The Scorpion Mud Turtle (Kinosternon scorpioides Linnaeus 1766) is the most widely distributed chelonian species in the Neotropical region, ranging from northeastern México to northwestern Argentina (Rueda-Almonacid et al. 2007; Berry et al. 2012; Tomas et al. 2015; Cáceres-Martínez et al. 2017; Rhodin et al. 2021). Three subspecies are currently recognized (Iverson et al. 2013): K. s. scorpioides, K. s. cruentatum, and K. s. albogulare. The first occupies most of the species' range, from eastern Panamá to northern Argentina (Vanzolini et al. 1980; Cabrera and Colantonio 1997; Cabrera 1998; Berry and Iverson 2011; Tomas et al. 2015) including the Argentinian provinces of Salta, Jujuy, Tucumán, Formosa, and Chaco (Cei 1993; Cabrera 1998; Acosta et al. 2013; Nigro et al. 2013).

Despite its wide distribution across a diversity of biomes, Scorpion Mud Turtle populations seem to inhabit similar wetlands. The species was described as “bottom walking” in permanent, semipermanent, and temporary water bodies (Berry and Iverson 2011; Berry et al. 2012). With the exception of some authors (Pritchard and Trebbau 1984; Moll 1990; Barrio-Amorós and Narbaiza 2008) who have described the species as carnivorous, it is mostly considered a generalist, opportunistic omnivore (Vanzolini et al. 1980; Rueda-Almonacid et al. 2007; Carvalho et al. 2008; Berry and Iverson 2011; Berry et al. 2012; Montes-Correa et al. 2017; Pereira et al. 2018). Only 2 published articles provide solid field data on the diet of K. scorpioides (Moll 1990; Carvalho et al. 2008), but only that of Carvalho et al. (2008) corresponds to K. s. scorpioides. Other information on the diet of the subspecies is anecdotal (Fiasson 1945; Cunha 1970; Vanzolini et al. 1980; Hernández-Ruz et al. 2016) or based on captive individuals (Pritchard and Trebbau 1984; Monge-Nájera and Moreva-Brenes 1987). The feeding habits of the southernmost populations were previously unknown.

Two decades ago, the conservation status of the species was assessed as insufficiently known in Argentina (Richard and Waller 2000) and this remains unchanged until today (Prado et al. 2012). With the exception of few local reports (Richard 1990a, 1990b; Acosta et al. 2013; Nigro et al. 2013), in the 85 yrs since the first report of K. s. scorpioides in Argentina (Freiberg 1936), no specific research has been performed on the species in the country.

Our contribution constitutes a first approach to the knowledge of the ecology of Kinosternon s. scorpioides in the Dry Chaco of Argentina, providing information on 1) the environmental characteristics of field capture locations, and 2) the diet of the subspecies based on stomach flushing and fecal samples.

METHODS

Study Area. — An extensive sampling effort was conducted in several localities of the Chaco ecoregion (sensu Morello et al. 2012) as part of a population study of Acanthochelys pallidipectoris (Freiberg 1945). Many localities were sampled along the Dry Chaco portion of eastern Salta, western Formosa, and western Chaco provinces (101 localities sampled) and in the Humid Chaco portion of northern Santa Fe, eastern Formosa, and eastern Chaco provinces (106 localities sampled). We made 6 field trips during the austral spring and summer months (October 2016, September 2017, March 2018, January and March 2019, March 2020), spending a total of 70 d in the field. Turtles were actively searched in several kinds of natural and artificial water bodies such as ditches along main, secondary, and unpaved roads, ponds within and bordering the Chaco forest, peridomiciliary lagoons, and urban-related water reservoirs.

