Walker2007), and reduction of animal functional diversity (Chillo and Ojeda, Reference Chillo and Ojeda2012; Periago et al., Reference Periago, Chillo and Ojeda2014). In Argentina's drylands, where grazing by domestic animals is the main land use, most studies have focused on the effect of this disturbance on animal populations (e.g. species decrease in abundance at grazed sites) and its impact at the community level (i.e. habitat modifications, such as an increase in bare ground cover, and their impacts on different species’ populations) (Tabeni and Ojeda, Reference Tabeni and Ojeda2003, Reference Tabeni and Ojeda2005; Corbalán, Reference Corbalán2006; Tabeni et al., Reference Tabeni, Mastrantonio and Ojeda2007; Rodríguez, Reference Rodríguez2009). However, little is known about the effects of grazing on biological interactions, such as seed dispersal by animals.
Seed dispersal is crucial for plant populations because it allows new individuals to move away from their mother plant (escape hypothesis; Connell, Reference Connel, den Boer and Gradwell1971; Janzen, Reference Janzen1971), explore new areas (colonization hypothesis; Howe and Smallwood, Reference Howe and Smallwood1982; Nathan, Reference Nathan2006) and regenerate the population (Forget et al., Reference Forget, Ranking-De Merona and Julliot2001; Jordano et al., Reference Jordano, Forget, Lambert, Böhning-Gaese, Traveset and Wright2011). Animal species can contribute differently to the overall effectiveness of seed dispersal a plant receives from its assemblage of dispersal agents (Schupp et al., Reference Schupp, Jordano and Gómez2010), among them mammals. Some are endozoochorous dispersers, considered efficient primary seed dispersers that ingest the whole fruit and then defecate or regurgitate the seeds (Janzen, Reference Janzen and Janzen1983; Jordano, Reference Jordano and Fenner2000; Forget and Cuijpers, Reference Forget and Cuijpers2008; Vander Wall and Beck, Reference Vander Wall and Beck2012). Other mammal species, mainly rodents, are scatter-hoarders who store seeds in a large number of small caches in the ground for future consumption (Vander Wall and Beck, Reference Vander Wall and Beck2012); and as they eat some seeds but not all of them, they become potential seed dispersers (Vander Wall, Reference Vander Wall1990, Reference Vander Wall2002; Forget et al., Reference Forget, Hammond, Milleron, Thomas, Levey, Silva and Galetti2002). In contrast, some hoarding species are mainly seed predators, because they store food in underground caches, generally in unfavourable conditions for seed germination and seedling establishment (larder-hoarders; Longland et al., Reference Longland, Jenkins, Vander Wall, Veech and Pyare2001; Vander Wall and Beck, Reference Vander Wall and Beck2012).
In this context, we conducted a study in a protected area (Ñacuñán Reserve) surrounded by cattle range areas dominated by Prosopis flexuosa (‘algarrobo dulce’) woodland in a dryland biome of Argentina. Mammals removing P. flexuosa fruits provide different effectiveness to this tree species (Campos and Ojeda, Reference Campos and Ojeda1997; Campos et al., Reference Campos, Giannoni, Taraborelli and Borghi2007, Reference Campos, Peco, Campos, Malo, Giannoni and Suárez2008; Campos and Velez, Reference Campos and Velez2015). Particularly wildlife species, some are opportunistic frugivores that disperse seeds through endozoochory (e.g. Lycalopex griseus, Dolichotis patagonum; Campos and Ojeda, Reference Campos and Ojeda1997; Campos et al., Reference Campos, Peco, Campos, Malo, Giannoni and Suárez2008; Campos and Velez, Reference Campos and Velez2015), whereas small rodents behave as scatter-hoarders (Microcavia australis, Eligmodontia typus; Giannoni et al., Reference Giannoni, Dacar, Taraborelli and Borghi2001, Reference Giannoni, Campos, Andino, Ramos-Castilla, Orofino, Borghi, de los Ríos and Campos2013; Campos and Velez, Reference Campos and Velez2015; Campos et al., Reference Campos, Campos, Giannoni, Rodríguez, Albanese and Cona2017) or seed predators through larder-hoarding (Graomys griseoflavus, Akodon dolores; Giannoni et al., Reference Giannoni, Campos, Andino, Ramos-Castilla, Orofino, Borghi, de los Ríos and Campos2013).
