Feeding preferences and nutritional niche of wild water buffalo (Bubalus arnee) in Koshi Tappu Wildlife Reserve, Nepal

Abstract The nutritional characteristics of food resources play an important role in the foraging behavior of animals and can provide information valuable to their conservation and management. We examined the nutritional ecology of wild water buffalo (Bubalus arnee; hereafter “buffalo”) in the Koshi Tappu Wildlife Reserve of Nepal during autumn using a multidimensional nutritional niche framework. We identified 54 plant species as being foraged by buffalo. We found that buffalo consumed graminoids and forbs 2–3 times more frequently than browse items. Proximate analyses of the 16 most frequently foraged plants indicated that buffalo diets were highest in carbohydrate (40.41% ± 1.82%) followed by crude protein (10.52% ± 0.93%) and crude fat (1.68% ± 0.23%). The estimated macronutrient balance (i.e., realized nutrient niche) of the buffalo diet (20.5% protein: 72.8% carbohydrate: 6.7% lipid) was not significantly different than the average balance of all analyzed food items based on 95% confidence regions. Our study suggests that buffalo are likely macronutrient specialists, yet may be generalists in the sense that they feed on a wide range of food items to achieve a nutrient balance similar to that available in forage items. However, the four most frequently consumed items tended to be higher in protein energy than less frequently consumed foods, suggesting some preference for higher protein forage relative to relatively abundant carbohydrates. Although limited in scope, our study provides important information on the nutritional ecology of buffalo, which may be useful for the conservation and management of this endangered species.

Nutrient balancing is the phenomenon by which animals homeostatically regulate their intake of foods to maintain a relatively consistent nutrient intake in the face of sometimes considerable variation in the nutritional composition of food items consumed (Simpson & Raubenheimer, 2012). Thus, intraspecific variation in diet composition due to environmental differences may not equate to significant differences in overall dietary nutrient composition. This phenomenon has been observed in geographically distinct populations of mountain gorillas, which regulated the composition of nutrients in their diets despite consuming different forage items (Gorilla beringei; Rothman, Plumptre, Dierenfeld, & Pell, 2007). Other species can tolerate widely different dietary nutrient compositions across their range, including omnivorous wild boars (Sus scrofa) and brown bears (Ursus arctos; Senior, Grueber, Machovsky-Capuska, Coogan, Raubenheimer, Stenhouse, Coops, & Nielsen, 2018). Therefore, considerable insight into an animal's feeding strategies can be gained by examining diet at the level of both foods and nutrients in relation to food availability.
Researchers have moved beyond the traditional categories of dietary specialization (i.e., generalist versus specialist) in terms of the range of foods consumed to also encompass the nutritional and other characteristics (e.g., structural components) of foods in a multidimensional nutritional niche (Coogan, Raubenheimer, Zantis, & Machovsky-Capuska, 2018;. At one level, an organism's degree of specialization can be described in terms of the nutritional composition of foods consumed, where an animal with a diet consisting of foods varying broadly in nutritional composition can be considered a food composition generalist. At another level, the nutritional composition of a population's overall diet can be used to assess the realized nutritional niche of that population. The range of realized nutritional niches of a species can be used to infer that species fundamental nutritional niche. Species with a wide range of realized niches (i.e., large fundamental nutritional niche) may be considered nutrient generalists. Conversely, species with a narrow range of realized niches (i.e., small fundamental nutrient niche) can be considered nutrient specialists. Finally, the multidimensional nutritional niche framework also considers the physical and non-nutritional properties of foods, such that an animal with the ability to consume a variety of foods that vary structurally can be considered a food exploitation generalist.
The wild water buffalo (Bubalus arnee; hereafter referred to as "buffalo") is a species that has been the subject of intensive management over the previous 60 years, and is listed by the IUCN as Endangered (Kaul, Williams, Rithe, Steinmetz, & Mishra, 2019).
Throughout their current range, buffalo select low-lying alluvial grassland habitats (Heinen & Paudel, 2015). While little research has explicitly studied the diet of buffalo, feeding observations suggest they are predominantly grazers, but have been observed foraging on forbs and browse, especially new growth (Choudhury, 2014).
In Thailand, 45% of buffalo diet was composed of 3 grass species (Chaiyarat, 2002). A population of buffalo in Assam, India, has regularly been observed foraging on water hyacinth (Eichhornia crasspies), an invasive forb that has become common in the freshwater systems of the area (Choudhury, 2014). Buffalo are also known to raid anthropogenic crops such as rice, sugarcane, and jute from agricultural lands on the fringes of their home range, which has led to buffalo-human conflicts in and around protected areas (Choudhury, 2014).
In Nepal, buffalo were restricted to the Koshi Tappu Wildlife Reserve (KTWR) until 2017 when a second population was established in Chitwan National Park. To date, research into the ecology of the KTWR buffalo has focused on population growth and genetic integrity. Censuses of the KTWR population have been conducted sporadically since 1976. These censuses have described an average annual population growth rate of 3.3% throughout that time (Dahmer, 1978;Heinen, 1993;Heinen & Kandel, 2006;Heinen & Singh, 2001;Khatri, Shah, & Mishra, 2012). This growth rate is consistent with population growth rates of other large, long-lived ungulates that have adequate habitat and are not subject to predation pressure (Clutton-Brock, 1989;Heinen & Paudel, 2015). The most recent census conducted in 2018 reported 441 individuals in the KTWR's buffalo population (KTWR, 2018). Females are typically found in herds of 13-17, and bachelor herds have been observed being comprised of 9-12 individuals (Heinen, 1993).
In this paper, we sought to further the understanding of the nutritional ecology of buffalo in the KTWR to facilitate its conservation and management. We used multidimensional nutritional niche concepts to evaluate the foraging choices of free-ranging buffalo during the autumn in the KTWR. First, we identified plant species that were foraged by buffalo. Then, to understand aspects of the food exploitation level of their nutritional niche, we evaluated the relative frequency (RF) of graminoids, forbs, and browse foraged by buffalo.
Next, we explored the nutrient balance of foraged species to gain insight into the degree of nutrient specialization and the realized nutrient niche for buffalo in the KTWR. We predicted that the realized nutrient niche for buffalo would be highest in the proportion of carbohydrate energy, moderate in protein, and with the lowest proportion for lipid, in keeping with the dietary nutrient balance of other Nepalese herbivores (Aryal, Brunton, et al., 2015;Koirala et al., 2019).