Turtle Catching and Sample Collection. — Turtles were captured either by muddling (Bury et al. 2012) or systematically passing a trawl net through the water. Each captured turtle was sexed, weighed (using a digital balance ± 1 g), maximum straight carapace length measured (using calipers ± 0.1 mm), photographed, global positioning system–referenced (with a Garmin^® GPSMAP 64s), and marked (Cagle 1939) before its release at the site of capture. Turtles were stomach-flushed in situ following the method of Legler (1977), and fecal samples were collected only on those occasions when turtles liberated feces during the flushing process. The array of aquatic prey items in the environment was sampled by passing a hand net (50-cm diameter, 0.1-mm mesh) through the vegetated margins and bottoms of each sample site 30 times. All samples were stored in 70% ethanol. Photographic vouchers of the species were deposited in the Herpetological Collection of the Universidad Nacional del Nordeste, Corrientes Province, Argentina (UNNEC-A 00024-00031).

Environmental Variables and Description of Capture Sites. — Each site in which we trapped turtles was characterized as follows: 1) perimeter and mean depth of water body were measured and 2) physicochemical variables were measured with a real-time data logger (Lutron^® WA-2015), including pH, conductivity (µS · cm^–1), dissolved oxygen (mg · L^–1), and air and water temperature (°C). Water temperature was taken at the middle of the water column and air temperature was taken 1 m above the ground near the edge of the pond. The time of day and the exact point in the water column varied according to the time each turtle was trapped and the maximum depth of each body of water, respectively.

As a complement, a larger-scale characterization of the capture sites was carried out using the software QGis 3.16.4 Hannover (Quantum GIS Development Team 2021). For this purpose, we used the 2019 annual 30-m-resolution land use and land cover map (MapBiomas Chaco Project 2021) and we also considered a buffer area of 3 km around the capture sites. This area is slightly larger than the maximum displacement of 2.5 km reported for the species by Bedoya-Cañón et al. (2018). Zonal histograms (number of pixels for each cover category) were used as an exploratory approach to identify the type and proportions of land cover and uses within the buffer areas. Once identified, 4 categories were defined: 1) woody vegetation areas (map codes 3, 4, 6, 45), 2) herbaceous vegetation areas (map codes 43, 44), 3) agricultural and livestock areas (map codes 15, 19, 36), and 4) nonvegetation land areas (map code 22). A detailed description of the legend with its corresponding code is available for download on MapBiomas website (MapBiomas Chaco Project 2021). We generated binary masks from each category (i.e., with a value of 1 for the interest coverage and 0 for the rest), and zonal analysis functions were applied to extract descriptive statistics of each buffer area (count, sum, mean). All layers were reprojected to the WGS84/UTM zone 20S coordinate reference system.

Diet Analysis. — Every food item was identified at the lowest possible taxonomic level (for the purpose of the analysis usually at order or family levels), counted, and measured in volume. For volume measurements of animal items, we employed 2 methods depending on the prey size: the water displacement technique based on the Archimedes principle, or the ellipsoid method of Dunham (1983). The volume of plant material was measured using a millimeter capsule in which each cell corresponds to 1 mm³ (Hyslop 1980).

The index of relative importance (IRI) (Pinkas 1971), IRI = %OF × (%V + %N), was used to determine the contribution of each item to the diet, where %OF, %N, and %V represent the percentage values of occurrence frequency, number frequency, and volume of each food item category, respectively. The highest value of IRI was used to rank the remaining IRI values according to 4 prey-item categories: accidental (0%–25%), accessory (25.1%–50%), secondary (50.1%–75%), and fundamental (75.1%–100%). The IRI was calculated for items grouped at the family level, but in a few cases of difficult identification they were grouped at the order level, and in certain groups (e.g., aquatic coleopterans and anurans) larval and adult stages were considered as different items.

The selection of prey resources was evaluated using the preference index (W_i = U_i/D_i; Savage 1931; Atienza 1994). The variable U_i represents the quotient u_i/u₊ and the variable D_i is the quotient d_i/d₊, where u_i is the observed number of units used of resource i, and u₊ the total number of resources used. Similarly, d_i is the number of units available in the environment of resource i, and d₊ the total resource availability. The index ranges from 0 (maximum negative selection) to infinity (maximum positive selection), 1 being the value corresponding to no selection or random food intake. This index was calculated for aquatic items grouped at least at order level (or higher in some categories of difficult identification such as Hirudinea, Crustacea) and discriminating between larva and adult stages where appropriate (terrestrial items were not sampled; therefore, the calculation was impossible). Diet and available items with numerical frequency lower than 1 were not taken into consideration for the index calculation.