We focused on studying the effects of cattle grazing activity on P. flexuosa seed removal by terrestrial wildlife mammals at two times of the fruiting period. We had the following questions: (1) does grazing affect total seed removal by mammals?; (2) is there any temporal difference in total seed removal during different times of the fruiting period of Prosopis?; and (3) are there any differences in seed removal by different functional mammal groups between the grazed and ungrazed sites during the beginning and by the end of the fruiting period? If the ungrazed site is functioning as a conservation site, we expected a higher total seed removal in ungrazed than in grazed woodland, because undisturbed woodland provides mammals with more food and refuge (Tabeni and Ojeda, Reference Tabeni and Ojeda2005; Corbalán, Reference Corbalán2006). Because two different time periods were considered, we expected a high total seed removal at the end of the fruiting period, after peak fruit fall, when removal and consumption of fallen fruits and seeds are usually more pronounced (Janzen, Reference Janzen1974; López de Casenave et al., Reference López de Casenave, Cueto and Marone1998). Finally, we expected that seed removal by the functional mammal groups in both grazed and ungrazed woodlands would be different at different times of the fruiting period due to their natural population fluctuations and habitat preferences (Corbalán and Ojeda, Reference Corbalán and Ojeda2004; Tabeni and Ojeda, Reference Tabeni and Ojeda2005). In the study area, bare ground cover is higher at the grazed site than in the protected woodland (Tabeni and Ojeda, Reference Tabeni and Ojeda2005) and, while some native mammals that disperse Prosopis seeds, such as M. australis and D. patagonum, are associated with open habitats in the cattle-grazed woodland (Tabeni and Ojeda, Reference Tabeni and Ojeda2003; Rodríguez, Reference Rodríguez2009), and L. griseus prefers open shrubby areas (Jiménez et al., Reference Jiménez, Lucherini and Novaro2008), other species (G. griseoflavus and A. dolores) are more abundant in areas with dense plant cover such as the protected woodland (Tabeni and Ojeda, Reference Tabeni and Ojeda2003, Reference Tabeni and Ojeda2005; Corbalán, Reference Corbalán2006).
Materials and methods
Study site
This study was conducted in the Man and Biosphere Ñacuñán Reserve (34°02′S, 67°58′W, 12,800 ha; Mendoza Province, Argentina) and in an adjoining cattle ranching area (both 540 m elevation); these sites are located in the Monte biogeographic province (central Monte; Cabrera, Reference Cabrera1971; Roig et al., Reference Roig, Roig-Juñent and Corbalán2009). The climate is semi-arid, with cold dry winters (–13 to 10°C) and warm rainy summers (20 to 42°C). Average annual rainfall is 329.4 mm (Estrella et al., Reference Estrella, Boshoven, Tognelli, Claver and Roig-Juñent2001). Three main types of plant communities occur in the area: mesquite woodland, creosotebush scrubland, and sand dunes (Roig, Reference Roig1971). The mesquite woodland is the most representative community; it consists of a tree layer of P. flexuosa and Geoffroea decorticans (mean canopy height being 7 m), a shrub layer with dominance of Larrea divaricata, L. cuneifolia and Condalia microphylla, and a grass layer of Pappophorum spp. Trichloris crinita and Digitaria californica (Roig, Reference Roig1971).
At the beginning of the nineteenth century, the area of Ñacuñán suffered severe logging of mesquite woodland and cattle overgrazing (Abraham and Prieto, Reference Abraham, Prieto, García Martínez and González Jácome1999; Rojas et al., Reference Rojas, Prieto, Álvarez and Cesca2009). The reserve was established by law in 1961 with the aim of protecting the mesquite woodland (Roig, Reference Roig1971), and it was fenced in 1972. This reserve offers a unique situation for research because it is the only area in the Monte where grazing by domestic animals is excluded. The adjoining grazed field is under continuous grazing and also presents other related activities (firewood extraction, subsistence hunting, and establishment of watering points, fences, roads and human settlements). Due to environmental constraints in these rangelands, 20 ha/AU (hectares per animal unit) is a commonly used stocking rate (Guevara et al., Reference Guevara, Grünwaldt, Estevez, Bisigato, Blanco, Biurrun, Ferrando, Chirino, Morici, Fernández, Allegretti and Passera2009). Previous studies in the area showed greater abundance of adult P. flexuosa trees in ungrazed than in grazed woodlands (62 and 32 trees in 0.4 ha, respectively; Aschero and Vázquez, Reference Aschero and Vázquez2009).
