Estimation of tiger densities in India using photographic captures and recaptures

Ecology, Dec, 1998 by K. Ullas Karanth, James D. Nichols

The model selection algorithm of CAPTURE identified [M.sub.0] as the most appropriate model for each of the four study areas (Table 2). Our a priori expectation was that tiger capture probabilities would be heterogeneous over individuals. Model [M.sub.h] had the second highest model selection criterion for all areas except Kaziranga, where [M.sub.h] again received a high model selection criterion score. Model [M.sub.h] fit the data well in all four study areas. However, the test ([M.sub.0] vs. [M.sub.h]) for heterogeneity provided no evidence of this effect on any of the three areas for which it was tested (Table 2). We also tested for temporal variation in capture probability ([M.sub.0] vs. [M.sub.t]), but found no evidence (Table 2). Although model [M.sub.0] was the apparent model of choice for all four areas, the estimator of population size associated with this model is known to be sensitive to violations of the underlying model assumption of homogeneous capture probabilities (e.g., Otis et al. 1978). The population size estimator of model [M.sub.h], in contrast, is known to be robust to violation of underlying model assumptions (Otis et al. 1978, Burnham and Overton 1979). We thus report population size estimates computed under both of these models, although we have more confidence in the [M.sub.h] estimates, because of estimator robustness.

Estimated capture probabilities per occasion (the probability that a tiger in the sampled area is photographed on a single sampling occasion) were consistent across the four areas, ranging from 0.12 to 0.26 under model [M.sub.0]. The estimated average capture probabilities under model [M.sub.h] were slightly lower, as expected, ranging from 0.11 to 0.22 (Table 3). The estimated probability that a tiger was photographed on at least one occasion (estimated by [Mathematical Expression Omitted]) ranged from 0.92 to 1.00 under model [M.sub.0] estimates, and from 0.79 to 1.00 under model [M.sub.h] estimates (Table 3). At Pench, we estimated only five adult tigers, and the estimates for the other three areas ranged from 24 to 33 adult tigers, depending on the estimation model used and the location (Table 3).

The mean maximum distances moved by tigers caught at least twice ranged from nearly 3 km to [greater than]5 km, yielding estimated boundary strip widths ranging from 1.38 km to 2.67 km (Table 4). The estimated effective areas sampled by the traplines ranged from just [greater than] 120 [km.sup.2] to [greater than]280 [km.sup.2]. Because of the robustness [TABULAR DATA FOR TABLE 3 OMITTED] of the model [M.sub.h] estimates of population size, we used these estimates in conjunction with Eq. 3 to estimate tiger density. The estimated density at Pench was only 4.1 tigers/100 [km.sup.2], but the estimates for the other three areas ranged from 11.5 to 16.8 tigers/100 [km.sup.2] (Table 4).

The values of the relative abundance indices for leopards calculated from camera trap data were: 5.44 captures/100 trap-nights of sampling effort for Nagarahole, 2.28 for Pench, 0.50 for Kanha, and 0.18 for Kaziranga. The estimated densities of principal prey species from our line-transect surveys are reported in Table 5. based on the natural size differences prevailing among these species, following Karanth and Sunquist (1995), we categorized them by mass as large prey ([greater than] 176 kg), medium-sized prey (21-175 kg), and small prey ([less than]20 kg).