The Shannon diversity index (Shannon 1948), H′ = –Σ (P_i × ln P_i) (where P_i = n_i/N), was calculated to compare the diversity of aquatic prey items (of animal origin) between the diet of turtles (H′d) and their environmental availability (H′a). This index was also employed to compare the environmental availability between provinces, and the diet diversity among populations of the species (in this case the vegetal fraction was included).

The feeding habit of the species was classified using the volumetric percentage of plant material in the diet as proposed by Espinoza et al. (2004) with modifications: carnivore = 0%–30%, omnivore = 30.1%–70%, and herbivore = 70.1%–100%.

To evaluate the feeding strategy, niche width contribution and prey importance within and between populations, we employed the graphic method proposed by Amundsen et al. (1996) which is in turn a modification of the method suggested by Costello (1990). This method graphs the prey-specific abundance (P_i; y axis; takes values between 0 and 100) against the occurrence frequency (OF; x axis; takes values between 0 and 1). P_i is the percentage a given prey comprises among all prey items for those predators where this prey actually occurs: P_i = (ΣS_i/ ΣS_ti) × 100, where S_i is the number of i prey comprised in the stomach content, and S_ti is the total stomach content in only those predators which presented i prey in their stomach.

As a complement to the graphic method, we also calculated the amplitude of the trophic niche according to the standardized index of Levins (Colwell and Futuyma 1971): Bst = (B – 1)/(n – 1), where n is the number of prey categories, and B = 1/(ΣP_i²), P_i being the numeric proportion of prey categories (P_i = n_i/N). This index takes values from 0 (narrow niche) to 1 (broad niche).

Statistical Analyses. — We used a t-test with a p = 0.05 to determine whether there were significant differences in straight-line carapace length and weight between sexes in adult specimens, and also to assess the significance between the H′ index values. The significance level of the preference index was obtained after comparing a statistic based on W values (S = [W_i – 1]²/se[W_i]²; se[W_i] = =[1 – D_i]/[u₊D_i]) with a χ² test of one degree of freedom (Manly et al. 1993). The Bonferroni correction was previously applied. All analyses were carried out using the R program, version 4.1.1 (R Core Team 2021).

RESULTS

Captures. — The species was actively searched for at 101 and 106 localities from the Dry and Humid Chaco, respectively (Fig. 1), and was found at 9 Dry Chaco localities from Anta and General San Martin departments (8 turtles from the province of Salta in 2018), Bermejo department (7 turtles from the province of Formosa in 2019), and General Güemes department (8 turtles from the province of Chaco in 2020). It was not found at the localities in the Humid Chaco. Table 1 provides sex and measurements by sex divided by province. All turtles were in good health condition (Fig. 2) and females were, although not significantly, smaller and heavier than males (length: t = –1.11, df = 19, p = 0.28; mass: t = 0.26, df = 19, p = 0.80). In some localities from Salta the species was found sharing ponds with the Chaco Side-necked Turtle, A. pallidipectoris.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Table 1. Captures (n), mean (m), standard deviation (SD), and minimum–maximum (R) of maximum straight carapace length (SCL, mm), and mass (W, g) of Kinosternon scorpioides scorpioides from Salta, Formosa, and Chaco provinces, Argentina.

