Predicting the current and future risk of ticks on livestock farms in Britain using random forest models

The most abundant tick species in northern Europe, Ixodes ricinus, transmits a range of pathogens that cause disease in livestock. As I. ricinus distribution is influenced by climate, tick-borne disease risk is expected to change in the future. The aims of this work were to build a spatial model to predict current and future risk of ticks on livestock farms across Britain. Variables relating both to tick hazard and livestock exposure were included, to capture a niche which may be missed by broader scale models. A random forest machine learning model was used due to its ability to cope with correlated variables and interactions. Data on tick presence and absence on sheep and cattle farms was obtained from a retrospective questionnaire survey of 926 farmers. The ROC of the final model was 0.80. The model outputs matched observed patterns of tick distribution, with areas of highest tick risk in southwest and northwest England, Wales, and west Scotland. Overall, the probability of tick presence on livestock farms was predicted to increase by 5-7 % across Britain under future climate scenarios. The predicted increase is greater at higher altitudes and latitudes, further increasing the risk of tick-borne disease on farms in these areas.

Distribution and prevalence of ticks and tick-borne disease on sheep and cattle farms in Great Britain

INTRODUCTION

The most abundant and widespread tick species in Great Britain, Ixodes ricinus, is responsible for the transmission of a range of pathogens that cause disease in livestock. Empirical data on tick distribution and prevalence are required to inform farm management strategies. However, such data are largely unavailable; previous surveys have been rare and are usually relatively localised.

METHODS

A retrospective questionnaire survey of farmers was used to assess the reported prevalence of ticks on livestock across Great Britain. Spatial scan statistics and kernel density maps were used to assess spatial clustering and identify areas of significantly elevated risk, independent of the underlying distribution of respondents. Logistic regression models were used to identify risk factors for tick presence.

RESULTS

Tick infection risk to livestock is shown to be spatially aggregated, with areas of significantly elevated risk in north Wales, northwest England and western Scotland. Overall, the prevalence of farms reporting tick presence was 13% for sheep farms and 6% for cattle farms, but in "hot spot" clusters prevalence ranged between 48-100%. The prevalence of farms reporting tick-borne disease overall was 6% for sheep and 2% for cattle, but on farms reporting ticks, prevalence was 44% and 33% for sheep and cattle farms, respectively. Upland farming, larger flock sizes, region and the presence of sheep on cattle farms were all significant risk factors for tick presence.

CONCLUSIONS

These data have important implications for assessing both the risk of tick-borne disease in livestock and optimising approaches to disease management. In particular, the study highlights the need for effective livestock tick control in upland regions and the southwest, and provides evidence for the importance of sheep as tick maintenance hosts.