The largest felid in the world, the tiger Panthera tigris, is threatened due to habitat destruction, prey depletion and poaching (Nowell & Jackson 1996). Of these threats, the poaching of tiger prey is considered the most insidious because of its potential to not only impact tiger populations by increasing rates of human-tiger conflict, but also indirectly affect them by decreasing their prey base and presumably, depressing their reproduction (Karanth & Stith 1999). Despite considerable investment and public support, conservation efforts aimed at mitigating these threats have experienced limited success, and tiger numbers and distributions have continued to contract (Dinerstein et al. 2007). As human populations in tiger range states continue to grow, so too will the pressure they exert on tiger populations and habitats.

Accurate knowledge of a species' dietary habits is one of several prerequisites for effective conservation, and it is fundamental to conservation initiatives such as habitat prioritization, protection and restoration. Given ample food and refuge from human persecution, tigers are capable of surviving in almost any vegetative association (Sunquist et al. 1999); tiger range encompasses a large variety of ecosystems and tiger distribution and density are closely tied to that of its primary prey (Miquelle et al. 1999, Karanth et al. 2004). Consequently, understanding the tiger's trophic requirements is essential to predicting its response to human-mediated environmental change (Sunquist et al. 1999).

Given its importance to the conservation of the species, it is not surprising that tiger diet has received considerable attention. Studies of tiger diet have been conducted throughout much of the range of the species, including India (Karanth & Sunquist 1995, Chundawat et al. 1999, Biswas & Sankar 2002, Sankar & Johnsingh 2002, Bagchi et al. 2003, Reddy et al. 2004, Andheria et al. 2007), Nepal (Seidensticker 1976a, McDougal 1977, Sunquist 1981, Johnsingh 1992, Seidensticker & McDougal 1993, Støen & Wegge 1996, Wegge et al. 2009), Bangladesh (Khan 2004), Thailand (Rabinowitz 1989), Bhutan (Wang & MacDonald 2009) and Russia (Miquelle et al. 1996). These studies indicate that tigers generally prey upon 8-15 species, and while they occasionally consume prey weighing up to 1,000 kg, the majority of tiger diet consists of small prey (i.e. ≤ 20 kg; e.g. Rabinowitz 1989) and medium-sized cervids (i.e. of 50-200 kg; e.g. Karanth & Sunquist 1995).

With few exceptions (e.g. Karanth & Sunquist 1995), most studies of tiger diet have been of limited spatio-temporal scale, potentially hindering our understanding of variation in the tiger's diet (Table 1). Stochastic events such as drought and outbreaks of disease (McDougal 1977) result in fluctuations in prey abundance and vulnerability (Schaller 1967, Johnsingh 1983, Dave & Jhala 2011); short term studies that encompass a single season or year may not detect variation in tiger diet resulting from these fluctuations. In addition, tigers exhibit intrasexual territoriality and maintain large home ranges (Sunquist 1981), and small study areas are unlikely to encompass the home ranges of more than a few tigers. In combination, the short duration and small study areas of most investigations of tiger diet may inhibit our understanding of tiger diet.

To better understand the spatio-temporal dynamics of tiger diet in relation to prey abundance and the effect of study duration and study area size (hereafter extent) on estimates of the tiger diet, we used heretofore unpublished data on tiger diet in combination with two previous studies of tiger diet conducted in Royal Chitwan National Park, Nepal (McDougal 1977, Sunquist 1981).

