ASSESSMENT OF POPULATION CHARACTERISTICS AND HABITAT USE OF PODOCNEMIS VOGLI (MÜLLER 1935) DURING A REPRODUCTIVE SEASON AT THE BOJONAWI PRIVATE NATURE RESERVE (VICHADA DEPARTMENT, COLOMBIA)

.— We conducted a short study in the Bojonawi Private Nature Reserve (Vichada department, Colombia) in three types of aquatic macrohabitats to characterize habitat use, population structure and reproductive phenology of Podocnemis vogli during one reproductive season. We also reevaluated the degree of sexual dimorphism in this species. To maximize capture rates, we used three methods: funnel traps, trammel and duration of the study and on the methods employed (Bernhard & Vogt, 2012; Bernardes et al., 2014; Alzate-Estrada et al., 2020). For some podocnemidid species it has been shown that characteristics of different water bodies are associated with different probabilities of detecting individuals of different size classes. For example, for both P. lewyana and P. erythrocephala, shallower water bodies are associated with a greater proportion of small individuals being registered (Gallego-García & Castaño-Mora, 2008; Bernhard & Vogt, 2012; González-Zárate et al., 2014). However, the greater ease of detecting such individuals in these types of habitat does not mean they avoid deeper water bodies (Páez et al., 2015; Alzate-Estrada et al., 2020). Several other studies on other turtle species also have documented differences in detectabilities depending on sex class distributions, but sex ratios differed, with an excess of males in the river and females in the lagoon. As with other congeners, P. vogli is a sexually dimorphic species where females attain larger sizes. In this study we documented that pre-cloacal tail length and nail length on the anterior third toe are the most useful secondary sexual characteristics for sexing subadult and adult individuals. Some females larger than 22 cm straight-line carapace length had calcified eggs in their oviducts. This species is abundant within the reserve and appears to be a habitat generalist, including the use of large rivers. More extensive and intensive monitoring is needed to document specific class survivorship rates, detectabilities and population dynamics. results illustrate how caution should be used in interpreting data on sex ratios that are collected in a few habitat types over short sampling periods; we have shown sex ratios in P. vogli to vary spatially over short distances and more long-term studies also might show that sex ratios in each macrohabitat may vary temporally.

and/or size class (reviewed by Tesche & Hodges, 2015). Finally, the different capture methods used to study podocnemidids are known to have biases in terms of the sexes and size classes they sample De Souza et al., 2013;Bernardes et al., 2014;Portelinha et al., 2014). In the few long-term studies on these species, one of the patterns that emerges is that most of the population is comprised of juvenile individuals, small females, and males, but the scarcer larger females are difficult to recapture Bernardes et al., 2014;Portelinha et al., 2014;Alzate-Estrada et al., 2020).

INTRODUCTION
A basic knowledge of life history characteristics and population structure is essential for predicting how changes in population vital rates due to natural or human impacts might affect the population's distribution, patterns of habitat use, and permanence (Heppell, 1998;Böhm et al., 2016;Ripple et al., 2017). For example, in neotropical freshwater turtle species of the Family Podocnemididae, changes in vegetative structure (both floating and riparian), in the width and depth of water bodies, and in the permanence of nesting beaches may lead to changes in home range use and population dynamics De Souza et al., 2013;Miorando et al., 2013;Portelinha et al., 2014;Barahona & López, 2015;Páez et al., 2015;Eisemberg et al., 2016). In addition, in Podocnemis species, both sex ratios and proportions of individuals in the different age/size classes in different habitat types (rivers, channels, wetlands) are not static, because they are affected by seasonal hydrological cycles that differentially affect movement patterns by members of these different groups (Fachin-Terán et al., 2006;Naveda-Rodríguez et al., 2018;Ponce De Leão et al., 2019;Alzate-Estrada et al., 2020).