Tree species
The study tree, P. flexuosa (Fabaceae, Mimosoideae), is a key species in its natural distribution range (Álvarez and Villagra, Reference Álvarez and Villagra2009; Álvarez et al., Reference Álvarez, Villagra, Villalba, Cony and Alberto2011; Campos and Velez, Reference Campos and Velez2015); it provides the human population with services such as shade and valuable goods like wood and food (Ladio and Lozada, Reference Ladio and Lozada2009).
Prosopis flexuosa blooms in spring (October to December) and fruits start to ripen in summer (February). When the ripe fruits fall, their persistence beneath tree canopies is short because animals remove most of the fruits and seeds within no more than six weeks after they reach the ground (Villagra et al., Reference Villagra, Marone and Cony2002; Campos et al., Reference Campos, Giannoni, Taraborelli and Borghi2007). The fruit is an indehiscent pod, with a relatively soft exocarp (Burkart, Reference Burkart1976) and a thick mesocarp that contains the major portion of sugars and proteins (Kingsolver et al., Reference Kingsolver, Johnson, Swier, Teran and Simpson1977). Seeds are covered by a bony endocarp and a coat that acts as a barrier inhibiting germination (Warrag, Reference Warrag1994). Germination is accelerated by the passage of seeds through the digestive tracts of animals, with some costs to seed viability (Campos and Ojeda, Reference Campos and Ojeda1997; Campos et al., Reference Campos, Peco, Campos, Malo, Giannoni and Suárez2008).
Seed removal experiment
We randomly chose adult Prosopis trees separated by a minimum distance of 500 m, both in the protected and the grazed area. Below the canopy of each tree, we placed 20 Prosopis fruits containing 15 seeds each (n = 300 total seeds per tree). Over a 48-h period, we monitored seed removal and the identity of the mammal species removing seeds from beneath each tree using camera traps (Moultrie, M-900i) mounted on trees at a height of 1.80 m. The supply of whole fruits and the short fruit exposure time helped us prevent fruit removal by ants (C.M. Campos, personal observation). The cameras took three consecutive pictures whenever animal movement was detected, with a 30-s delay between shoots. The location of the cameras allowed us not only to identify animal species, but also to count the number of remaining fruits after the animal had left the place (Grünewald et al., Reference Grünewald, Breitbach and Böhning-Gaese2010). We offered whole Prosopis fruits and, as seeds can be counted, we used the number of seeds removed by species and by functional mammal groups as response variable in the model (see ‘Data analysis’ section). Those seeds that had been moved by an animal far from the camera trap's field of view were considered removed.
To assess whether seed removal by mammals changed at two different times during the fruiting period of P. flexuosa, we conducted the experiment at the beginning of the fruiting period when fruits can be found on the ground beneath the trees (March 2015), and by the end of the fruiting period when almost no fruits are available on the ground (May 2015). The production of Prosopis fruit is highly variable between years, depending on environmental factors (Dalmasso and Anconetani, Reference Dalmasso and Anconetani1993; Álvarez and Villagra, Reference Álvarez and Villagra2009). A period of low fruit production has been observed from 2013 onwards (C.M. Campos, personal observation).
Data analysis
Because of the small sample size, we did not include in the model those species with an unknown effect on Prosopis seeds (‘unknown’ group; see Table 1). As our objective was to compare the effect of cattle grazing activity on seed removal by wild mammals, we also excluded from the analysis data on trees from which seeds were removed by domestic animals. After excluding these trees, the final number of camera traps placed on trees at the beginning of the fruiting period was 40 for protected and 47 for grazed sites; and at the end of the fruiting period it was 20 for protected and 17 for grazed sites.
