3.2. Contact structure: domestic dogs
For the transmission of lyssavirus (the cause of rabies) between dogs, the probability of transmission given contact (a bite) has been estimated to be 49% (95% CI, 45−52%) (Hampson et al., 2009). The epidemiology of transmission is not subject to great uncertainty because the parameter estimate is not highly variable: this is information that can be obtained through controlled laboratory studies and detailed field observations. Therefore, understanding rabies spread at the wild−domestic interface depends on knowledge of the pattern of contacts between domestic and wild dogs. Canine rabies lyssavirus can only be transmitted between live animals, so the role of the environment in indirect transmission is not relevant. Thus, with documented knowledge of the probability of transmission given contact, the challenge to model rabies spread within populations and between populations at the wild−domestic interface is to characterise and measure effective contact.
Within Indigenous communities across Australia, it is common practice to allow domestic dogs to roam freely both within and around the communities, with some individuals traveling to more distant bushland areas (Bombara et al., 2017a; Hudson et al., 2017; Maher et al., 2019). Such roaming behaviours increase the likelihood of disease spread within and between communities, as well as creating an opportunity for disease transmission beyond communities via interactions with wild dogs.
In northern Australia, and specifically within the NPA, the roaming distributions of domestic dog − and inferences regarding contacts − have been studied extensively. For domestic dogs, the approach has used GPS units attached to collars (Figure 1A). With periods of monitoring from 3 days up to 2 months, GPS data (Figure 1B) allow dogs’ home ranges (HR) and utilisation distributions (UD) to be estimated (Figure 1C). The Biased Random Bridge (BRB) analytical method has application to such GPS data in which the period between GPS fixes is semi-regular but there can be considerable time periods when no fix is recorded (for example, a dog lying underneath a car). The BRB method places kernel functions over each movement of the dog between two consecutive GPS fixes, therefore more realistically tracing the dog’s movements. Kernel smoothing is carried out considering not only recorded GPS fixes, but also interpolated locations. The dog is assumed to move between two successive fixes according to a biased random walk, i.e. the roaming pattern has a drift along each track, represented by a diffusion parameter (Dürr and Ward, 2014). The estimated UD has a probabilistic interpretation; for example, at the 95% isopleth, this is the area in which the dog spent 95% of its time during the monitoring period.
In initial studies in the NPA, Dürr and Ward (2014) estimated UDs for roaming dogs of up to 20 ha (95% isopleth). In addition, within this short-term study dogs were recorded to be taken in vehicles on hunting trips of up to 25km in distance and 6 hrs in duration (Dürr and Ward, 2014). In another study, seven community dogs were identified with an estimated HR > 20 ha (maximum 40 ha), which included the bushland surrounding these Indigenous communities, areas in which many dogs were observed to spend considerable periods of their time (Bombara et al., 2017a). In a study using longer monitoring periods (up to 67 days), three classes of domestic dogs were characterised: stay-at-home, roamers and explorers (Hudson et al., 2017). Explorer dogs showed large variability in their roaming patterns, but their HRs continued to increase as the monitoring period increased. In further analyses, Maher et al. (2019) examined the temporal activity of roaming domestic dogs and found that dogs roamed further during the dry season and that their daily roaming behaviour peaked between 1700 and 1800 hours. Some dogs were recorded up to 2.5 km from their residence in this study. The studies conducted in the NPA suggested that mostly, domestic dogs remained within their communities with ample opportunities to contact other domestic dogs. However, it was also observed that some of these dogs could roam considerable distances from their communities into bushland areas, demonstrating a potential wildlife−domestic interface.