Feeding of small Neotropical felids ( Felidae : Carnivora ) and trophic niche overlap in anthropized mosaic landscape of South Brazil

Understanding the diversity of a community and its dynamics is especially important in altered habitats such as agricultural ields, where this information can help biodiversity conservation programs. In an altered landscape of the interior Atlantic Forest, western Paraná State, Brazil (25o41’ to 25o20’S and 53o56’ to 54o35’W), samples (310) were collected and a total of 110 samples could be determined for some small Neotropical felids, including 39 Leopardus guttulus (oncilla), 38 L. wiedii (margay) and 33 Puma yagouaroundi (jaguarondi). The diets of these felids contained typical synanthropic prey such as Mus musculus in 44% (L. guttulus), 32% (L. wiedii) and 15% (P. yagouaroundi) of the total samples. This observation and the sample collection in agricultural places demonstrate that felids can use this anthropized landscape. The small mammals (<100 g) were more common in the diet of these species. Nevertheless, the correction factor was shown to have signiicant eiciency in correcting estimates of biomass ingested for two of the three species of felids, and we therefore recommend that these species be used in future studies. Even with this observed coexistence, the food niche of the three species showed a large overlap.


Introduction
Knowing how a species uses and shares resources is essential for understanding the mechanisms that allow the coexistence of species (e.g., guild of small Neotropical felids), since diferences in resource use may be a key factor in a competitive scenery (SCHOENER 1974) and may determine the diversity in a community (PULLIAM, 2000;DAVIES et al., 2007;DI BITETTI et al., 2010).
Competition for resources may be expected in guilds if morphological and behavioral similarity exists between species, when competitive exclusion of subordinate species may be a consequence (DONADIO; BUSKIRK, 2006;MEACHEN-SAMUELS;VALKENBURGH, 2009;MORIN, 2011).However, resource partition or displacement character can reduce competition and facilitate coexistence among species (MACARTHUR; LEVINS, 1964;MORIN, 2011).Among the resource partition factors, there can be a partition in horizontal (type) and vertical (forest stratum) habitat, time and diet (AUGUST, 1983;DAYAN;SIMBERLOFF, 1998;DI BITETTI et al., 2010).In the food dimension, the use of abundant resources or segregation in prey size can minimize competition, allowing the coexistence of species (ROSENZWEIG, 1966;MOTTA-JUNIOR, 2006).The feeding habit permeates every aspect of the life history of an animal, where it is essential to understand the ecology of a species (CRAWSHAW; QUIGLEY, 2002).
Two of these species (L.wiedii and P. yagouaroundi) have a wide geographical distribution, from southern Mexico to southern Brazil (NOWELL; JACKSON, 1996).Leopardus guttulus has a smaller distribution area, which includes only the southern distribution area of the old denomination L. tigrinus (NASCIMENTO, 2010;TRIGO et al., 2013).This distribution area includes Southeast, South and West-Central regions of Brazil, and northeastern Argentina and Paraguay (NASCIMENTO, 2010).Although these three small Neotropical felids are considered threatened, little ecology information is available about them in situ (CALLEIA et al., 2009;DI BITETTI et al., 2010;IUCN, 2014).For example, only four dietary studies are available for L. wiedii (KONECNY, 1989;WANG, 2002;ROCHA-MENDES et al., 2010;BIANCHI et al., 2011), and these do not include data in agricultural landscapes and in sympatry Feeding of small Neotropical felids with two other species, L. guttulus and P. yagouaroundi.For the latter two species, there is also little information about their feeding in a sympatry context (TÓFOLI et al., 2009;SILVA-PEREIRA et al., 2011).
These three small Neotropical felids occur in disturbed places (DI BITETTI et al., 2010;KOSYDAR et al., 2014) and their known feeding repertoire includes insects, amphibians, reptiles, birds and mammals, mostly small species (KONECNY, 1989;WANG, 2002;TÓFOLI et al.. 2009;ROCHA-MENDES et al., 2010;BIANCHI et al., 2011;SILVA-PEREIRA et al., 2011).On the other hand, medium-sized mammals are also reported in the feeding of these small Neotropical felids (WANG, 2002;TÓFOLI et al., 2009).Thus, to compensate for a possible overestimation of the biomass of medium-sized species and a consequent undervaluation of small-sized species, it is important to use correction factors of biomass consumed (OLIVEIRA, 2002).A correction factor is not available for the Neotropical felids, except for P. concolor (ACKERMAN et al., 1984), and a bobcat correction factor has been used for smaller species such as L. pardalis (VILLA MEZA et al., 2002).Since L. guttulus, L. wiedii and P. yagouaroundi are known in the interior Atlantic Forest, a region that has undergone great disturbance in natural habitat, and since the information on feeding habitat for these small wildcat species is anecdotal, the objectives of this study were: (1) to provide information on the diet breadth and food niche overlap for these three species when in sympatry; and (2) to develop and test correction factors for the biomass consumed by L. guttulus and L. wiedii.

