Diet of Dermatemys mawii, an Aquatic Turtle That Relies Heavily on Terrestrial Vegetation
Abstract
Dermatemys mawii is a critically endangered freshwater turtle endemic to Central America. In the wild, these turtles are thought to be wholly herbivorous as adults and feed on a variety of vegetation; however, no studies have quantitatively assessed potential dietary differences based on biotic and abiotic factors. The purpose of our study was to describe and quantify the wild diet of D. mawii and assess differences based on habitat, maturity, and sex. We evaluated the stomach contents of 66 turtles legally harvested by local hunters for personal consumption throughout the country of Belize. Percent volume (by displacement) and percent frequency of each stomach item were used to calculate an index of relative importance (IRI). One algal and 6 plant families contributed to an overall diet composition consisting of leaves, flowers, stems, seedpods, seeds, and fruit. Rocks and invertebrates were also consumed, although we believe these to be incidental consumption. The leaves of the riparian tree Inga edulis were present in 73.1% of turtle stomachs and accounted for almost half of the total volume of all stomach contents combined. We used Spearman rank correlation coefficients to test the null hypothesis that there was no correlation in the rankings of stomach items (i.e., there were differences) when comparing turtles by habitat, age, and sex. There were significant differences in the ranking of food items between river and lagoon habitats, with lagoon turtles relying heavily on the algae Nitella sp.; however, the stomach contents from both habitats were equally diverse (Hrivers = 1.68, Hlagoons = 1.64). There were no differences in IRIs between adults and juveniles or between males and females. Our results emphasize the importance of habitat in D. mawii diet selection and the importance of leaves from riparian plants species that are shed into their aquatic habitats.
Nutritional ecology provides a framework for understanding patterns of feeding and food selection (Choat and Clements 1998). The field of nutritional ecology is integrative and draws from various disciplines to develop hypotheses in an effort to explain phenomena related to the consumer and feeding, such as diet selection in herbivores. The optimal digestion theory posits there is a relation between diet and digestion such that variation in food selection, rate of intake, and digestive structure and function yield optimal digestion and absorption (Sibly 1981; Karasov et al. 2011). Therefore, the digestive strategy that is most beneficial to the consumer is one that maximizes the net amount of energy and/or nutrients absorbed from the ingested food per unit of time (Hume 1989). In herbivores, this may mean preferential selection of plant species or parts that are highly digestible (Choat and Clements 1998). Herbivorous species have demonstrated selective consumption in controlled feeding trials. In the species Pseudemys nelsoni, hatchling turtles consumed only Hydrilla ventricillata leaves and left the less digestible stems intact (Bjorndal and Bolten 1992).
Dermatemys mawii is a freshwater turtle endemic to Central America. This critically endangered species has been overhunted for local human consumption (Moll 1986; Polisar 1994, 1995; Polisar and Horwich 1994), and populations continue to decline due to overharvesting (Rainwater et al. 2012). Belize is the only country that still allows legal harvest of D. mawii (Section 12:02 (b) & (d) under the Fisheries Act Chapter 210 of the Laws of Belize Revised Edition 2000–2003). In the wild, adult D. mawii are herbivorous and consume aquatic vegetation and windfall plant matter (Moll 1989; Polisar 1992; Vogt et al. 2011). Herbivory in extant reptiles is not as prevalent as it is in mammals. Of the approximately 4100 species of terrestrial mammals, 25% are considered herbivores; of the approximately 3060 species of reptiles, only 3% are considered herbivores (Choat and Clements 1998). Previous studies have indicated that D. mawii are probably herbivorous throughout their lives (Moll 1989; Polisar 1992) and therefore may not undergo ontogenetic dietary shifts, at least not in terms of carnivory to herbivory. Moll (1989) and Polisar (1992) noted that D. mawii feed on a variety of vegetation when it is opportunistically available. Moll (1989) quantified the stomach contents of D. mawii adults and juveniles during the dry season from several water bodies throughout Belize. Although statistical comparisons were not made, Moll (1989) suggested similarities of diets between age-groups and noted minor differences among habitats (e.g., river, lagoon, or estuary). Anecdotal evidence by Polisar (1992) emphasized habitat as an important factor in diet selection and additionally observed seasonal differences in feeding behavior. In river systems during the dry season, emergent vegetation along the bank and windfall plant matter were readily consumed; during the rainy season, when floodwaters had inundated riparian forests, terrestrial plant species were also consumed. Dermatemys mawii in lagoons ate submergent aquatic plants during the dry season, when low turbidity contributed to increases in light and plant growth; in the rainy season, lagoon turtles used increases in water level to explore additional food sources such as emergent and riparian vegetation in flooded areas (Polisar 1992).
