The use of time-series analysis to forecast bont tick (Amblyomma hebraeum) infestations in Zimbabwe.

Abstract

Studying the dynamics of tick infestations on cattle is an essential step in developing optimal strategies for tick control. Successful strategic tick control requires accurate predictions of when tick infestations will reach predetermined threshold levels. In the case of Amblyomma hebraeum, earlier work has shown that there is no consistent pattern of seasonal activity. This means that a statistical model for predicting A. hebraeum infestations cannot reliably use climatic factors as the only independent variables. An alternative method is to apply time-series, or auto-regressive moving-average (ARMA), analysis which uses only the past population patterns to predict future trends. This technique was applied to a data set consisting of 108 weekly tick counts of A. hebraeum (adult males, standard females, flat females and standard nymphs), conducted at an experimental station in southeastern Zimbabwe. The ability of the ARMA models to fit and predict actual tick infestations was judged using two sets of criteria. The first set focused on the goodness-of-fit, and used the adjusted R2 values, Q statistic and the Akaike Information Criteria. The second set of criteria measured the forecasting accuracy of an estimated equation, and consisted of regressing a 9-period forecast against an actual out-of-sample data set not used in the estimation process. The root mean square error of the forecast was also considered when comparing several models for the same data set. Using these criteria, the models estimated using the ARMA technique were judged to both fit and forecast with sufficient accuracy to warrant their use in strategic tick control. Although the success of using ARMA to forecast A. hebraeum is partly due to the non-seasonal behavior of the species, the results presented here suggest that it is worthwhile exploring the use of ARMA techniques to model the dynamics of other tick species. Where independent variables exert considerable influence on the dynamics of a tick species, these variables can be incorporated into an ARMA-style model.