| Study area
The KTWR was established in 1976 to preserve the last Nepalese population of buffalo and act as a migratory bird sanctuary (Heinen & Paudel, 2015). The KTWR lies on the floodplains of the Saptakoshi River in the South-East Terai region of Nepal (Sah, 1997). The reserve has subtropical climate, with an elevation ranging between 75 and 100 m above sea level. Nepal has four climatic seasons, including spring (March-May), summer (June-August), autumn (September-November), and winter (December-February). The reserve covers a 175-km 2 core area with a 173 km 2 buffer zone. The KTWR incorporates two municipalities of the Saptari district, two municipality of the Udayapur district, and one municipality and one rural municipality of Sunsari district, with a combined population of 84,423 people among 14,865 households (KTWR, 2018). The KTWR is mostly comprised of alluvial grasslands (56%) and large sand/gravel deposits (22%) with some forest (1%), lakes and ponds (0.01%), marshes and swamps (6%), rivers and streams (10%), and, in the buffer zone, agricultural land (5%) (Chettri, Uddin, Chaudhary, & Sharma, 2013). In 2009, a botanical survey described 670 species of vascular plants in the reserve (Siwakoti, 2009). Natural predators of buffalo (e.g., tigers, Panthera tigris; leopards, Panthera pardus; and dholes, Cuon alpinus) have been extirpated from the KTWR for at least 40 years (Heinen & Paudel, 2015). Likewise, large mammalian herbivores such as gaur (Bos gaurus) and blue bull (Boselaphus tragocamelus) have declined in numbers and are now rare in the KTWR.