View Fullscreen

Environmental Features of the Capture Sites. — The species was searched for in a variety of water bodies, but it was consistently found in 3 types of environments: 1) flooded ditches along unpaved secondary roads, 2) peridomiciliary water reservoirs, and 3) natural water reservoirs (Fig. 3). Ditches on the sides of roads accumulate water, sometimes in large quantities, generating a continuous channel along the road that can reach up to 140 m long (and even more) with low depth (usually no more than 50 cm) depending on rain intensity. The peridomiciliary water reservoirs were constructed by people to stock water for farm animals and were placed near houses in a Chaco forest matrix outside the urban nucleus. These water bodies varied in size; however, they were usually moderately large (with a maximum recorded perimeter and depth of 230 m and 70 cm, respectively) with a variety of shapes. They frequently harbored all kinds of garbage on the bottom and margins (vehicle parts, bottles, plastic bags, etc.). Finally, natural water reservoirs were medium-sized (68–140 m perimeter and 35–50 cm depth), irregular in shape, and always surrounded by the typical Chaco forest matrix away from the vicinity of the houses. The common feature that links the 3 types of water bodies where turtles were found is the strong dependence on seasonal rains to fill them, most being temporary and dependent on the duration and intensity of the rainy season. In general, we detected low levels of conductivity (113.3–431 µS · cm^–1), high water temperature values (23°C–33.8°C), pH values consistently neutral (7.3–7.8) to slightly alkaline (8.48, in a pool from Chaco), and strongly varying values of dissolved oxygen: 1.2–2.3 mg · L^–1 (Salta), 4.6 mg · L^–1 (Formosa), 6.5–13.2 mg · L^–1 (Chaco).

View Figure

• Download as Powerpoint
• Download Figure

The environmental matrix that surrounded most inhabited water bodies was composed of primary and secondary native forest but also grassland, with a coverage ranging from 61% to 100% (Fig. 4). The sites in Salta presented the highest percentage of agricultural areas in the surroundings, with a maximum percentage of cover between 22% and 38%, and contrasting with the very low values found for Chaco (< 3%) and Formosa (< 1%; Fig. 4).

View Figure

• Download as Powerpoint
• Download Figure

Diet Composition. — Stomach contents were obtained from turtles from Salta (5), Formosa (7), and Chaco (8) provinces. Fecal matter was obtained from turtles from Salta (8), Formosa (6), and Chaco (1). Across all sample sources, 38 different prey item categories were recorded. These included invertebrates, vertebrates, and plant material (see Table 2). Turtles from Chaco and Salta had consumed the lowest number of animal prey types (17 in Chaco and 18 in Salta), whereas those from Formosa ate 25 animal prey items. We found that 60.9% of the turtles had plant material (seeds) in the tract (14 turtles), and was found in 57.1% of stomach samples and 60.0% of fecal samples. All seeds ingested by turtles from Formosa were grass seeds (probably Echinochloa sp.), while those seeds ingested by turtles from Salta and Chaco were from a single type of unidentified bush or tree.

Table 2. Percentages of relative importance index (%IRI) of each prey category for stomach contents (SC), fecal material (FM), and both combined (T) for specimens of Kinosternon scorpioides scorpioides from Salta, Formosa, and Chaco provinces. NI = not identified; bold italic values = fundamental item; bold values = secondary item; italic values = accessory item; roman values = accidental item.

View Fullscreen

The complete list of item categories with the corresponding percentage of IRI value (%IRI) discriminated by province and sample source is shown in Table 2. In general, the IRI ranked seeds as the most important dietary item (fundamental item in stomach and fecal samples from Formosa, and in fecal samples from Chaco and Salta), followed by anurans (secondary item in fecal samples from Chaco: 57.26%), anisopteran naiads (accessory item in fecal samples from Salta: 42.58%), planorbid snails and belostomatid water bugs (accessory items in stomach and fecal samples from Formosa: 27.67% and 25.08%, respectively). The stomach samples from turtles from Salta and Chaco differed somewhat from the summary above: stomachs from Salta turtles showed ostracods as a fundamental item (in this case the seeds were accessory: 39.7%) and stomachs from Chaco turtles indicated copepods as a fundamental item (100%, seeds also being fundamental: 87%). Other food items were a minor part of the diet and were classified as accidental (%IRI values lower than 25%; see Table 2).

