A Comparison of Two Methods of Estimating the Size of a Population of Giant Tortoises on Aldabra

Abstract

SUMMARY From a sample census of giant tortoises in the south-eastern region of Aldabra atoll, the size of the markable population was estimated as 87 300. A subsequent recapture study of the same population gave an estimate of 68 100. Sampling errors are unlikely to account for the difference, which is significant at the 1% level. After taking into account obvious violations of the underlying assumptions the difference remains. The sample census estimate is thought to be the more reliable. largely based on data from this sample census. The bulk of the tortoise population is in the south-east of Grande Terre (for map see Bourn & Coe 1978) where recapture work was subsequently conducted. An alternative estimate of the size of the population in this area, the Petersen estimate, was obtained from mark-recapture analysis. As it was hoped that the marking programme would form a foundation for a more detailed description of the population dynamics, it was desirable that the basic assumptions of mark-recapture analysis be seen to be satisfied. As the number of tortoises recaptured (221) was large compared with many recapture studies there is an unusual opportunity to test the main assumption underlying the Petersen method, that marked and unmarked animals have the same probability of being caught in the second sample. Seber (1973) observed that it is not always easy to detect departures from this assumption and quoted examples to show that the Petersen estimate may be unreliable even when the assumptions appear to be satisfied. He recommended that 'if the Petersen estimate is to be used extensively for a given species then it should be compared with other methods'. In this paper the Petersen estimate of population size is compared with the sample census estimate, and the advantages and disadvantages of both methods are considered.