Not only does spatial and temporal variation in local habitat characteristics complicate the understanding of podocnemidid population structure, the results of studies also depend heavily on the location and duration of the study and on the methods employed Bernardes et al., 2014;Alzate-Estrada et al., 2020). For some podocnemidid species it has been shown that characteristics of different water bodies are associated with different probabilities of detecting individuals of different size classes. For example, for both P. lewyana and P. erythrocephala, shallower water bodies are associated with a greater proportion of small individuals being registered (Gallego-García & Castaño-Mora, 2008;González-Zárate et al., 2014). However, the greater ease of detecting such individuals in these types of habitat does not mean they avoid deeper water bodies Alzate-Estrada et al., 2020). Several other studies on other turtle species also have documented differences in detectabilities depending on sex class distributions, but sex ratios differed, with an excess of males in the river and females in the lagoon. As with other congeners, P. vogli is a sexually dimorphic species where females attain larger sizes. In this study we documented that pre-cloacal tail length and nail length on the anterior third toe are the most useful secondary sexual characteristics for sexing subadult and adult individuals. Some females larger than 22 cm straight-line carapace length had calcified eggs in their oviducts. This species is abundant within the reserve and appears to be a habitat generalist, including the use of large rivers. More extensive and intensive monitoring is needed to document specific class survivorship rates, detectabilities and population dynamics.
At the national level, P. vogli is classified in Colombia as Least Concern (LC) based upon its apparent abundance (Morales-Betancourt et al., 2015). However, it is possible that population sizes are declining in some places due to increased harvest that represent a replacement phenomenon, as hunters shift their attention away from more depleted podocnemidid species such as P. expansa and P. unifilis (Morales-Betancourt et al., 2012a;Martínez-Callejas et al., 2015).
The establishment of a demographic baseline on abundance, age/size structure, sex ratios, and habitat use for P. vogli populations that have not suffered substantial harvest or habitat degradation are vital as a reference to guide management efforts for more impacted populations (Rhodin et al., 2018). Given that the Bojonawi Private Nature Reserve (BPNR) is relatively well conserved and protected, our goals were to document these aspects of its P. vogli population during the reproductive season, which allowed us to evaluate as well the minimum size of sexual maturity in females and document which secondary sexual characteristics are most useful for sexing subadult and adult individuals. We expected to find differences in the proportions of size classes and genders of P. vogli among macrohabitats, given that during the reproductive season females of species in the Family Podocnemididae tend to disperse to areas appropriate for nesting. We also expected to find differences among capture methods, because every capture method suffers from some particular biases related to the size and/or sex of the animals that are captured.

Study Site
This study was conducted from 6 January to 6 April during the 2017 reproductive season in the floodplain of the Orinoco River in the BPNR within the buffer zone of the Tuparro National Park and UNESCO Biosphere Reserve (Fig. 1). The BPNR is located 15 km southwest of the city of Puerto Carreño (Vichada department, Colombia) at an elevation of 51 masl (6.09799°N, 67.48321°W). The region has a unimodal rainfall regimen, with a dry season extending from December to March. The mean annual temperature is 28 ºC and the mean annual precipitation is 2200 mm (Fundación Omacha, 2020). The BPNR is located along the Orinoco River and is mainly composed of savannah grassland, rocky outcroppings of the Guiana Shield, "morichales" (riparian vegetation dominated by the Moriche palm, Mauritia flexuosa), and clear and white-water channels draining seasonally flooded forests. We chose to focus field work during the reproductive season to be able to detect the presence of oviductal eggs within the reproductive tracts of P. vogli females.
We selected three different aquatic macrohabitats for this study: the main channel of the Orinoco River, the El Pañuelo lagoon, and a small tributary channel to the Orinoco River, the El Tesoro channel; the mean linear distance between the sites was 1.55 km (SD = 1.31; Fig. 1). The river and lagoon habitats were wide (average of 1218 m and 157 m, respectively), with warm waters of low transparency. In contrast, the channel was narrow (with an average width of 5.6 m), with considerable vegetative cover and lower water temperatures (Gómez, 2018;Parra-Henao et al., 2019). All macrohabitats have been shown to differ in their physiochemical characteristics, sedimentation levels, nutrient contents, current flow, water transparency, and depth (Lasso et al., 2014).