Materials and Methods
The extreme west of Paraná State, in southern Brazil (25º41' to 25º20'S and 53º56' to 54º35'W), is located between 120 and 540 m a.s.l.(SANTOS et al., 2006), and has an annual rainfall between 1,500 and 2,200 mm and mean temperature between 16 and 22 ºC (DE ANGELO et al., 2011).The study locality belongs to the ecoregion of the Upper Paraná Atlantic Forest or interior Atlantic Forest (DE ANGELO et al., 2011), which up to the middle of the last century consisted of submontane semi-deciduous forest and alluvial forest (SALAMUNI et al., 1999).After the human occupation that occurred in the 1950s and the extensive conversion of forest to agricultural areas, only 6% of original forest remains in formations of the Upper Paraná Atlantic Forest (GIRAUDO et al., 2005).However, these remaining forests are fragmented and form, along with the agricultural matrix, a mosaic landscape where annual crops such as soybean (Glycine max) and corn (Zea mays) predominate.In this region, the National Park of Iguaçu (Brazil) is a major conservation unit and connects to the National Park of Iguazú (Argentina) to form a continuous native forest of over 2,500 km 2 (DI BITETTI et al., 2003).
Carnivore feces were collected between 2007 and 2009, covering foot edges, trails, paths and roads in the small forest fragments, in the agricultural matrix and in the interior of Iguaçu National Park (ParNa Iguaçu).The areas of this study (Figure 1) were selected due proximity and easy accessibility and are presented below.In addition, samples sporadically collected next to ParNa Iguaçu and gastrointestinal samples of roadkill of small Neotropical felids were also included in this analysis.
A total of 310 samples were collected.Of these, 298 were fecal samples and twelve were gastrointestinal samples directly removed from dead felids found on the road.All these samples were collected under authorizations 025-2007, 18,347 and 12,200 After being collected, the samples were stored in plastic bags, identiied and further processed by washing in 0.02 mm iltering mesh cones and drying at 60 ºC for 24 h.During sample screening (fecal or gastrointestinal samples), we separated and stored the contents that allowed the identiication of animal species by assessing the microstructure of the guard hairs.The current identiication method used for these three species follows three procedures: identiication of the medullary pattern, cuticular pattern, and width to length ratio of the cuticular scales of the hair shaft (QUADROS; MONTEIRO-FILHO, 2010).The medullary pattern, called trabecular with fringed margins, is similar in all small Brazilian Neotropical felids (VANSTREELS et al., 2010).The cuticular character, however, divides Feeding of small Neotropical felids them into two pairs of species: the losangic pattern clusters L. guttulus with P. yagouaroundi, and L. wiedii shows the foliaceous pattern (QUADROS; MONTEIRO-FILHO, 2010).Each small Neotropical Brazilian felid can be identiied using the width to length ratio of the scales (Figure 2) on the shaft of the guard hairs (QUADROS; MONTEIRO-FILHO, 2006;2010).The hairs of predators and prey were identiied using reference data (QUADROS, 2002; subsequently published as QUADROS; MONTEIRO-FILHO, 2010;and MIRANDA et al., 2014).Samples were identiied to the species level using the guard hairs, which enter the gastrointestinal system by ingestion during selfcleaning behavior (ECKSTEIN;HARTS, 2000).Other items (prey hairs, teeth, exoskeleton, feathers, seeds) were identiied by comparative analysis (see table results for details) with museum pieces (Collection of Zoology Department, Mammology Section at Paraná Federal University (DZUP-CCMZ-UFPR)) or pieces obtained in the study area (e.g., Zea mays, Triticum aestivum and Glycine max).
The occurrence frequency of food items were calculated by dividing the number of samples in which a given item was found by the total number of samples (resulting in a percentage), and the percentage of occurrence of each item was calculated by dividing the number of samples in which a given item was found by the total number of all items (resulting in a percentage).Frequency of occurrence indicates whether an item is frequent in the diet (KONECNY, 1989;MARTINS et al., 2008), whereas percentage of occurrence indicates the relative importance between items (MAEHR; BRADY, 1986;MARTINS et al., 2008).