Methods

Analysis

We analyzed the contribution of each prey species to tiger diet and expressed it as percent occurrence (number of occurrences of each prey/total number of occurrences × 100) and percent biomass contribution (biomass of each prey type consumed/total biomass consumed × 100). We estimated percent occurrence to allow comparison with previous studies. However, small prey species, having proportionately more indigestible material (e.g. hair and bones) per unit weight than large prey species, produce more scats per volume consumed, resulting in overestimates of their relative contribution to tiger diet. Consequently, the percent occurrence is not an accurate indication of the contribution of a particular prey to the diet when there is a large variation in prey size (Floyd et al. 1978, Ackerman et al. 1984). To correct this overestimation, we applied the regression equation developed for cougars Puma concolor by Ackerman et al. (1984):

where Y is the weight of prey consumed/scat produced, and X is the live weight of the prey. This equation relates prey weight to the number of scats produced/unit weight consumed. We calculated Y for each species and multiplied it by the number of occurrences of the species to estimate the relative biomass of each prey type consumed. When possible, we used the estimated live weights of prey consumed by tigers in Nagarahole, India (Karanth & Sunquist 1995); otherwise, we used average prey weights (Schaller 1967, Lekagul & McNeely 1977, Nowak 1991).

Slight differences in the methods of the three studies required minor data manipulations to maintain analytical consistency. In the interest of focusing on the tiger's natural prey, McDougal excluded scats containing domestic livestock, and to be consistent, we did likewise. Consequently, we disregarded all occurrences of domestic livestock (one in our study and two in Sunquist) in tiger scats. Similarly, we discarded all scats composed entirely of vegetation. In his analysis, Sunquist failed to distinguish between chital, hogdeer and muntjak hairs, resulting in a single prey category (Axis spp. or Muntiacus muntjak). Again, for consistency, we aggregated all occurrences of chital, hogdeer and muntjak into the single prey category for all three studies. For calculations of percent biomass and mean mass of prey consumed, we took the mean weight of all three prey species combined.

We used data on prey abundance collected during companion studies that were conducted in 1974, 1977 and 1980, corresponding with the time periods when the scats were collected (Seidensticker 1976b, Mishra 1982, Tamang 1982) to assess the relationship between prey abundance and tiger diet using Spearman rank correlations. We summed the density estimates for chital, hog deer and muntjak into a single category for comparison with our combined prey category. We calculated the total number of prey species consumed to determine trophic niche breadth.

To examine the effect of study duration on estimates of tiger diet, we combined our data with those of Sunquist and qualitatively compared the percent biomass contribution and maximum number of species consumed in either study to the aggregate for both studies combined. To examine the effect of study extent on estimates of tiger diet, we compared the percent biomass contribution and the number of species preyed upon from McDougal and Sunquist to their aggregate estimates. To quantify the effect of study duration and study extent on estimates of tiger diet, we compared the percent biomass contribution and number of species consumed from all three study sites to their aggregate estimates.

Results

We collected 77 scats and compared these data with those from the 123 and 55 scats collected by McDougal and Sunquist, respectively. We excluded 15 scats which contained grass, leaves, domestic livestock or unidentifiable animal remains from further analyses. Tigers preyed upon a minimum of eight, five and 16 species during McDougal, Sunquist and our study, respectively. Cervids constituted the primary prey of tigers for all three studies with > 65% percent occurrence (Table 2). In our study, tigers preyed primarily on sambar, differing from studies McDougal and Sunquist in which the combined prey category of chital, hog deer and muntjak formed the bulk of tiger diet. The discrepancy in primary prey among studies differed in terms of biomass contribution (Fig. 2); sambar contributed the largest percent biomass for McDougal and for our study and nearly so for the study by Sunquist. However, when we increased study duration (aggregating data from Sunquist and our study), sambar contributed the most biomass to tiger diet, replacing the combined prey category of Axis spp. or Muntiacus muntjak. For all other species, increasing study duration or extent resulted in estimated biomass contributions between those estimated for each study singly, but did not change their order of importance (see Fig. 2). Mean mass of prey consumed was 55.3 kg for McDougal, 50.2 kg for Sunquist and 36.7 kg for our study. Mean mass of prey consumed when we extended the study duration was 41.1 kg, was 53.8 kg when the extent was increased, and was 47.4 kg when duration and extent were increased.