| Field methods
We conducted field surveys during the autumn (November 2017) following the hot, wet summer monsoon season and avoiding the cool, dry winter season. During the winter vegetation, dies during the dry period and summer monsoon floods limit buffalo to foraging on islands and in croplands (Chettri et al., 2013). Therefore, we conducted surveys for buffalo foraging during November, when the KTWR buffalo have the greatest access to forage and travel widely throughout their study area.
We identified and sampled vegetation in bison foraging plots following a modified version of Ngoti (2017). We established 50 foraging plots (5 m × 5 m) throughout the KTWR in areas where buffalo are typically observed, and where fresh buffalo dung and signs of foraging were present ( Figure 1). We located and identified buffalo dung with the help of a local KTWR guides who had knowledge of buffalo ecology and behavior. We were careful to visually identify dung based on its physical characteristics to prevent misidentification of domestic cattle dung as belonging to buffalo. In addition, rarity of other large herbivores in the park facilitated buffalo dung identification. We established the square 5 m × 5 m foraging plots using a measuring tape, where the perimeters were set using wooden pegs in each of the four corners with plastic ropes delineating each of the four sides of the plot. Once the plot was set, we recorded plot-level field data, such as plot ID number, date, latitude, longitude, habitat type, presence or absence of cattle, and existing F I G U R E 1 Koshi Tappu Wildlife Reserve, Nepal, and the distribution of foraging plots established in November 2017. Foraging plots are numbered from 1 to 50 plant species. Plants that were grazed, browsed, or debarked were carefully identified and recorded during data collection following a modified protocol from Ngoti (2017). We did not attempt to quantify the amount or proportion of foraging on different species in plots, rather we simply identified whether a species was foraged or not and that species functional foraging group (i.e., forbs, graminoids, or browse). Plants were identified with the help of our local guide as well as local residents familiar with the flora in the park.
We collected representative samples of foraged plants of the same species from nearby plants within the foraging plots that were not foraged. We therefore assumed that the plants we collected for analysis had the same nutritional properties as foraged plants.
We clipped plant samples for collection in a manner that mimics the way the plant was foraged by buffalo at that site. In general, we clipped grasses above the organic debris covering the ground, and leaves of browse items were clipped up to approximately 1.8 m off the ground. We sealed samples of each plant species in airtight plastic bags and transported for nutrient content analysis. In field, plant sample identification was verified at the Department of Plant Resources herbarium in Kathmandu, Nepal.

| Diet and nutritional analysis
After collecting plants, and before conducting nutritional analysis, we estimated the buffalo's diet by calculating the frequency and relative frequency (RF) of all foraged species using the following equations from Fracker and Brischle (1944): Relative frequency (% ) = (Frequency of species x/Total frequency of each foraged species) × 100 TA B L E 1 Plant species foraged by wild water buffalo (n = 54) and the relative frequency (RF %) that foraging was observed for each species Note: The 16 most frequently foraged plants are shaded in grey and were used in analysis of nutrient composition of wild water buffalo diets. The functional forage group (FG) for each species is also listed as graminoid (G), forb (F), or browse (B).
Due to research constraints, we selected 16 plant species for nutritional analysis that were foraged in at least 5 plots (i.e., RF > 2%).  (Pearson, 1976). Raubenheimer, 2011). To evaluate nutrient balance, we first converted macronutrients to units of metabolizable energy using conversion factors of 9 kcal/g for lipids and 4 kcal/g for carbohydrates and proteins (Coogan, Raubenheimer, Stenhouse, & Nielsen, 2014;Merrill & Watt, 1973). Following conversion, the metabolizable energy values for each macronutrient in individual food items were summed together and then expressed as a percentage of the sum of total metabolizable energy. Food items were plotted as Cartesian points within RMTs based on their metabolizable energy content.
We drew convex hull polygons (Wijeweera & Kodituwakku, 2018) around the foraged plants plotted in the RMT to visually evaluate the range (or breadth) of macronutrient compositions in food items of buffalo, and thus the degree of macronutrient generalism or specialization exhibited by buffalo based on their foraging preferences.
We then weighted the macronutrient proportions of forage items by their RF to estimate the buffalo's November diet, their realized macronutrient niche for that period (Machovsky-Capuska et al., 2016).
Our RMTs revealed that the 16 most frequently foraged species by buffalo ranged in percent metabolizable energy from 58% to 88% for carbohydrate, 7% to 31% for protein, and 1% to 16% for lipids ( Figure 4). The four species that accounted for greater than 25% of the foraged plants occupied a nutrient space that was relatively higher in protein concentration and lower in carbohydrates than the 12 other species. However, the mean nutrient composition of the 16 most frequently consumed forage plants (19.4% protein: 73.7% carbohydrate: 6.9% lipid) was not significantly different than the estimated nutrient balance (20.5% protein: 72.8% carbohydrate: 6.7% lipid) of their weighted diet (i.e., the estimated dietary nutrient proportions of forage items weighted by the RF of food items consumed) based on the 95% confidence region (Figure 4). TA B L E 2 Results of the proximate analysis describing the nutritional composition as a percentage of the 16 most frequently foraged plants of wild water buffalo F I G U R E 2 Mean content of plant components in species foraged by wild water buffalo. The mean content of protein, ash, and moisture is not significantly different as shown by "A," the mean content of the crude fat is significantly lower than other nutrients shown by "B," and the mean content of carbohydrate is significantly higher shown by "C."