The analysis of the animal portion of the diet of each province indicated a high number of invertebrates, particularly arthropods, in the diet of K. s. scorpioides: 88.9% of the animal categories preyed upon by Salta turtles were invertebrates (66.7% arthropods). Similar results were found for Chaco and Formosa turtles: 88.0% (Formosa) and 76.5% (Chaco) of the animal portion were invertebrates (arthropods representing 76.0% and 100%, respectively).

Although presenting differences in numeric and occurrence frequencies, turtles from the 3 provinces preyed upon the similar set of aquatic invertebrates: microcrustaceans, aquatic coleopterans, planorbid snails, dragonfly naiads, and aquatic heteropterans. Ostracods (87.5%) were on top of the rank of occurrence frequency in the animal portion for Salta turtles, followed by the adult aquatic coleopteran family Hydrophilidae (75%), dragonfly naiads of the order Anisoptera (50%) and the water bug family Belostomatidae (37.5%). Based on numeric frequency, the most important categories were ostracods (7.9%), planorbid snails (3.8%), and the adult aquatic coleopteran family Hydrophilidae (3%).

Samples from Formosa turtles differed slightly from those in Salta: the occurrence frequency was dominated by the adult aquatic coleopteran families Hydrophilidae (85.7%) and Dytiscidae (71.4%), ants (71.4%), the water bug families Belostomatidae and Corixidae (both 57.1%), and planorbid snails (57.1%), whereas the numeric frequency was dominated by planorbid snails (1.1%) followed by the aquatic coleoptera family Dytiscidae (0.4%).

Finally, samples from Chaco showed copepods as the item with the highest frequency of occurrence (62.5%), followed by ostracods (50%), the adult aquatic coleopteran families Hydrophilidae and Dytiscidae (50% each), cladocerans, and dragonfly naiads of the order Anisoptera (both 37.5%). The numeric frequency of these samples is predominated by copepods (37.2%) followed by cladocerans (9.4%) and the water bug family Dytiscidae (7.9%).

Tadpoles, ostracods, anisopteran naiads, aquatic coleopterans, and water bugs were the item categories shared by the turtles from the 3 provinces. In fact, aquatic coleopterans and water bugs were present in both stomach and fecal samples from the turtles of the 3 provinces, and the other items were distributed in one or more different samples.

The values of the preference index indicated that the items in which there was a significantly higher consumption than expected by chance were microcrustaceans, aquatic gastropods, and hirudineans (Salta); adult aquatic coleopterans and aquatic gastropods (Formosa); and microcrustaceans (Chaco; see Table 3).

Table 3. Preference index values (W_i) for the diet of Kinosternon scorpioides scorpioides from Salta, Formosa, and Chaco provinces followed by the corresponding interpretation according to the significance test proposed by Manly et al. (1993). S+ = positive, S– = negative; NS = no selection.

View Fullscreen

The values of the diversity index H′ for the available aquatic prey in each province (H′a) were significantly higher than those obtained for the diet of the species (H′d) consumed by the turtles: Salta H′a: 2.06 vs. H′d: 1.75 (t = 6.11; df = 483; p < 0.0001); Formosa H′a: 1.74 vs. H′d: 1.57 (t = 2.40; df = 391; p = 0.017); and finally, Chaco H′a: 1.74 vs. H′d: 0.79 (t = 18.08; df = 892; p < 0.0001). Similarly, we found significant differences in paired comparisons of dietary indexes among provinces (H′d vs. H′d; below diagonal in Table 4). Paired comparisons of environmental availability indexes between provinces (H′a vs. H′a; above diagonal in Table 4) were only significant for Salta in comparison to the other 2 provinces, which did not differ significantly from each other.

Table 4. Left side of table (below diagonal): pairwise comparisons of diet diversity of Kinosternon scorpioides scorpioides (H′d) between provinces (H′d vs. H′d); right side of table (above diagonal): pairwise comparisons of environmental diversity (H′a) between provinces (H′a vs. H′a).