Capture methods
To maximize capture rates, we used three capture methods (funnel traps, trammel nets, and manual capture), following the recommendations for working with freshwater turtles of De la . Because of the specific conditions in each macrohabitat during the reproductive season, it was only possible to sample them in a standardized manner (using funnel traps); for this reason, all statistical comparisons of macrohabitats are based solely on these data. We made a total of three sampling periods in each of the three macrohabitats, with five days of capturing effort in each sampling period and a two-week time interval between the start of each period per macrohabitat. In each sampling period, we installed 15 funnel traps; nine were double funnel traps (five with 0.8 m diameter hoops and 1.5 m in length, four with 0.38 m diameter hoops and 1.6 m in length) and six were single funnel traps (with 0.8 m diameter hoops and 1.05 m in length). We set the traps in shallow locations with little current near basking sites. We checked the traps three times daily (09:00, 15:00, and 21:00 h). We used three different types of bait: 10 traps were baited with fish, usually guabina (Hoplias malabaricus) and corn (Zea mays), and five traps were baited with corn and banana (Musa paradisiaca). In addition, when possible, we placed in all traps a raceme of leaves of "chigo" (Campsiandra comosa). Total funnel trap sampling effort (16200 trap-h) was standardized equally among macrohabitats: 120 h (five days) * 3 (sampling periods) * 15 (traps) = 5400 trap-h/macrohabitat. After setting or checking the traps, we focused on the other two capture methods. When turtles were visually detected (between 09:00 and 23:00 h), manual capture was attempted. In the El Pañuelo lagoon, manual capture was conducted in shallow water accessible from the shoreline or from a small boat. In the El Tesoro channel, we also manually captured individuals encountered while snorkeling. In the Orinoco River it was not possible to use this method due to the depth, current, and turbidity of the water. The effort for this method was not standardized; of a total of 23 person-h of searching, only 4 h occurred in the El Pañuelo lagoon vs. 19 h in the Tesoro channel.
Capturing turtles with the trammel net was only possible in the lagoon macrohabitat, because in the other two macrohabitats the lead weights of the net could not settle on the bottom evenly, providing the turtles a means of escape. Prior to passing the trammel net (mesh size of 5 cm, net size of 80 m wide X 5 m high) the site to be netted was baited by submerging in the area a mix of dry dog food and ground corn. In total, we conducted 15 passes in appropriate areas (of moderate depth, with few submerged objects), for a total effort of 6 person-h.

Measuring and marking individuals
All captured individuals were marked, weighed, measured, sexed, and photographed prior to their release at the capture location. Turtles were marked by cutting notches on marginal scutes using the modification of the Cagle (1939) system proposed by Rueda-Almonacid et al. (2007). Depending on the size of the individual, turtles were weighed using 1 or 20 kg dynamometer scales (Pesola, ± 10 or ± 200 g, respectively). The morphometric variables of straight-line carapace length (SCL), straight-line carapace width (SCW), maximum plastron length (MPL), and pre-cloacal tail length (PreCTL) were measured with digital 15 cm or 50 cm calipers (Mitutoyo and Haglof, ± 0.01 mm and ± 5 mm, respectively). We also measured the length of the  third nail (NL, ± 0.01 mm) on the front left extremity, given that this trait has been reported to be useful for sexing some turtle species (Rueda-Almonacid et al., 2007).

