Abstract:
African horse sickness (AHS) is a disease that is endemic to sub-Saharan Africa and is
caused by a virus potentially transmitted by a number of Culicoides species (Diptera:
Ceratopogonidae) including Culicoides imicola and Culicoides bolitinos. The strong association
between outbreaks of AHS and the occurrence in abundance of these two Culicoides species has
enabled researchers to develop models to predict potential outbreaks. A weakness of current
models is their inability to determine the relationships that occur amongst the large number
of variables potentially influencing the population density of the Culicoides species. It is this
limitation that prompted the development of a predictive model with the capacity to make
such determinations. The model proposed here combines a geographic information system
(GIS) with an artificial neural network (ANN). The overall accuracy of the ANN model is 83%,
which is similar to other stand-alone GIS models. Our predictive model is made accessible to a
wide range of practitioners by the accompanying C. imicola and C. bolitinos distribution maps,
which facilitate the visualisation of the model’s predictions. The model also demonstrates how
ANN can assist GIS in decision-making, especially where the data sets incorporate uncertainty
or if the relationships between the variables are not yet known.