Limited studies (Moll 1989; Polisar 1992; Vogt et al. 2011) have contributed to a general, qualitative understanding of D. mawii diet. However, no studies have quantitatively assessed dietary differences based on biotic and abiotic factors. Given the diversity of plant functional groups (e.g., terrestrial, riparian, and aquatic), plant species, and plant parts available to D. mawii, a more robust understanding of differences in diet is important and can contribute to the conservation of this critically endangered species. The purpose of our study was to generate a more complete understanding of food resources used by D. mawii. In this study, we describe the diet of wild D. mawii and quantitatively compare the diets of turtles from different habitats, of different maturity classes, and of different sexes.
METHODS
Sample Collection and Preparation. — Stomach contents of D. mawii were collected from turtles that were butchered from October 1989 through June 1990 and September through November 1990 in Belize. Turtles were legally collected by local hunters for personal consumption from the Belize River, Batlass Creek, Rio Bravo, Mussel Creek, Hector Creek Lagoon, Spanish Creek, and White Water Lagoon. Sexual maturity of females was determined by the presence of oviductal eggs, corpora lutea, corpora albicans, and/or enlarged vitellogenic follicles (Polisar 1996); sexual maturity of males was based on secondary sexual characteristics such as yellow to orange dorsal head coloration and enlarged tail size (Polisar 1996). Straight-line carapace length (SCL) was also measured. Stomach contents were stored in ethanol until they could be analyzed. Identification and quantification occurred within a year of collection. Each diet item was identified to the lowest possible taxon, and the volume of each taxon was quantified by water displacement. Small pieces of stem and seedpods could not be reliably distinguished, so they were combined. Taxa less than 0.1 ml in volume were considered trace amounts.
The aquatic systems from which turtles were harvested can be divided into 2 distinct systems: rivers and lagoons. River habitats are fast-moving and turbid and generally devoid of submergent aquatic vegetation; lagoon habitats have slow-moving water with submergent aquatic vegetation. Rivers included in this study were the Upper, Middle, and Lower Belize River, and lagoons consisted of Batlass Creek, Cox Lagoon, Hector Creek Lagoon, Spanish Creek, and White Water Lagoon (a map of these collection sites along with further descriptions of individual waterways can be found in Polisar 1996). Some turtles (n = 18) were kept in captivity between the time they were harvested from the wild to the time they were slaughtered. The stomach contents of turtles that were kept for more than 1 d (n = 11) were removed from the analyses because their diet could have been supplemented and/or digestion could have altered diet identification and quantification.
Quantifying Stomach Contents. — We assessed stomach contents by percent volume and percent frequency of occurrence. We also used an index of relative importance (IRI) that had been adjusted for herbivory (Bjorndal et al. 1997) to quantitatively evaluate stomach contents using the following formula:
where F is percent frequency of occurrence and V is percent volume. Individually, percent frequency of occurrence and percent volume can lead to the misinterpretation of the importance of a food item. The IRI accounts for the volume and frequency in which each item occurs and allows for a relative comparison of importance. IRIs were calculated for stomach contents by item (e.g., leaf, flower, and rock) to get a general understanding of what is being consumed. We then calculated IRIs for food items by the lowest taxa to better understand what parts of which species were being consumed to account for potential functional differences in taxa (e.g., terrestrial species vs. aquatic species). All turtles were assessed together and were then compared by habitat, maturity class, and sex in separate analyses.