Sexing and estimating size at sexual maturity
Individuals were sexed based on their body sizes, PreCTL, shape of the anal scale (with a "V" shape typical of females vs. a "U" shape typical of males, Morales-Betancourt et al., 2012b), and the presence of yellow spots on the head, especially in the snout area in males. Following the recommendations of Ramo (1982), males were considered sexually mature at body sizes of 13 cm SCL, and females at 20 cm SCL; these reproductive sizes were based upon an inspection of gross anatomy, including the size of follicles within the ovaries and the development of oviducts in females or the morphology of the testes and presence of sperm in the epididymis of males. In this study, we inspected the reproductive condition of captured females using a Bondway sonogram (BW560V) by examining the inguinal region in individuals large enough to accommodate the probe, attempting to detect the presence of developing eggs. To facilitate comparison with previous studies of this species (Alarcón-Pardo, 1969;Ramo, 1982), we used the following five size classes and categories for the population structure analyses: juveniles (SCL < 9 cm), subadult females (9 cm ≤ SCL < 20 cm), subadult males (9 cm ≤ SCL < 13 cm), adult females (≥ 20 cm) and adult males (≥ 13 cm).

Statistical analyses
The overall relative abundances in the three macrohabitats were estimated as the number of captures per trap-h, using the standardized funnel trap data. We tested for differences in the number of individuals captured during each sampling period in each macrohabitat, using a heterogeneity Χ 2 to see if the observed abundances differed from a random pattern. We also used a chi-squared goodness-of-fit test to inspect for equal sex ratios (1♀:1♂). Differences among body sizes between macrohabitats were inspected using an ANOVA on the SCL data. A multivariate MANOVA was used to examine which morphometric variables differed between sexes and a linear discriminant analysis was used to determine which variables were most useful for sexing subadult and adults.
We examined movements within the study site based on the data from the recaptured turtles using Advance ArcMap version 10.2.2.3552 of the ArcGis software (version 10.2.2). All analyses were conducted using RStudio (version 1.0.153, RStudio Inc., 2017) by the R Project software (version 3.4.0, R Development Core Team, 2017).

RESULTS
With the three capture methods, we obtained a total of 149 P. vogli individuals, but only recaptured 11. Of these turtles, only three were juveniles, three were sub-adult males, 34 were subadult females, 67 were adult males and 42 were adult females. The number of individuals captured varied among methods and macrohabitats (Table 1). During the day, we frequently observed P. vogli individuals basking or floating on the surface in all three macrohabitats.
We captured 81 of the 149 individuals using the standardized funnel trap method, as well as obtained eight of the 11 recaptures in these traps. Relative abundances based upon these data was highest in the El Pañuelo lagoon (0.007 individuals/trap-h), and lower but comparable in the Orinoco River (0.005 individuals/ trap-h) and El Tesoro channel (0.004 individuals/trap-h). Of the total captures obtained using funnel traps, 19.1 % were captured during the first sampling period, 38.2 % during the second period, and 42.7 % during the last sampling period. These differences did not differ significantly from a null hypothesis of equal captured proportions in each period (X 2 2 = 4.74, P = 0.09). We found no differences in the size distributions of individuals captured with the funnel traps in the three macrohabitats (ANOVA, SCL, F 2,80 = 0.251, P = 0.778). Capture success with the funnel trap method was greatest between 09:00-21:00 h, representing 76.4 % of all captures vs. only 23.6 % in the 21:00-09:00 interval.