The niche breadths of the species were determined using the standardized Levins index and Smith´s measure (KREBS, 1999).The standardized Levins (B sta ) is widely used and was calculated as a comparative measurement with other works.This index is expressed on a scale of zero to one, where values close to zero represent a narrow niche and those close to one a broad niche.Smith's measures (FT) were calculated to measure the conidence interval and to determine if niche breadth difers between felid species.This index varies like the standardized Levins with zero representing a narrow niche and one a broad niche.The Pianka index (KREBS, 1999) was used to assess the degree of overlap between feeding niches, which is also expressed on a scale from zero to one, where values close to zero represent little  niches were recalculated 1000 times.We did not consider that the observed overlap was equal to the expected overlap when the occurrence was equal or less than 5% of randomized values (α ≤ 0.05).
To calculate the consumed biomass, we used the information on the live weights of prey (PAGLIA et al., 2012).The estimates of biomass consumed by predators were calculated as the product of the weight of the prey and the frequency of occurrence observed in stool samples.These values were corrected by a linear factor (ACKERMAN et al., 1984).For this, whole carcasses of rat (Rattus novergicus) and quail (Coturnix sp.) were ofered to adult captive L. guttulus (n = 21) and L. wiedii (n = 14) maintained at the ITAIPU Binacional Wildlife Breeding Center.Carcasses were ofered to all specimens of L. guttulus and L. wiedii simultaneously at 6:00 p.m. for eight days, with four consecutive days for quail and for rat.On average, we ofered 384 ± 42 (SD) g of quail and 355 ± 50 g of rat to L. guttulus, and 412 ± 86 g of quail and 385 ± 32 g of rat to L. wiedii.On the next morning of each day, the remnants of ofered carcass were measured for each individual and the weight consumed calculated.All ofered and consumed biomass values were used to calculate a linear regression for L. guttulus and for L. wiedii, and to obtain a linear correction factor.
A total of 168 measurements for L. guttulus and 112 for L. wiedii were utilized in the linear regression.Normality of the data was tested prior to regression analysis.The linear equation obtained ( X b a y .ˆ− = ) was used as a correction factor for the biomass consumed, in which y ˆis the value of consumed biomass corrected, a is the constant representing the intercept of the line with the vertical axis, b is the constant representing the slope of the line and X is the observed biomass or the live weight of the prey.Diference hypothesis between uncorrected (expected) and corrected biomass (observed) was tested with a Wilcoxon signed-rank test (T) considering α = 0.05.This test was used because number of prey for all felids was less than 20 and observed and expected data were considered paired.Due to the small number of specimens of P. yagouaroundi in captivity, the correction of biomass for this felid was calculated with the correction factor of L. wiedii, because this felid has a similar body weight as P. yagouaroundi (NOWELL; JACKSON, 1996).Regarding the correction factor of consumed biomass, the linear regression between the values of biomass supplied and consumed provided the following linear equations: L. guttulus y ˆ = 197.9+ 0.19 * X, and r 2 = 0.006; L. wiedii y ˆ = 154.6 + 0.35 * X, and r 2 = 0.03.According to the behavior of biomass correction by the linear equations obtained, biomass consumption values were corrected beginning at 245 g for L. guttulus and 238 g for L. wiedii.
To evaluate the importance of prey size, we adapted the method developed by Emmons (1987) that organizes prey size into biomass classes and calculates the percentage occurrence of biomass prey considering the total sum of prey biomass consumed and its frequency.
Here, we considered the followings biomass classes: weight less than 100 g, between 100 and 1000 g and more than 1000 g.

