Abstract
A series of 59 transect surveys was conducted in selected wet forest habitats, along the coast of West Africa, to estimate the density distribution of African Hinge-back tortoises (Kinixys homeana and K. erosa). Line transect data were fed into a simple model to derive a detection function. The parameters estimated by the model produced an elaborate characterisation of tortoise distribution, which proved to be useful in the formulation of hypotheses about tortoise densities. Line transect data were analysed by DISTANCE, with a series of key and the series adjustment: the uniform function, the 1-parameter half-normal function, and the 2-parameter hazard-rate function were considered as key functions; the cosine series, simple polynomials, and Hermite polynomials were considered as series expansions. The detection function was estimated separately for Kinixys homeana and K. erosa, and for transects grouped for each study area by considering all the combinations of the above key functions and series expansions. The Akaike Information Criterion (AIC) was computed for each candidate model and used for model selection. The best model of the detection function, for both the tortoise species was the uniform function with no series expansion. Model results indicated that the density of the two species was inversely related at the local scale, and complementary across the region; such that the density of one species increases from West to East while the other one declines. Overall, the comparison of density estimates between the two tortoises is consistent with a former hypothesis suggesting inter-specific competition and consequent resource partitioning. Other causes may contribute to explain the observed patterns, including the low productivity of rainforest habitats and long-term human perturbation.