View Fullscreen

Finally, the percentage by volume of plant material indicated an omnivore preference in Salta and Formosa turtles, with seeds representing 35.12% and 36.96% of the diet, respectively. In contrast, turtles from Chaco showed a tendency toward a carnivorous diet, with only 21.71% of the total volume being plant material. Different populations of the species seem to display different feeding strategies (Fig. 5). The graphic method showed that Salta turtles practiced a mixed feeding strategy with varying degrees of specialization toward seeds and freshwater bivalves, and a generalization with most turtles preying upon many types of resources simultaneously (Fig. 5A). In Formosa there was a clear specialization on seed consumption (Fig. 5B). Finally, turtles from Chaco practiced a more general strategy by preying upon many items with no predominance of any of them (Fig. 5C). The standardized index of Levins indicated a narrow trophic niche of the turtles in the 3 provinces (Bst Salta: 0.04; Bst Formosa: 0.003; Bst Chaco: 0.21).

View Figure

• Download as Powerpoint
• Download Figure

DISCUSSION

Knowledge of the habitat requirements, feeding ecology, abundance, and distribution of a given species constitutes vital information for assessing its conservation status and for planning the best strategies to protect it. This study was intended to provide knowledge about the biology of K. s. scorpioides at the southern limit of the species range. This widely distributed subspecies was reported in many biomes in the Neotropics: Venezuelan Llanos (Fiasson 1945); Dry Tropical Forest in Colombia (Cáceres-Martínez et al. 2017), Ecuador (Cisneros-Heredia 2006), and Brazil (Carvalho et al. 2008; Barreto et al. 2009; Hernández-Ruz et al. 2016); Yungas Forest in Argentina (Prado et al. 2012); Dry Chaco in Argentina, Bolivia, and Paraguay (Gonzales 1998; Acosta et al. 2013; Buskirk 2007); Amazonia, Caatinga, Cerrado, Coastal Atlantic Rainforest, and Pantanal in Brazil (Vanzolini et al. 1980; Andrade 2019); Brazilian Submontane Tropical Rainforest (Costa et al. 2010); and Altitudinal Atlantic Rainforest patches inserted in the Caatinga in northeastern Brazil (Freitas et al. 2019). Regardless of the biome, the type of aquatic habitat that K. s. scorpioides seems to prefer is shallow water with a soft muddy bottom, and subject to seasonal and daily fluctuations in physicochemical parameters. The literature reveals that K. s. scorpioides inhabits warm (23°C–30°C: Pereira et al. 2007; present work) and acidic (pH 4.26–6.65: Pereira et al. 2007) to nearby neutral (pH 6–7.83: present work) waters, with low to moderate dissolved oxygen values between 1.46 and 8.40 mg · L^–1 (this work also includes the range published by Pereira et al. 2007), and conductivity varying from 70 to 480 µS · cm^–1 (data from Pereira et al. 2007 that includes the range observed by our team). The strong dependence on seasonal rains appears to be valid only for microhabitats in dry areas of the subspecies range (e.g., Caatinga, Dry Chaco). The diversity of aquatic environments in which this species is mentioned in the literature includes shallow ditches (Vinke and Vinke 2001; Buskirk 2007); temporary ponds and small lagoons (Gonzales 1998); seasonally flooded lakes (Barreto et al. 2020); flooded lowlands (Fiasson 1945); marshes, wells, and irrigation channels (Bedoya-Cañón et al. 2018); lakes, marshes, and slow-flowing streams (Carvalho et al. 2008); artificial ponds to stock water for cattle (Tomas et al. 2015); margins and overflow areas of streams, rivulets, and rivers (Cisneros-Heredia 2006; Costa et al. 2010; Cáceres-Martínez et al. 2017); and even rivers, but without further details (Pereira et al. 2018). Finally, Montes-Correa et al. (2017) mentioned a lot of habitat types used by the species without discriminating which of them correspond to the subspecies K. s. scorpioides. We verified some of these environments in our surveys along the Dry Chaco of Argentina: ditches on secondary roads, natural temporary ponds surrounded by a forest matrix, and peridomiciliary artificial ponds to stock water for cattle.