Because the IRI of each food item is a rank, IRI values cannot be directly compared between groups (e.g., rivers vs. lagoons). Therefore, we used Spearman rank correlation coefficients to test the null hypothesis that there was no correlation in the rankings of items (i.e., there were differences) when comparing groups (Martin et al. 1996; Hart et al. 2002). Rankings that are highly correlated are significant and indicate there are no differences in the IRIs between those 2 groups (i.e., rankings that were significantly correlated were similar across groups). When the Spearman rank tests determined there were differences between groups, we used a Kruskal-Wallis test to determine where differences existed in individual percent volume consumed (Bjorndal et al. 1997). All data analyses were completed with the software R 3.1.2 (R Development Core Team 2014).
RESULTS
Overall Diet Composition. — A total of 66 turtles were used in this study. Turtles from rivers (n = 49) ranged in size from 32.0 to 44.2 cm SCL and consisted of adults (n = 27), juveniles (n = 20), and unknown maturity (n = 2). Adult female turtles (n = 11) and adult male turtles (n = 16) ranged in size from 34.2 to 44.2 cm SCL and 36.0 to 41.0 cm SCL, respectively. Juveniles ranged in size from 32.0 to 38.4 cm SCL. Turtles from lagoons (n = 15) ranged in size from 32.3 to 42.3 cm SCL and consisted of adults (n = 3), juveniles (n = 11), and 1 turtle of unknown maturity. Only 1 female and 2 males represented the adult population from lagoons. Two turtles were of unknown origin, maturity, and sex.
For all turtles combined, stomach content composition was varied and included multiple plant parts (leaves, flowers, stems/pods, seeds, and fruit) from 6 plant families, 1 algal family, invertebrates, and rocks (Tables 1 and 2). Leaves made up a majority of the diet by a large margin for percent frequency, percent volume, and index of relative importance (IRI). Other important food items (in decreasing order of importance) were flowers, stems/pods, and algae (Nitella sp.). Nitella sp. is a plant-like macro-algae that forms thick mats along the bottoms of lakes and swamps. Flowers and stems/pods were found frequently (46.3% and 34.3% of stomachs, respectively), but stems/ pods accounted for less than 5% of the total volume of combined stomachs. Nitella sp. was found only in the stomachs of juveniles (n = 6, frequency 9%), and all except 1 of those turtles were from lagoon habitats. When present, Nitella sp. occurred in large volumes (10.5–43 ml). All other items had an IRI value of less than 1% and included fruits, seeds, and nonplant stomach contents, such as small rocks and invertebrates. Fruits consisted only of figs from Ficus sp. (family Moraceae) and were found only in the stomachs of turtles from river habitats; seeds included beans (family Fabaceae) and unidentified species.
Nonplant stomach contents included small rocks and invertebrates (flatworms and trace amounts of annelid worms and insects, presumably consumed incidentally) and were found occasionally (ranging between 3.0% and 10.4% of stomachs). Each item contributed to less than 1% of the total volume of combined stomachs, but all nonplant categories were present in males, females, adults, and juveniles. Individual rocks ranged in size (volume by displacement) from less than 0.1 to 2 ml. Insects consisted only of beetle parts, such as the elytron (wing covering); flatworms ranged in volume in individual stomachs from less than 0.1 to 1 ml. We could not determine the species of flatworms or whether they were symbiotic or consumed (either intentionally or incidentally).
