The use of discriminant analysis for the estimation of sampling biases for female tsetse

Abstract

ABSTRACT. In this paper we investigate whether the technique of discriminant analysis can be used to estimate sampling biases for female tsetse. Discriminant analysis was first applied to laboratory samples of female tsetse, Glossina morshans morsitans Westwood, to test whether flies of known history could be assigned to the correct day of the pregnancy cycle on the basis of their fat, haematin and corrected residual dry weight. Following the satisfactory results from the laboratory samples, the same technique was applied to field samples of G.m. centralis Machado captured by electric traps and hand nets in Zambia and of G.palpalis palpalis (Robineau‐Desvoidy) captured in biconical traps at five sites in Ivory Coast. The results show that flies on day 1 of the pregnancy cycle were most likely to be caught, with a second peak of day‐6 and day‐7 flies, while very few day‐8 or day‐9 flies were caught. These major peaks in fly trappability coincide with the known feeding habits of female tsetse, and indicate synchrony of feeding by many members of the population immediately after larviposition and again as the larva in utero moults from the second to third instar. G.palpalis is relatively more available at this later stage of its pregnancy cycle to the capture methods used than is G.morsitans. A third feed may be taken at a more variable point in the pregnancy cycle. This method of estimating the sampling biases of female tsetse could allow an estimate of total population size, as long as the absolute sampling efficiency of flies on any one day of the pregnancy cycle could be established by, for example, mark‐release‐recapture experiments.