The presence of the subspecies in peridomiciliary ponds suggests tolerance of low–moderate densities of human settlements (never exceeding 17,000 inhabitants, see data below), including polluted water bodies and watercourses. Our survey did not demonstrate the species in association with the periphery of towns, but this had been verified 2 decades ago (1997) in Laguna Yema (2744 habitants; Instituto Nacional de Estadística y Censos [INDEC] 2010), in the Dry Chaco ecoregion on Formosa Province, and many years ago, in towns placed in the Yungas–Dry Chaco transition zone (Ledesma and Yuto: Jujuy Province; Tartagal: Salta Province; see Freiberg 1967), and in the Yungas ecoregion (Orán and Tabacal: Salta Province; see Freiberg 1936). Nowadays, those towns mentioned by Freiberg (1936, 1967) support between 12,579 (El Tabacal plus Hipólito Yrigoyen) and 82,413 people (Orán), exceeding 50,000 people in Ledesma plus Yuto and Tartagal (INDEC 2010). During the years in which the species was cited there, the human population was significantly lower: it ranged between 2589 people (Yuto; INDEC 1971) and 16,740 people (Tartagal; INDEC 1971), while the entire Oran department (including Tabacal and Orán, among other towns) housed 16,624 people (according to the national census of 1930: see Benclowicz 2011). But low population density does not imply pristine environments. Most of the towns mentioned by Freiberg were born as important centers of sugar cane production (Tabacal, Ledesma) and oil wells (Tartagal), with the presence of intense deforestation processes that prevail until today. The Yungas region has been subject to an intense deforestation process related to the production of bananas, citrus, coffee, cattle, and soy, particularly in the east of Tartagal, in the transition with the Chaco region (Rodríguez and Silva 2012). Large deforestation processes associated with coal mining have been reported for Tropical Dry Forest populations of the subspecies in northeastern Colombia (Cáceres-Martínez et al. 2017). This tolerance of human-impacted habitats was reported early on by Pritchard and Trebbau (1984), and the turtle seems to occur even in habitats polluted by household waste and sewage, as reported by Pereira et al. (2018) from along the Cariús River at Nova Olinda municipality in northeastern Brazil with a population of 15,000.

Our analysis of the land cover and land use matrix surrounding the water bodies where we found K. s. scorpioides, together with field observations, revealed a prevalence of primary and secondary Chaco forest (at least 60%), but with a certain tolerance to areas subject to deforestation for crops and cattle, as in certain locations in the province of Salta. Thus, the available evidence supports the idea that K. s. scorpioides is capable of inhabiting a wide range of aquatic environments (even those polluted and disturbed by human presence), and it tolerates in some degree the fragmentation of the forest matrix. Nevertheless, a periodic monitoring program of certain populations is needed to better understand this landscape, since it has been demonstrated for other turtles that there may be a lag period between the beginning of anthropogenic disturbance and its effect on population demography (Eskew et al. 2010).

Diet composition is a very important aspect of the life history of a turtle species: feeding studies provide data on the available environmental resources needed to satisfy the energetic needs of individuals (Macip-Ríos et al. 2010). The feeding data obtained here for the southernmost populations of K. s. scorpioides confirmed the omnivorous feeding habits proposed in previous literature (Vanzolini et al. 1980; Rueda-Almonacid et al. 2007).