List of mammal species that removed Prosopis seeds and number of cameras by which species was observed at least once at grazed and ungrazed sites and during the beginning and end of the fruiting period
Functional groups are: SP (seed predator), UNK (‘unknown’), ED (endozoochorous disperser) and SH (scatter-hoarder).
The total number of seeds removed by each species, pooling across individual visits at every tree (response variable), was considered an estimator of the quantity component of seed dispersal effectiveness (Schupp et al., Reference Schupp, Jordano and Gómez2010). The variables considered in our model were: site (with two levels: ungrazed and grazed), fruiting time of P. flexuosa (with two levels: beginning and end) and functional group [with three levels: seed predator (SP), scatter-hoarder (SH) and endozoochorous disperser (ED)], and interactions between them. The sign of parameters having significant effects was used to interpret the results (McCulloch and Searle, Reference McCulloch and Searle2001; Bolker et al., Reference Bolker, Brooks, Clark, Geange, Poulsen and Stevens2009).
To assess whether seed removal by mammals differed between sites (question 1), between different times during the fruiting period of Prosopis (question 2), and among the functional mammal groups at both sites during each time of the fruiting period (question 3), we fitted a generalized linear model (GLM) with a Poisson error structure (McCullagh and Nelder, Reference McCullagh and Nelder1989). As we detected a higher amount of zeros than expected for a Poisson distribution, we fitted a zero-inflated model. Following the model building and selection guidelines by Zuur et al. (Reference Zuur, Ieno, Walker, Saveliev and Smith2009), we obtained and compared the AIC (Akaike information criterion; Akaike, Reference Akaike1981) values from zero-inflated and zero-altered (hurdle) models. Both kinds of models were built with a Poisson and a negative binomial distribution to correct overdispersion. The model we finally chose was the zero-altered negative binomial model (Zeileis et al., Reference Zeileis, Kleiber and Jackman2008; Zuur et al., Reference Zuur, Ieno, Walker, Saveliev and Smith2009), which was the one with the lowest AIC value. We included a random tree effect in the model in order to account for pseudoreplication among visits of each species to a tree (Zuur et al., Reference Zuur, Ieno, Walker, Saveliev and Smith2009).
All statistical analyses and graphs were performed using R 3.3.1 language and environment (R Development Core Team, 2016). We used the hurdle() function from the ‘pscl’ package for the building of a zero-altered negative binomial model (Zeileis et al., Reference Zeileis, Kleiber and Jackman2008).
Results
During the 280 camera-trap nights (200 for the beginning and 80 for the end of the fruiting period), 65.4% (n = 24,315) of the total seeds offered (n = 37,200) were removed, with mammals being the only seed removers. We recorded ten species removing Prosopis seed, with G. griseoflavus, M. australis and L. griseus being the species recorded (at least once) by the highest number of camera traps (Table 1). Results on domestic mammals are provided for information purposes only. Cows and horses removed seeds from 16 Prosopis trees (Table 1); whereas cows removed seeds during the beginning and end of the fruiting period (mean ± SE number of removed seeds was 26.54 ± 8.4 and 21.75 ± 15.83, respectively), horses were only recorded during the beginning of the period (13.08 ± 7.27).
Regarding our first question, 69.5% of offered seeds were removed from the grazed area and 61% from the ungrazed site. Pooling across all species, there was no difference between sites in total number of seeds removed (Z = –1.88; P = 0.5: Table 2). Considering times, 64.25% of offered seeds were removed during the beginning of the fruiting period and 67% at the end. There was no statistically significant difference in the number of total seeds removed between times (Z = –1.91; P = 0.5; Table 2).
Results of zero-hurdle model of seeds removed per site (grazed and ungrazed), times during the fruiting period of Prosopis (beginning and end), by different functional groups (FG) of terrestrial mammals (SP: seed predator, ED: endozoochorous disperser and SH: scatter-hoarder) and interactions among variables
SE, standard error. Significant differences at P < 0.05 are in bold.