Results
Of the 310 samples, 157 (51%) were identiied as being from wild felids, most of which (110 or 35%) were one of the three species of small Neotropical felids under study.Of these, 98 were identiied from stool samples and 12 from gastrointestinal contents of felids that were dead on the roads.Only one sample was found in the stomach of L. wiedii and contained two specimens of Didelphis sp.Other hair samples were collected in the distal intestine.
All dead felids were found at the intersection between river and road.Among the other identiied samples, six were from L. pardalis (ocelot), 33 from P. concolor (puma) and eight from Panthera onca (jaguar), all of which were collected inside Iguaçu National Park.

Feeding of small Neotropical felids
The frequency of occurrence of all prey species in their respective live-weight ranges revealed that small prey were the most important for the three carnivores (less than 100 g).On the other hand, when the biomass consumed was analyzed, prey between 100 and 1000 g were more important to L. guttulus and P. yagouaroundi, and prey greater than 1000 g for L. wiedii (Tables 1, 2, 3).Prey heavier than 1000 g were not detected in the diet of L. guttulus.In these samples, insect exoskeletons (Blattaria, Hymenoptera, Lepidoptera and Orthoptera) and grasses (Poaceae) were less important in biomass.However, these were found respectively in 29 (26%) and 19 (17%) samples for all three species.The Akondontini rodents and Monodelphis marsupials were the most important food items for the three small Neotropical felids.
1 Prey identiication method: H = microscopic pattern of hair, T = tooth or bone structure, SD = Scale and/or ectoderm, F = Feather, S = grass, E = exoskeleton portion, SE = seed; 2 Total identiied items; 3 Number of records for this prey species ÷ total records for all prey species; 4 Number of records for this prey species ÷ total number of fecal samples; 5 Approximate weight of adult prey; 6 Corrected biomass; 7 Total consumed biomass; 8 Frequency of occurrence of biomass = values of BC (kg) x100 ÷ total sum of BC (kg).
Feeding of small Neotropical felids 1 Prey identiication method: H = microscopic pattern of hair, T = tooth or bone structure, SD = scale and/or ectoderm, F = feather, S = grass, C = exoskeleton portion; 2 Total identiied items; 3 Number of records for this prey species ÷ total records for all prey species; 4 Number of records for this prey species ÷ total number of fecal samples; 5 Approximate weight of adult prey; 6 Corrected biomass; 7 Total consumed biomass; 8 Frequency of occurrence of biomass = values of BC (kg) x100 ÷ total sum of BC (kg).
The trophic niche breadth of the three species was similar (see Smith's measure in Table 4).However, L. guttulus was the broadest, followed by the niches of L. wiedii and P. yagouaroundi.All three species had high levels of trophic niche overlap (average ≈ 90%).For L. guttulus and L. wiedii, this overlap was larger than that for L. guttulus and P. yagouaroundi or P. yagouaroundi and L. wiedii (Table 4).All observed niche values were higher than expected (α<0.05).