Using the hand-capture method, we captured 28 and recaptured two individuals in the two macrohabitats where this technique could be implemented (lagoon and channel), for a success rate of 1.3 individuals/h. We did not inspect statistically for differences with this method between those two macrohabitats due to very different sampling efforts among them (4 h vs 19 h). With the trammel net in the El Pañuelo lagoon we captured 40 turtles, plus one recaptured individual that had been initially trapped in a funnel trap, for a success rate of 6.83 individuals/h. Sex ratios differed among macrohabitats (heterogeneity X 2 2 = 8.922; P < 0.01), being unbiased in the El Tesoro channel (1♀:1♂; X 1 2 = 0; P = 1.0), significantly male biased in the Orinoco River (1♀:2.6♂; X 1 2 = 4.84; P < 0.01), and significantly female biased in the El Pañuelo lagoon (1♀:0.5♂; X 1 2 = 4.33; P < 0.01), yielding an overall balanced sex ratio for the population over all three macrohabitats (1♀:1.07♂; X 2 2 = 0; P = 1.0). Taken as a whole, there was significant sexual dimorphism between the sexes (subadult and adult females vs. subadult and adult males; HOTELLING MANOVA test: F 1,145 = 19.262; P < 0.001), with females being heavier, larger and wider (SCL, MPL, SCW) than males, with a maximum SCL of 30 cm and body weight of 3 kg (Table 2, Fig.  2). The original variables that best distinguished the sexes for all sizes (except juveniles) were PreCTL, NL, MPL, and SCL (Linear discriminant analysis: LD1 = 95.5%; LD2= 4.5%; Table 3).
The minimum body size for females where it was possible to insert the sonogram probe into the inguinal region was 22 cm SCL. From a sample of 27 females with a CL larger than 22 cm SCL, nine were found to have oviductal eggs in differing stages of calcification, with the last such individual recorded near the end of February. The smallest female detected to carry oviductal eggs was 24.1 cm SCL.
The mean linear distance moved by the recaptured individuals was 380 m (range 34 -1602 m). Males (n = 9) moved approximately twice as far as females (n = 6), with a male mean linear distance = 468 m (range: 34 -1602) and a female mean linear distance of 226 m (range: 92 -544). Given the limited sample sizes, we did not compare these data statistically.

DISCUSSION
During the 2017 reproductive season at the BPNR, P. vogli did not show a marked segregation in terms of macrohabitat use, with both sexes and most size classes occupying each macrohabitat type sampled, including the main channel of the Orinoco River.
Other studies with this species have reported it to exhibit a preference for lentic habitats, such as lagoons and shallow channels (Alarcón-Pardo, 1969;Ramo, 1982). This discrepancy between previous characterizations and our results may be related to diverse causes, including the time of year when field work was conducted (De Souza et al., 2013;Eisemberg et al., 2016;Ponce De Leão et al., 2019;Alzate-Estrada et al., 2020), the capture methods employed Bernardes et al., 2014;Portelinha et al., 2014;Parra-Henao et al., 2019), and the effort invested in sampling different types of habitat with different methods (Bernhard, 2010;Páez et al., 2015). Due to our use of a standardized method (funnel traps) appropriate for all three of the macrohabitat types we sampled, and the comparable sampling effort in each, we were able to show that the relative abundance of P. vogli in the river was comparable to the relative abundance of individuals in the lentic channel macrohabitat. Studies that rely on other sampling methods were incapable of detecting P. vogli in lotic habitats, such as manual capture, could give an erroneous impression that the species avoids large rivers (Rueda-Almonacid et al., 2007;Morales-Betancourt et al., 2012a). The fact that previous studies have not reported use of river habitat by P. vogli also might be related to geographic variation in habitat preferences. We worked in the Orinoco River floodplain, while previous studies with this species were conducted in llanos piedmont or in seasonally flooded pluvial plains (Alarcón-Pardo, 1969;Ramo, 1982) where P. vogli individuals may indeed not enter larger water bodies. Regardless, our study is the first to show that in some locations P. vogli individuals may be present in substantial numbers in large riverine habitats. Unfortunately, the low number of recaptures in this study precluded the estimation of detectabilites for each sex and/or size class to rigorously estimate actual densities in each macrohabitat, instead of having to compare estimates of relative abundances.
