Spatio-temporal analysis of dog ecology and rabies epidemiology at a wildlife interface in the Lowveld Region of South Africa

Abstract

It is estimated that free-roaming dogs comprise on average 75% of dog populations. Interactions between free-roaming dogs and wildlife occur across the globe but little is known about these interactions in the Lowveld region of South Africa, where wildlife areas are fenced off from surrounding communities. Even with extensive fences dogs are still entering the reserves. The study site, a private reserve in eastern Mpumalanga Province on the western boundary of the Greater Limpopo Transfrontier Park, has a high density of human settlements on its boundary. These communities own dogs, many of which are free-roaming. Between January 2009 and March 2014, 170 stray dogs were destroyed inside the reserve and 65.3% of the samples returned a positive result for rabies. Dogs are not limited to the reserve edges and have been documented several kilometres into the reserve. Eleven geographical factors were used in the spatial analysis: (i) camps or lodges in the reserve, (ii) fence line of the reserve, (iii) water points within the reserve, both natural and man-made (excluding rivers but including pans, dams and waterholes which hold water for most of the year), (iv) access roads from gates to camps and lodges, (v) access gates into the reserve, (vi) pickets (field ranger accommodation) and general staff accommodation, (vii) rivers in stream order from 1 to 6, (viii) vulnerable points for erosion along the fence line, and (ix) villages bordering the reserve. GPS locations of dogs shot were used to obtain nearest distance to each factor. Generalized linear models (GLM) were then used to analyse the spatial data of distance of dogs shot to the nearest factor. Dogs were significantly more likely to be shot further away from pickets and closer to minor rivers. There was a significant interaction between these two factors (p < 0.0001). Dogs that were shot further from villages (odds ratio 1.42, 95% confidence intervals 1.18 1.71, p = 0.0002) and closer to water (odds ratio 0.41, 95% confidence intervals 0.21 0.81, p = 0.009) were more likely to test positive for rabies. A univariate GLM, with distance to fence as the only explanatory variable, showed a significant association between this and rabies test result: for every 1 km further away from the fence the odds of a dog testing rabies positive increases by 1.68 (95% confidence intervals 1.20 2.36, p = 0.002). However the fence is likely not to be the influencing factor but rather other factors close to the fence. To remove the effect of the fence an analysis of the subset of dogs found further than 200m into the reserve found a positive association between distance from a village and a positive rabies test result (odds ratio 1.58, 95% confidence intervals 1.18-2.32, p = 0.007). Temporal analysis of the data shows a higher average monthly number of dogs shot during the wet season (Oct-Mar) as well as a higher variance although the cycle is not strictly seasonal. An upgrade of the reserve fence for security reasons coincided with a decrease in number of dogs destroyed in the reserve in 2012-1014. Home range analysis of most susceptible predators was collected from daily sightings data and overlaid with rabies hotspots, which gives management an indication of the need to vaccinate predators with home ranges in close proximity to the fence. The spatial results gives management an indication to increase efforts to destroy free-roaming domestic dogs further into the reserve as the likelihood of a positive rabies result is greater. Ultimately efforts should be focused outside the reserve in the communities to eradicate rabies from the host (domestic dogs) before there is a significant transfer of rabies to wildlife.