Diet items were also evaluated by taxa and included species from the following plant families: Fabaceae, Malvaceae, Moraceae, Poaceae, Typhaceae, Hydrocharitaceae, and Characeae (algal) (Table 2). Plant parts of species within the Fabaceae (legume) family were the most important by a large margin for percent frequency, percent volume, and IRI. The leaves of Inga edulis (within the Fabaceae family; locally “bri-bri”) were the most consumed item. Inga edulis leaves were found in 73.1% of stomachs and accounted for 48.5% of the total volume of all stomach contents combined. A relatively small volume of leaves (10.5% of total volume of combined stomach contents) could not be identified but were considered the second most important food item in terms of percent frequency, percent volume, and IRI; unidentified leaves were found in 62.7% of stomachs. The flowers of I. edulis were also found frequently (38.8% of stomachs), but the range of volume within each stomach was large (0.1–24.5 ml). Fruits (i.e., figs) from a Ficus sp. were also present. Leaves from Paspalum sp. and an unidentified grass in the Poaceae family were regularly found in stomachs (19.4%) but always in quantities of less than 1 ml and often in trace amounts (< 0.1 ml). However, grass leaves were found in males and females, juveniles and adults, and in both lagoons and rivers, making them the second (after Fabaceae) most ubiquitous plant family present in stomachs in terms of frequency for sex, maturity class, and habitat.
Diet Comparison by Habitat: Rivers Versus Lagoons. — Leaves are the main food item for D. mawii from both rivers and lagoons (Table 1). However, there were significant differences in the rankings of food items (IRIs, percent frequencies, and percent volumes) between river and lagoon habitats (Spearman rank correlations; Table 3). Comparisons were made for each of the 7 stomach items by volume; fruits, rocks, and insects were not present in lagoons and were not included in the analyses (Kruskal-Wallis tests, Bonferroni adjusted alpha of 0.007). Consumed volumes of Nitella sp. were significantly greater (p = 0.004) in turtles from lagoons. Although turtle stomachs from lagoons lacked fruit, insects, and rocks, a Hutcheson t-test comparing the Shannon diversity indices between rivers and lagoons indicated no significant difference in diversity between habitats (Hrivers = 1.68, Hlagoons = 1.64, p = 0.819). Turtles from both rivers and lagoons had consumed equally diverse stomach content categories.
When comparing diet items by taxa (Table 2), diets of turtles from rivers and lagoons were not correlated (IRIs, percent volume, and percent frequency), indicating that there were significant differences in rankings of food items by taxa between the two habitats (Spearman rank correlations; Table 4). In the stomachs of turtles from rivers, leaves of I. edulis were the most important food item by percent frequency, percent volume, and IRI. A majority of the stomachs also had flowers from I. edulis (53.1%) and leaves from unidentified species of plants (63.3%). In both habitats, a large majority of the unidentified species of leaves was terrestrial. Stem pieces and/or seedpods (taxa unknown) were also common in river turtle stomachs (frequency 32.7%) and ranged in volume from trace amounts to 14.5 ml. However, volumes of stems/pods in individual stomachs were often less than 1 ml and may have been from incidental ingestion during leaf consumption; all stomachs that contained stems/pods also contained much larger quantities of leaves. Items found only in the stomachs of river turtles included I. edulis flowers and stems, stems from Pithecellobium sp. (family Fabaceae), bean seeds from unidentified species in the Fabaceae family, fruits (i.e., figs) from unidentified species of the Moraceae family, and rocks. In turtle diets from lagoon habitats, unidentified species of leaves had the highest percent frequency and IRI. The next most important food items by taxa were Nitella algae and leaves from I. edulis. Items unique to stomachs of lagoon turtles were leaves from Najas sp. (consumed by 1 juvenile female), cattail leaves (family Typhaceae; consumed by 1 juvenile female), and flowers of P. aquaticus (consumed in large quantities by a juvenile male [51.5 ml] and a juvenile female [65.0 ml]). Grass blades (family Poaceae) and invertebrates (flatworms, annelids, and beetles) were also present in turtle stomachs from both habitats. Grasses and invertebrates each contributed less than 1% of the total IRI for diets of turtles from both rivers and lagoons. A Hutcheson t-test between the Shannon diversity indices indicated no significant difference in diversity of turtle diets between habitats (Hrivers = 2.12, Hlagoons = 2.17, p = 0.729). Diets of turtles from rivers and lagoons were equally diverse in terms of plant parts consumed by taxa.