Sex ratios in turtle populations may vary on a local scale due to differences in habitat preferences of males and females (Portelinha et al., 2014;Viloria & Forti, 2015) and differences in mortality rates of each sex in different habitat types (Fachín-Terán % Vogt, 2004;Folt et al., 2016). In our study, we documented sex ratio skews in favor of males in the river macrohabitat and females in the lagoon macrohabitat. One possible explanation for this is the proximity of suitable nesting habitat (savannah grassland, Barrio-Amorós & Narbaiza, 2008) along the margins of the lagoon where females were found in greater numbers, combined with a tendency in males to be more vagile in general, as has been shown in other podocnemidid species (Fachín-Terán et al., 2006;Bernhard, 2010;Alzate-Estrada et al., 2020). Our    results illustrate how caution should be used in interpreting data on sex ratios that are collected in a few habitat types over short sampling periods; we have shown sex ratios in P. vogli to vary spatially over short distances and more long-term studies also might show that sex ratios in each macrohabitat may vary temporally.
Funnel traps successfully captured individuals of most size classes in all three macrohabitats, although juveniles were dramatically less commonly trapped than individuals larger than 9 cm. In contrast, the trammel net method was far more efficient in the macrohabitat (lagoon) where it was possible to use. In only six hours of sampling with this method, we captured roughly 50% of the total number of individuals sampled during 16,200 trap-h of effort with the funnel traps in all three macrohabitats. But the trammel net method had a significant bias, because no juveniles and fewer subadults than adults were obtained.
Manual capture of individuals that were detected visually also was a viable method in water bodies that were clear and with slow currents. While the results obtained with all methods indicate that P. vogli is primarily active during the day (71.9 % of captures), through use of the manual capture method we also documented for the first time that nocturnal foraging in P. vogli occurs.
It was possible to confidently distinguish between sex and sub-adult and adult individuals (SCL > 9 cm) based on secondary sexual characteristics, with PreCTL the most reliable morphometric variable (males have longer PreCTLs). This agrees with Alarcón-Pardo (1969) and Ramo (1982), where males were reported to have thicker and longer tails, up to double the length of females with comparable body sizes. This character also has been shown to be useful for sexing other podocnemidid species (Pritchard & Trebbau, 1984;Iverson, 1995;Bernhard, 2010). The second most useful morphometric character to distinguish sexes in this study was NL, but only for adult individuals. This trait also has been used to sex turtles belonging to genera in the Family Emydidae, where males have longer nails on their third digits as well (Rueda-Almonacid et al., 2007). The longer nails apparently function to help males grasp females during courtship and mating Ceballos et al., 2012;Escalona et al., 2012).
Although we may not have been able to detect reproductively active females with SCL < 22 cm (due to the size limitations imposed by our sonogram probe), the smallest female we detected with oviductal eggs using the sonogram had a SCL = 24 cm, which agrees with reports from other areas in Colombia and Venezuela (Alarcón-Pardo, 1969;Ramo, 1982;Pritchard & Trebbau, 1984). In terms of the phenology of the reproductive season, Alarcón-Pardo (1969) and Portocarrero-Aya (2008) reported that P. vogli in Colombia nest from November to January, while we documented oviductal females using the sonogram in February, which agrees with the dates for nesting reported by Ramo (1982) for Apuré, Venezuela.
The few recaptures during our relatively short study revealed only limited local movements of individuals, with no movements documented between macrohabitats. However, we did document one case of a long-distance movement involving a change in macrohabitat with our capture of the female marked during a previous study (in 2015 in the Bita River, 6.15669°N, 67.50761°W; Morales-Betancourt & Lasso, unpublished data), having dispersed 9.3 km. Ramo (1982) also documented comparable movements by male P. vogli during the period of flooding in the rainy season in the Venezuelan llanos. Future more intensive mark-recapture efforts in our study site will probably reinforce and extend the results obtained during our three-month study indicating that P. vogli individuals are capable of moving substantial distances during their ontogeny, as has been shown in other podocnemidid species (Fachín-Terán et al., 2006;Bernhard, 2010;Alzate-Estrada et al., 2020).