The number of seeds removed by different functional mammal groups varied within sites and between times during the Prosopis fruiting season (Table 2). The total number of seeds removed by seed predators and by the scatter-hoarder was higher than that removed by endozoochorous dispersers (Tables 2 and 3; Fig. 1). Considering groups with the highest number of Prosopis seed removed (seed predators and scatter-hoarder), the model showed site-related differences during the fruiting period of Prosopis. The total number of seeds removed by seed predator species was higher at the end of the period and at the ungrazed site, whereas the total number of seed removed by the scatter-hoarder was higher during the beginning of the fruiting period and at the grazed site (Table 2; Fig. 1).
Boxplots of seeds removed per functional group at grazed and ungrazed sites during (A) the beginning and (B) the end of the fruiting period of Prosopis. Functional groups are: ED (endozoochorous disperser), SH (scatter-hoarder) and SP (seed predator).
Mean number ± SE of seeds removed and number of visits of each mammal species, and mean number ± SE of seeds removed from both sites (grazed and ungrazed) and times (beginning and end) during the fruiting period
Discussion
Regarding our predictions, we expected a higher total seed removal by wild mammals at the ungrazed site and at the end of the fruiting period but, contrary to expectations, total seed removal did not differ between grazed and ungrazed sites or between times. However, we did find differences in the number of seeds removed between the functional mammal groups at the different times studied and at both sites.
By using camera traps, we were able to identify wild mammal species removing fruits, quantify the number of visits, and estimate the number of Prosopis seeds removed by each species. As a result, we found that species recorded by the highest number of camera traps (G. griseoflavus, M. australis and L. griseus; Table 1) had the highest number of feeding visits to the trees and also removed the highest amount of seeds (Table 3). Mammals with an ‘unknown’ contribution to the Prosopis seed dispersal effectiveness were recorded by a lower number of cameras than small and medium-sized mammal species.
The role of small rodents as being responsible for most of the seed removal was previously reported for the Neotropics (e.g. Sánchez-Cordero and Martínez-Gallardo, Reference Sánchez-Cordero and Martínez-Gallardo1998; Iob and Vieira, Reference Iob and Vieira2008). It is also known that, in drylands around the world, invertebrates and small rodents are the main groups removing seeds and acting as seed predators or seed dispersers (Abramsky, Reference Abramsky1983; Linzey and Washok, Reference Linzey and Washok2000; Marone et al., Reference Marone, López de Casenave and Cueto2000; Kelt et al., Reference Kelt, Meserve and Gutiérrez2004). Consistent with this general description, results from our study site showed that the main functional mammal groups removing Prosopis seeds were small rodents: two sigmodontine species (seed predators; G. griseoflavus and A. dolores) and one hystricognath species (scatter-hoarder; M. australis). Seed removal by medium-sized mammals (L. griseus and D. patagonum; Table 1) was lower than that by small rodents.
Endozoochorous dispersers such as L. griseus and D. patagonum play an active role in dispersing Prosopis seeds (Campos and Ojeda, Reference Campos and Ojeda1997; Campos et al., Reference Campos, Peco, Campos, Malo, Giannoni and Suárez2008); however, none of these species feed exclusively on Prosopis fruits. Also, both species are associated with open habitats (Jiménez et al., Reference Jiménez, Lucherini and Novaro2008; Rodríguez, Reference Rodríguez2009). Following their habitat requirements, and despite seed removal by this functional group not being statistically different between sites in our study, we found a tendency towards a higher number of trees visited (Table 1) and a higher number of seed removed (Table 2) by these medium-sized mammals, particularly by L. griseus, at the grazed site compared with the ungrazed site. The tendency found here indicates a high seed removal activity of these native species at a site where they could be more exposed to risks from human activities, thereby increasing their vulnerability to being hunted (Campos et al., Reference Campos, Tognelli and Ojeda2001; Tabeni et al., Reference Tabeni, Spirito and Ojeda2013).