F I G U R E 3
The mean relative frequency at which graminoids, forbs, and browse species were foraged by wild water buffalo (Bulbalus arnee) in the Koshi Tappu Wildlife Reserve may have foraged plants within survey plots. However, populations of other large ungulate herbivores in the KTWR are sparse (Khatri, Shah, Shah, & Mishra, 2010) and the guides employed to identify signs of buffalo have expertise in distinguishing buffalo sign from other local herbivores. Furthermore, we did not look at metabolizable energy from digestible fiber (e.g., Aryal, Brunton, et al., 2015), which has been shown to influence food selection in other herbivores . Yet, given the consistency with This result conflicts with the results of Chaiyarat (2002), the only other study that explicitly tried to describe the diet of buffalo, but in Thailand. Chaiyarat (2002) found that 90% of buffalo diet was composed of grasses based on fecal analysis in Huai Kha Khaeng Wildlife Sanctuary, Thailand. However, that diet was analyzed using microhistology, which has been critiqued as showing a positive bias toward grasses that tend to pass through the digestive system intact and are therefore more readily identified in fecal analyses than forb or browse species (Varva & Holecheck, 1980).
Additionally, the author did acknowledge that forbs and shrubs were present in buffalo diets (Chaiyarat, 2002). Our study is the first to describe forbs as a dominant part of buffalo diets. Modern The black convex hull polygon represents the mixture space of food items consumed by wild water buffalo. The red polygon highlights the nutrient mixture space of the four species that made up greater than 25% of foraged plants. (b) A closer look at the region of nutrient space occupied by the most commonly foraged plant species of wild water buffalo. The blue dot (with accompanying 95% confidence region outlined in blue) is the mean nutrient proportion of all forage items and represents what the nutrient composition of the diet would be if wild water buffalo foraged randomly on these species. The green dot is an estimate that describes the realized nutrient proportions of the wild water buffalo's diet as estimated by weighting the nutrient proportions of food items by the relative frequency (RF) they were foraged upon. Because both the mean nutrient proportion of food items and the weighted diet nutrient proportions lie within the 95% confidence region, we can infer that there is no significant difference between them Despite the limitations of this study, our research contributes important knowledge on the feeding preferences and nutritional con-  in Chitwan National Park (CNP), which still houses tigers. Theory predicts that buffalo will spend less time at a particular foraging station, more time being vigilant, and an overall lower quality diet while foraging on a "landscape of fear" in CNP (Hernandez & Laundre, 2005).

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
A complete list of foraged species with relative frequency statistics and macronutrient content of the 16 most frequently foraged species can be retrieved via Dryad: https://doi.org/10.5061/dryad. vt4b8 gtp0.