Diet Comparison by Maturity Class: River Adults Versus River Juveniles. — Quantitative comparisons of diet between adults and juveniles were conducted only for turtles from river habitats due to low sample size in lagoon systems (adult, n = 3; juvenile, n = 11). In comparisons of river adults and river juveniles, leaves were the most important food item for both maturity classes (Table 5). Adult diets lacked algae, and juvenile diets lacked fruits and seeds. There were no differences in IRI rankings of food items between adults and juveniles, but there were differences in percent frequency and percent volume between the maturity classes (Spearman rank correlations; Table 3). A Hutcheson's t-test comparing the Shannon diversity indices indicated no significant difference in diversity between maturity classes (Hadults = 1.66, Hjuveniles = 1.56, p = 0.588). Additional comparisons were made given that size rather than maturity class is also important in turtle digestive studies. Spearman rank correlations were used to examine the relationships between SCL and individual percent volume of food items consumed. There were significant but weak correlations between SCL and flowers (r = 0.428, p = 0.003), stems/pods (r = –0.344, p = 0.019), and rocks (r = –0.304, p = 0.040). As SCL increased, individual flower consumption increased while stem/pod and rock consumption decreased. All other correlations between percent volume of stomach item and SCL were nonsignificant (≥ 0.05).
Food items by taxa were also evaluated, and there were no differences in any rankings (Spearman rank correlation; Table 4), indicating that both maturity classes used the same taxa of plants with similar rankings of importance. Leaves of I. edulis were the most important food item by percent frequency, percent volume, and IRI. Additionally, a majority of adult stomachs (70.4%) also had flowers from I. edulis, which flowers in the dry season (Table 6).
Diet Comparison by Maturity Class: Lagoon Adults Versus Lagoon Juveniles. — Sample size was too small to quantitatively compare stomach contents of adult lagoon turtles and juvenile lagoon turtles, but similar to river habitats, juvenile stomachs lacked seeds, and adult stomachs lacked algae. Juveniles from lagoons also relied heavily on algae and flowers.
Diet Comparison by Sex: River Males Versus River Females. — Stomach contents of adults were considered only when analyzing males and females due to differences based on maturity class. There were no differences between males (n = 16) and females (n = 11) for diet items by food category (Spearman rank correlation; Table 3) or by taxa (Spearman rank correlation; Table 4). Leaves were the most important food item for both sexes (Table 5). When broken down by taxa, the leaves of I. edulis were the most important food item by percent frequency, percent volume, and IRI (Table 6); flowers of I. edulis were the second most important food item. Female stomachs lacked seeds, invertebrates, and rocks (Table 5).
DISCUSSION
Dermatemys mawii are primarily folivorous. Leaves of the riparian tree, I. edulis, constituted a bulk of the diet. Other important dietary items found in this study were also from riparian species in which leaves (or other plant parts) were shed into lagoons or rivers. Only 2 fully aquatic species (Nitella sp. and Najas sp.) were found in stomachs infrequently (1.5%–9.0% frequency). Emergent aquatic species from Typhaceae (unidentified species) and Poaceae (Paspalum sp.) were also present. It is unknown if D. mawii were consuming terrestrial species in large quantities due to specific nutrients and/or availability, and this requires further investigation. However, terrestrial plant species have more structural support and may well be harder to digest than aquatic species. As leaves from terrestrial species are the main component of D. mawii's diet, their digestion may require slower passage or specialized gut microflora that differ from other freshwater herbivorous turtles that rely mainly on leaves from aquatic species.
