Mapping tsetse habitat suitability in the common fly belt of southern Africa using multivariate analysis of climate and remotely sensed vegetation data.

The distribution of Glossina morsitans centralis, Glossina morsitans morsitans and Glossina pallidipes are described in part of southern Africa, using a range of multivariate techniques applied to climate and remotely sensed vegetation data. Linear discriminant analysis is limited in its predictive power by the assumption of common covariances in the classes within multivariate environment space. Maximum likelihood classification is one of a variety of alternative methods that do not have this constraint, and produce a better prediction, particularly when a priori probabilities of presence and absence are taken into account. The best predictions are obtained when the habitat is subdivided, prior to classification, on the basis of a bimodality detected on the third component axis of a principal component analysis. The results of the predictions were good, particularly for G.m.centralis and G.m.morsitans, which gave overall correct predictions of 92.8% and 85.1%, with a Kappa index of agreement between the prediction and the training data of 0.7305 and 0.641 respectively. For G.pallidipes, 91.7% of predictions were correct but the value of Kappa was only 0.549. Very clear differences are demonstrated between the habitats of the two subspecies G.m.centralis and G.m.morsitans.