We observed the consumption of a wide variety of aquatic and land invertebrates, small aquatic vertebrates, and even an astonishing number of seeds, particularly of an unidentified grass in Formosa, possibly of the genus Echinochloa. Interestingly, a high rate of ingestion (1138) of grass seeds was evidenced in fecal samples of 16 of 23 studied individuals of K. s. scorpioides from Serra dos Carajas (eastern Amazonia, Brazil; Carvalho et al. 2008). In addition, a plant-based diet was described based on an unspecified number of fecal samples taken from specimens of K. s. scorpioides from the Brazilian Amazonia (Vogt et al. 2009). That study also reported the presence of 711 unidentified grass seeds in 9 turtles, confirming that seeds are an important food resource for this turtle, contrary to the previous assumption that this subspecies is primarily a predator and scavenger in nature (Berry and Iverson 2011), and even that plant material is not consumed by this turtle (Vinke and Vinke 2001). Carvalho et al. (2008) also reported many arthropods, mainly aquatic but also terrestrial species that fell in the water, and some vertebrates that were also eaten by K. s. scorpioides from Serra do Carajas. In the latter study, water bugs of the genus Limnocoris, together with immature and adult dragonflies, dominated the animal fraction of the diet, with vertebrates representing only a minor portion, and probably ingested as carrion. Carvalho et al. (2008) highlighted the great difference they found between the low presence of anurans in fecal samples and the high abundance of them within the habitat where turtles were caught, attributing such difference to the study of feces only (higher digestibility leads to fewer remains in feces). The values of anuran ingestion in turtles of the 3 populations studied by our team show a low consumption of anurans independently of sample source (feces vs. stomach flushing), and environmental availability (always high), and this was also revealed by the analysis of preference: tadpoles were selected negatively in the 3 studied populations (a situation that could be interpreted as the turtles' difficulty in capturing tadpoles).

Plant consumption was marked in all populations with IRI values signaling seeds as a fundamental item and suggesting intentional consumption of plant material (Table 2). Costello's (1990) graphs showed a different contribution to the specialized feeding strategy of seed consumption in turtles from Salta and Formosa: some but not all turtles from Salta displayed an individual specialization on seed consumption (upper left part of the Fig. 5A and coinciding with an intermediate value of niche width), while those from Formosa specialized on seed consumption appear as a population specialization instead of an individual feature (upper right part of the Fig. 5B and reflected in the small value of the index of Levins). However, in Chaco, although the IRI also catalogued seeds as a fundamental item (Table 2), it could be said that, according to the interpretations of the graphical method (lower right part of the Fig. 5C), the plant material had been eaten occasionally by most of the individuals (general strategy). This fact would correlate with the highest value of the standardized index of Levins and with the carnivorous feeding habit according to the percentage of the volumetric proportion of the plant material.

The secretive habits of K. s. scorpioides and the lack of fieldwork designed specifically to detect the species explain the scarcity of records in Argentina. In fact, 77 yrs had passed since the first report on the species in the country (Salta Province; Freiberg 1936) until the first report for the province of Chaco (Acosta et al. 2013). Meanwhile, the distribution of the species was constructed by a multiplicity of casual findings (road-kill specimens, live specimens crossing a road, dry carcasses of dead turtles). As stated in the “Methods” section, we undertook an intense sampling effort in many environments across 4 provinces belonging to the Dry and Humid Chaco ecoregions. In Argentina, K. s. scorpioides inhabits most of the ecological complexes (sensu Morello et al. 2012) of the Semiarid Dry Chaco subregion (1, Middle and High Pilcomayo; 2, Pilcomayo-Bermejo Interfluvium; 3, Juramento–Salado Paleochannels; 4, Central Forest and Shrub Lands; 5, Bermejito–Teuco–Bermejo; and 6, Itiyuro Alluvial Fan) and in nearby areas of the Yungas ecoregion (Selva and Pedemontane Grassland), but with no records from the Humid Chaco ecoregion to the east. These ecoregions and ecological complexes embrace the Teuco–Bermejo (Bermejito and San Francisco rivers among others) and Pilcomayo River basins. Finally, the present study is the first to provide baseline information to approach an understanding of the habitat requirements and feeding habits of the southernmost known populations of K. s. scorpioides. We believe that our work could certainly have an impact on future projects, stimulating ecological research on this secretive turtle.