The development of high resolution maps of tsetse abundance to guide interventions against human African trypanosomiasis in northern Uganda

BACKGROUND: Vector control is emerging as an important component of global efforts to control Gambian sleeping sickness (human African trypanosomiasis, HAT). The deployment of insecticide-treated targets (“Tiny Targets”) to attract and kill riverine tsetse, the vectors of Trypanosoma brucei gambiense, has been shown to be particularly cost-effective. As this method of vector control continues to be implemented across larger areas, knowledge of the abundance of tsetse to guide the deployment of “Tiny Targets” will be of increasing value. In this paper, we use a geostatistical modelling framework to produce maps of estimated tsetse abundance under two scenarios: (i) when accurate data on the local river network are available; and (ii) when river information is sparse. METHODS: Tsetse abundance data were obtained from a pre-intervention survey conducted in northern Uganda in 2010. River network data obtained from either digitised maps or derived from 30 m resolution digital elevation model (DEM) data as a proxy for ground truth data. Other environmental variables were derived from publicly-available resolution remotely sensed data (e.g. Landsat, 30 m resolution). Zero-inflated negative binomial geostatistical models were fitted to the abundance data using an integrated nested Laplace approximation approach, and maps of estimated tsetse abundance were produced. RESULTS: Restricting the analysis to traps located within 100 m of any river, positive associations were identified between the length of river and the minimum soil/vegetation moisture content of the surrounding area and daily fly catches, whereas negative associations were identified with elevation and distance to the river. The resulting models could accurately distinguish between traps with high and low fly catches (e.g. < 5 or > 5 flies/day), with a ROC-AUC (receiver-operating characteristic - area under the curve) greater than 0.9. Whilst the precise course of the river was not well approximated using the DEM data, the models fitted using DEM-derived river data performed similarly to those that incorporated the more accurate local river information. CONCLUSIONS: These models can now be used to assist in the design, implementation and monitoring of tsetse control operations in northern Uganda and further can be used as a framework by which to undertake similar studies in other areas where Glossina fuscipes fuscipes spreads Gambian sleeping sickness.