The only other study that has quantified the diet of D. mawii is Moll (1989). Turtles in his study were collected in Belize from the Belize River (adult, n = 82; juvenile, n = 28), Progresso Lagoon (adult, n = 58; juvenile, n = 26), and Rio Grande Estuary (adult, n = 24; juvenile, n = 16) between January and April. Despite temporal and spatial overlap in the collection of turtles, there were major qualitative differences between the stomach contents in the present study compared with those in Moll's study. In Moll (1989), almost all of the food items were from aquatic or semiaquatic plants in comparison to the present study, in which a majority of the stomach contents was from terrestrial plants. Only 3 plant species were identified in the stomachs of turtles in both studies: Najas sp., Paspalum sp., and Ficus sp. Najas sp. was abundant and common in Moll's data set and found in the stomachs of adult and juvenile turtles from both rivers and lagoons; in the current study, Najas sp. was found in the stomach of only 1 juvenile from a lagoon. Paspalum peniculatum packed the stomachs and intestines of almost all turtles in Moll's study regardless of habitat or maturity class and was the most important food item consumed by D. mawii. In the present study, only 19.4% of D. mawii consumed the emergent, semiaquatic grasses of Paspalum sp., and individual consumption was so low that these grasses were generally a minor contribution to overall diet. Polisar (1992) suggested that seasonal timing of collection (January–April) influenced Moll's findings and that riparian species, such as I. edulis, may be a more important contribution to diet throughout the year, as was demonstrated in the present study, which had a longer collection period that spanned 10 mo. These results suggest that availability of food sources may be important in foraging strategies among D. mawii; however, further investigation is required, including examining temporal and spatial abundance of plant and algal species.
There were differences in diets between D. mawii from rivers and those from lagoons. River turtles relied more heavily on riparian species such as Inga edulis and Ficus sp., while lagoon turtles consumed more aquatic (Nitella sp. and Najas sp.) and emergent (Typhaceae) vegetation generally found only in slow-moving, freshwater systems. The presence and absence of specific taxa in the stomachs of turtles from lagoons versus rivers suggest that D. mawii may be feeding opportunistically on the plant matter available within each habitat and corroborates Moll's (1989) findings that describe consumption based on food availability. Moll also describes occasional consumption of windfall food sources such as fig fruits (rivers) and mangrove leaves (estuary section of the Rio Grande River). Although some of the plant species consumed are generally dictated by the habitat, the present study emphasizes the importance of windfall food sources from riparian species as a main component in the diets of D. mawii regardless of habitat.
Stomach content analyses may be biased because soft-bodied animals may be underrepresented, while fibrous and chitinous/keratinous materials (e.g., insect exoskeletons) may be overrepresented. However, quantitative comparisons made between habitat, sex, and age would be equally biased. An additional limitation of the present study would be a lack of seasonal comparisons due to the small number of samples collected during the rainy season (n = 6) compared with the dry season (n = 58). However, plant species from the study region generally flower and fruit during the dry season (Janzen 1967; Pavelka and Knopff 2004), as represented in the diversity of plant parts present in turtle stomachs (Tables 1 and 2).
Despite these limitations, the results of this study are still important to conservation efforts for D. mawii. Head-starting has been proposed as a conservation strategy for D. mawii (Vogt et al. 2011; Rainwater et al. 2012; Bishop et al. 2021). Head-started individuals are hatched and reared in captivity until they reach a certain size (variable by species), at which point they are released into the wild, where they presumably perform within their ecological niche and produce offspring. Identifying areas abundant with the specific taxa referenced in this study could be helpful in determining appropriate release habitat for head-started individuals. Depending on the size and age of released turtles, additional knowledge of wild food sources at earlier stages of life (e.g., hatchlings) would also be beneficial and warrant further study.
Dermatemys mawii in this study relied heavily on terrestrial plants that shed leaves, flowers, and fruit into the water despite the fact that D. mawii are predominantly aquatic, with only females leaving the water to nest (Vogt et al. 2011). Turtles in this study, regardless of sex or maturity class, probably consumed what was readily available in the habitat. Turtles from lagoons had a greater dependence on algae than river turtles, but turtles from both environments relied heavily on the leaves of I. edulis, a riparian tree species. Fruits of figs (Ficus sp.) and flowers of I. edulis were consumed only by turtles from rivers but were important components of their diet and may be consumed while seasonally available. In conclusion, D. mawii is primarily folivorous and relies heavily on the leaves of riparian plant species that are shed into their aquatic habitats, with site-specific dietary patterns influenced by dynamic interactions between changing water levels (that provide varying access to the aquatic-terrestrial food source gradients) and plant phenology that manifests along those same gradients.
Contributor Notes
†Deceased
Handling Editor: Vivian P. Páez