During the beginning of the fruiting period of Prosopis, the seed predators (G. griseoflavus and A. dolores) at the ungrazed site and the scatter-hoarder (M. australis) at the grazed site were the main functional groups removing seeds. Following our predictions, these results could be related to the habitat requirements of the species. From previous studies we know that grazing by large herbivores causes habitat modifications such as alteration of biomass and vegetation structure with consequent effects on small-sized mammal populations (Jones, Reference Jones2000; Flowerdew and Ellwood, Reference Flowerdew and Ellwood2001; Foster et al., Reference Foster, Barton, Lindenmayer and Toit2014). In the central Monte, the main seed predator G. griseoflavus avoids open habitats and is associated with patches of complex vegetation structure on grazed and ungrazed sites (Tabeni et al., Reference Tabeni, Mastrantonio and Ojeda2007). Akodon dolores, which only removed a low number of seeds from the ungrazed site during the end of the fruiting period (Tables 1 and 3), prefers habitats with high plant density inside the ungrazed area (Tabeni and Ojeda, Reference Tabeni and Ojeda2005; Corbalán, Reference Corbalán2006). In contrast, M. australis, which removed a higher number of seeds from the grazed site, mainly occupies habitats under grazing where vegetation structure allows it to build its colonies under plants with an umbrella-like pattern (Tognelli et al., Reference Tognelli, Campos, Ojeda and Roig1995).
Despite the different number of trees included in our experiment at both fruiting times considered, we found strong tendencies of differential seed removal by mammals on sites under two land management practices. During the end of the period, the seed predator group removed the highest number of seeds. This could be related to the seasonal increase in the population abundance of G. griseoflavus occurring in autumn (Ojeda, Reference Ojeda1989; Corbalán and Ojeda, Reference Corbalán and Ojeda2004). As an advantage from the seed point of view, the time when the peak fall of Prosopis fruits occurs (summer) does not coincide with the time at which the population of their main seed predator is naturally more abundant (autumn).
Our results represent a season of low Prosopis fruit production, considering the high inter-annual variability in crop size of this tree species (Dalmasso and Anconetani, Reference Dalmasso and Anconetani1993). However, the dynamics of seed removal by mammals found here may vary in a year of higher fruit availability; in that case, mammals may or may not become satiated by large fruit production events (Kelly and Sork, Reference Kelly and Sork2002). Also, a previous study in the area found Prosopis seed production to be higher at the protected than grazed sites (Aschero and Vázquez, Reference Aschero and Vázquez2009). Further research in a year of high Prosopis fruit production could explain the seed removal response by mammals in an ecological context of different fruit availability.
In summary, seed predator species removed more seeds compared with other species at the ungrazed site and during the end of the fruiting period, whereas the scatter-hoarding species removed more seeds from the grazed site at the beginning of the period. On the one hand, the nearly 50-year exclusion of large herbivores from the protected area seems to have generated more suitable microhabitats for seed predators and to have caused the expulsion of medium-sized mammals (Tabeni et al., Reference Tabeni, Spirito and Ojeda2013). On the other hand, grazing management could be advantageous for seed removal by a hoarding species and by medium-sized mammals that disperse seeds through endozoochory. Nevertheless, a lower percentage of Prosopis seedling survival was reported for the grazed area compared with the ungrazed site, probably due to intensive cattle trampling (Campos et al., Reference Campos, Campos, Mongeaud, Borghi, de los Ríos and Giannoni2011; Aschero et al., Reference Aschero, Morris, Vázquez, Alvarez and Villagra2016). As a first approach to addressing seed removal by mammals using camera traps, and from a plant perspective, a tentative suggestion would be to maintain, in the protected area, a heterogeneity of habitats that ensures that all local wildlife species play their functional roles (Campos et al., Reference Campos, Campos, Miguel and Cona2016), mainly enlarging the habitat associated with seed disperser species.
Acknowledgements
The authors thank all the staff of Ñacuñán Reserve for their cooperation; and the owner and the family in charge for allowing us to work in the cattle ranching area. We also thank three reviewers for comments on an earlier version of the manuscript. Nélida Horak assisted us in drafting the English version.
Financial support
This work was supported by Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET-PIP 112-201101-00601), Agencia Nacional de Promoción Científica y Tecnológica (PICT-2013-0478) and a graduate fellowship from CONICET to F.M.
Conflicts of interest
None.