Composition of tiger diet was unrelated to the density of prey (ρ = −0.10, df = 10, P = 0.74). Tigers preyed upon a total of 18 species (15 prey categories) across the three studies (see Table 2). Tiger diet was more varied during our study (N = 13 prey categories) than during the studies of McDougal (N = 6) or Sunquist (N = 5). Increasing the study duration (i.e. aggregating data from Sunquist and our study) resulted in the identification of one more prey category (belonging to Viverridae) consumed by tigers than in our study alone, and 10 more than in the Sunquist study. Increasing the size of the study area (aggregating data from McDougal and Sunquist) resulted in the identification of one more prey category (belonging to Viverridae) than in McDougal alone and two more than in Sunquist. Increasing study duration and study area size (aggregating data from all three studies) increased the number of prey categories by nine, 10 and two, for McDougal, Sunquist and our study, respectively.

Discussion

Composition of tiger diet for Sunquist and McDougal, which overlapped in time but not in space, were broadly similar. Overlapping confidence intervals indicate that, while there were slight differences in the percent occurrence of primary prey items among the two studies (i.e. sites), the differences were largely indistinguishable. Two species were unique to McDougal, whereas one species was unique to Sunquist; regardless, none of the species unique to either study contributed substantially to tiger diet (see Table 2).

Conversely, the composition of tiger diet for Sunquist and our study, which overlapped in space but not in time, differed dramatically, with sambar contributing far more, and the prey category containing chital, hog deer and muntjak contributing far less to tiger diet during the our study (see Table 2), to the extent that sambar became the largest contributor, in terms of biomass, to tiger diet (see Fig. 2). If tigers in Chitwan were non-selective predators, we would expect that an increase in the consumption of sambar was the result of either increased sambar abundance or decreased abundance of the other three species. However, studies of tiger prey selection from other tiger populations indicate that tigers are selective predators (Karanth & Sunquist 1995, Bagchi et al. 2003, Wegge et al. 2009), and the results of our rank correlation analysis support the conclusion that tigers are actively selecting certain prey types. Although our analysis did not incorporate prey selectively, the heavy reliance on sambar by Chitwan tigers is consistent with and provides further support for the finding that tigers actively select prey weighing > 176 kg (Karanth & Sunquist 1995).

Increasing study duration and study area size changed the relative contribution of prey species to tiger diet. Combining the results of Sunquist with McDougal and our study increased the percent biomass contribution of sambar while decreasing the percent biomass contribution of the prey category combining chital, hog and barking deer, with the end result that sambar, rather than the combined prey category, contributed more biomass to tiger diet in Chitwan; all other changes to prey biomass contributions were nominal and did not change their relative rank importance in tiger diet. Previous analyses have suggested that the percent occurrence of prey items tends to stabilize between 50 and 60 scats (Biswas & Sankar 2002, Bagchi et al. 2003), so a possible explanation for the observed differences is that only by combining the 50 scats from Sunquist with the 62 scats from our study were we able to get an accurate estimate of tiger diet. If this is the case, results from studies using fewer or similar numbers of scats to estimate tiger diet should be viewed more sceptically.

Similar to other studies of tiger diet (Karanth & Sunquist 1995, Miquelle et al. 1996, Biswas & Sankar 2002), tigers in Chitwan preyed heavily on medium- to large-sized large cervids. It remains unclear why the average weight of tiger prey was lower and the diversity of tiger diet higher during our study than during McDougal or Sunquist. It has been suggested that tigers include more species, as well as smaller species, in their diet when their primary prey (cervids) are unavailable (Sunquist et al. 1999, Reddy et al. 2004). However, owing to enhanced protection, cervid populations were most abundant during our study (Mishra 1982) suggesting that this was not the case. We speculate that the reason that a greater diversity and smaller mean size of prey items occurred during our study happened because we sampled a different segment of the tiger population. Although our scats were collected in the same study area as Sunquist, they were collected during activities geared towards understanding the role of juvenile dispersal in structuring the population (Smith 1993), and consequently, may have been more prone to collecting scats of juveniles and dispersers than were Sunquist and McDougal, which focused on resident, breeding adults during their studies. If this is the case, it suggests that diet may vary by tiger age classes, and future studies should attempt to determine age-specific dietary habits.

