Abstract

The study was undertaken with the objective of comparing the relative efficiency of some of the feasible methods of estimating the abundance of herbivores in the forest of Kerala and to suggest refinements in the existing methods. Both empirical and theoretical investigations were undertaken in order to meet the above objectives. Data collected from eight sanctuaries during the wildlife sensus conducted in 1993, were utilised to compare the relative efficiency of different detection function models for estimating the abundance of herbivores. The following species viz., elephant, sambar, spoted deer, barking deer, wild boar and gaur were considered for the study. Univariate half normal distribution was found promising with respect to precision of the density estimates. An examination of the theory showed that for a given set of detections, overestimation of distances in the field would lead to underestimation of density in the case of line transect sampling and vice versa with half normal detectionfunction. With Fourier series model for detection function, the direction of effect on density estimate was found to be governed by the range and distribution of the distance measurements. Simple linear regression equation fitted through the origin showed that there was underestimation by 2 m for every 100 m of actual distance which is negligible. the mean bias in the visual estimation of actual distance was not significantly different from zero. However, the coefficient of variation of visual estimatesof distance varied from 54 per cent in 0-20 m class to 34 per cent in 80-100 m class. For a given set of detections and transect length, increased disruption of distance values on an average was found to bring down the density estimates both in the case of half normal and Fourier series model.The sampling intensity required was found different for the different species. On an average, one transect of 2 km was found necessary for every 5 km2 of the area sampled. In line transect sampling, the form of thedetection function is found to vary with the local conditions associated with the forest type, weather condition, observer's fatigue etc. Random parameter model was formulated taking the two parameter negative exponential model as detection function. thebasic proposition was that apart from the estimation errors, the relation betweenn perpendicular distance and cumulative density function of the number of sightings can have different parameters in different locations and these can be viewed as random deviations from population level parameters. The model was tested utilising data collected for the species sambar from 10 wildlife sanctuaries at different periods. the method has the clear advantagge of being able to develop density estimates based on very few observations from a location which would be impossible through traditional methods. In the case of elephants and gaur, indirect evidence like dung density is a very strong indicator of the habitat use which is associated with animal density and therefore accurate estimation of dung density is important. An analysis of data on distance to dung piles, collected during the course of line transect sampling, indicated that Fourier series model is a good choice for detection function model is most of the vegetation types. The present study has shown that total count is inapplicable for estimation of animal abundance as it leads to heavy undercounting. Line transect sampling has a firm theoretical footing but suffers from low number of sightings arisingfrom low density of animals or poor detection percentage. However, indices of abundance based on indirect evidences would serve most of the practical purposes in wildlife management