Discussion
The success in identification of fecal samples for wild felids (51%) can be considered a satisfactory result since results above 40% are considered eicient (ROCHA-MENDES et al., 2010).Even with this good result, this technique can provide better results if considered as part of a tool set that includes footprints and biomolecular techniques (VANSTREELS et al., 2010).
Few identiied samples of small Neotropical felids were collected inside Iguaçu National Park (n = 27).In this region, a camtrap study detected more L. guttulus and L. wiedii in the least-protected sites where the big felids were less associated (DI BITETTI et al., 2010).
This pattern of small Neotropical felids in degraded areas was related to released ocelot and puma populations, which are possibly caused by a decline in jaguar populations (see also MORENO et al., 2006;PAVIOLO et al., 2008) and also explains that observed in this study.The other samples (n = 83) were collected in the bufer zone of Iguaçu National Park (10 km of the edge this conservation unit).These samples were collected at the edge of small forest fragments or the forests that surround small rivers.In both cases the sampling sites were transition areas between forest fragment and corn or soybean ields.
The noted predominance of mammals lighter than 1000 g in the diets of L. guttulus, L. wiedii and P. yagouaroundi (KONECNY, 1989;FACURE;GIARETTA, 1996;WANG, 2002;TÓFOLI et al., 2009;ROCHA-MENDES et al., 2010;SILVA-PEREIRA et al., 2011) may be related to the fact that these are the most abundant mammals in the Neotropical region (ROBISON;REDFORD, 1986;SOLARI;RODRIGUES, 1997;ABREU et al., 2008), and tend to increase in abundance in regions of altered forest habitat (LAURANCE et al., 2002).This is especially supported by the low occurrence of rodents in the diet of small Neotropical felids in landscapes where they are less abundant, with the highest occurrence of potentially more abundant items in the landscape, such as lizards (OLMOS, 1993).Furthermore, medium-and large-sized species (>1000 g) are less abundant and less diverse in altered landscapes (CHIARELLO, 1999).Thus, the predominance of small mammals in the diets of small Neotropical felids points to their possible opportunistic character, since they do not seem to maximize their energy input by selecting prey by size and consuming small items, which are potentially more abundant in the altered landscape (SILVA-PEREIRA et al., 2011).Sharing abundant resources is one of the factors that may explain the broad food niche overlap between these three species and their coexistence (see also WANG, 2002;MOTTA-JUNIOR, 2006;SILVA-PEREIRA et al., 2011;DI BITETTI et al., 2010).This is corroborated by the data of abundance of non-volant small mammals in the interior Atlantic Forest, where Akodontini is more abundant group (DE LA SANCHA, 2014).On the other hand, the consumption of large prey (>1000 g), for example, Dasyprocta azarae and Sapajus nigritus, by L. wiedii and P. yagouaroundi (WANG, 2002;TÓFOLI et al., 2009;ROCHA-MENDES et al., 2010;SILVA-PEREIRA et al., 2011), but not observed for L. guttulus (SILVA-PEREIRA et al., 2011), is other factor that may contribute to reducing the competition between these species (ROSENZWEIG, 1966).
These observations of large prey consuming small Neotropical felids are related to the use of carcasses (WANG, 2002;TÓFOLI et al., 2009).However, there is no evidence of such behavior in small Neotropical felids (SILVA-PEREIRA et al., 2011).The fact that these animals consume prey heavier than 1000 g highlights Feeding of small Neotropical felids the importance of using correction factors of biomass in studies of the feeding ecology of small Neotropical felids, since their capacity to consume biomass is less than the live weight of the species included in their food.The diference between the observed and corrected biomass was signiicant only for L. wiedii and P. yagouaroundi, illustrating that estimates are not corrected and are overestimated.Because of the correlation between the densities of the prey biomass available in the landscape and population density of carnivores (CARBONE; GITTLEMAN, 2002;HETHERINGTON;GORMAN, 2007), the accuracy of such information is important for the proper planning of the conservation of this group of species in situ.
Furthermore, it is important to note that the partition of circadian activity of these three species may be another niche dimension that facilitates their coexistence.Diurnal-crepuscular prey, e.g., Sapajus nigritus, Galicts cuja and Dasyprocta azarae (CULLEN et al., 2000;DI BITETTI et al., 2000;YENSEN;RARIFA, 2003), were consumed only by L. wiedii and P. yagouaroundi and not L. guttulus.In a circadian rhythm study conducted in this same area (DI BITETTI et al., 2010), the circadian activity of L. wiedii and P. yagouaroundi was nocturnal with small percentage at twilight.
It is also important to highlight the consumption of cursorial and arboreal prey by the L. wiedii, a felid considered adapted to arboreal life (OLIVEIRA, 1998).This information indicates that this species makes use of food resources on both forest strata (PASSAMANI, 1995;AZEVEDO, 1996;SOLORZÁNO-FILHO, 2006;CALLEIA et al., 2009).Another work has indicated that the margay moves on the ground (KONECNY, 1989), and eats mostly arboreal rodents (FACURE;GIARETTA, 1996), but this is based on few observations.On the other hand, reports on L. wiedii using the upper stratum of forests and consuming arboreal prey or birds are frequent (XIMENEZ, 1982;MONDOLFI, 1986;PASSAMANI, 1995;AZEVEDO, 1996;OLIVEIRA, 1998;SOLORZÁNO-FILHO, 2006;BIANCHI et al., 2011).In this fragmented place, arboreal species were found in stool samples of the three small Neotropical felids collected in the agricultural matrix, e.g., Caluromys lanatus, Sapajus nigritus, Marmosa paraguayanus (DI BITETTI et al., 2000;GRELLE, 2003).This consumption of arboreal prey also suggests that the maintenance of forest fragments is an important factor for these small Neotropical felids in this landscape with agricultural use.
Finally these results emphasize that these three felids use small prey as important food items, such as Akodontini rodents and Mus musculus.Nevertheless, we found that L. wiedii and P. yagouaroundi consumed medium-sized mammals.For this, the correction factor improved eiciency in estimates of consumed biomass, and it is important to use it in future studies.For this reason, we point out the need for correction factors for biomass in L. pardalis and P. yagouaroundi, which must be determined and tested in the future.The results of their broad feeding niche overlap indicate that the coexistence of these small wildcats may be facilitated by the consumption of abundant prey such as Akondontini rodents, beyond subtle stratiication in circadian activity, verticalization of habitat and body size of these felids.
, provided by IBAMA (Environment Institute of Brazil) and ICMBio (Biodiversity Conservation Institute of Brazil), and 224/10 IAP, provided by IAP (Environment Institute of Paraná State -Brazil).