Differences among studies in the number of prey species consumed suggest that studies of short duration and limited spatial extent may fail to identify rare prey items (i.e. those constituting < 5% diet biomass) in the diet of tigers. Whereas it is intuitively appealing to assume that common prey is important, the converse assumption, that rare prey are unimportant, is injudicious as rarity does not imply a relative value for prey categories. For example, in our study, prey categories that individually contributed < 5% of diet biomass, collectively contributed > 20% of the biomass consumed by tigers. Rather than being unimportant, prey that is only rarely consumed may prove vital during periods when more commonly consumed prey are unavailable (Schaller 1967).

Similarly, despite the tiger's apparent selection of medium to large prey (Karanth & Sunquist 1995, Biswas & Sankar 2002), one should not discount the importance of small prey to supplement tiger diet. During periods when larger prey are unavailable, perhaps due to disease outbreaks (McDougal 1977), seasonal changes in prey behaviour (Schaller 1967) or peaks in poaching pressure, small prey may provide sufficient sustenance to allow tiger persistence. However, we caution that, whereas small prey may collectively contribute substantially to tiger diet, it remains questionable whether, in the absence of larger prey items, small prey would be sufficient to support tiger reproduction (Sunquist et al. 1999). Modeling of cougar energetics suggests that raising young to independence requires > 250% more food than required by an adult female without young (Laundré 2005). Given that cougars and tigers have similar litter sizes and lengths of young dependency (Sunquist & Sunquist 2002), it seems highly unlikely that a tigress could raise cubs solely on small prey.

Scats used for all three studies were collected during the dry season, limiting our analysis to the annual, rather than seasonal, dynamics of tiger diet. However, it is likely that seasonal fluctuations in prey abundance and vulnerability resulting from birth pulses, rutting behaviour and migration may result in seasonal variation in tiger diet (Schaller 1967, Sunquist 1981). If possible, future research should aim to document seasonal variation in tiger diet. Of particular interest would be tiger diet during the wet season, a period when prey may be more diffuse due to the widespread availability of water, and which consequently may be a difficult period for tigers.

Our analysis provided anecdotal evidence of pseudoreplication (Hurlbert 1984). Studies of tiger diet are primarily conducted incidental to other research projects. Consequently, scats are normally collected opportunistically (i.e. haphazardly), rather than by randomized sampling design. Scats collected opportunistically, incidental to other activities (e.g. while walking transects or investigating a kill), are unlikely to be independent, and may result in biased estimates of the relative importance of certain prey (Marucco et al. 2008). This bias was evident in our study when two scats were identified as containing hairs of the Asiatic golden cat Pardofelis temminckii which were considered extremely rare in the park during our study. Upon closer scrutiny, we realized that both scats had been collected on the same day within a short distance of one another, likely representing the predation of a single golden cat. Presumably, a single tiger had killed a golden cat, and then defecated several times along that same game trail. In this case, the non-independence of the scats did not have a large effect on estimates of tiger diet; however, it is easy to envision a scenario in which the non-independence of scats would result in significant bias. For example, if a tiger kills a large animal such as a gaur, it typically remains with the kill for several days, alternately feeding, resting and defecating. A single kill as large as a gaur Bos gaurus will result in numerous tiger scats. If all the scats from that gaur kill are collected, the importance of gaur to tiger diet is likely to be overestimated while the importance of all other prey items will be underestimated. Future research on tiger diet should attempt to minimize these potential biases by adopting a sampling framework that is statistical, rather than haphazard. To accomplish this, researchers could use molecular methods to identify the individual tigers responsible for each scat and sample accordingly (Prugh et al. 2008). Use of this method has the added advantage of allowing researchers to investigate intersexual (Pilgrim et al. 2005) and individual variation in diet (Prugh et al. 2008). However, we acknowledge the difficulty of collecting viable genetic material given the climate in most of tiger range. Alternatively, researchers could minimize the potential issue of pseudoreplication by randomizing the placement of sampling transects.