FIGURE 1 :
FIGURE 1: Location of sampling areas in western Paraná State, southern Brazil.(A) The study (black) area in South America, and; (B) in the triple frontier: Argentina, Brazil and Paraguay.(C) Forest remnants and watercourses in study area.
overlap and those close to one extensive overlap.The prey records observed were randomized, and the overlap

is connected to the secondary riparian forest of the Represo River and together covers 70 ha. On this farm, samples were collected on secondary roads in agricultural places, trails in the interior of secondary forests and edges of
these forest fragments.

TABLE 1 :
Relative frequency and biomass of prey consumed by oncilla (Leopardus guttulus), collected in Southwest Paraná State, Atlantic Forest domain, Brazil.
1 Identiication method: H = microscopic pattern of hair, T = tooth or bone structure, SD = scale and/or ectoderm, F = feather, S = grass, C = comparison of exoskeleton portion; 2 Total identiied items; 3 Number of records for this prey species ÷ total records for all prey species; 4 Number of records for this prey species ÷ total number of fecal samples; 5 Approximate weight of adult prey; 6 Corrected biomass; 7 Total consumed biomass; 8 Frequency of occurrence of biomass = values of BC (kg) x100 ÷ total sum of BC (kg).

TABLE 2 :
Relative frequency and biomass of prey consumed by margay (Leopardus wiedii), collected in Southwest Paraná State, Atlantic Forest domain, Brazil.

TABLE 3 :
Relative frequency and biomass of prey consumed by jaguarondi (Puma yagouaroundi), collected in Southwest Paraná State, Atlantic Forest domain, Brazil.