Could a rabies incursion spread in the northern Australian dingo population? Development of a spatial stochastic simulation model

Abstract

Australia, home to the iconic dingo, is currently free from canine rabies. However northern Australia, including Indigenous communities with large free-roaming domestic dog populations, is at increased risk of rabies incursion from nearby Indonesia. We developed a novel agent-based stochastic spatial rabies spread model to evaluate the potential spread of rabies within the dingo population of the Northern Peninsula Area (NPA) region of northern Australia. The model incorporated spatio-temporal features specific to this host-environment system, including landscape heterogeneity, demographic fluctuations, dispersal movements and dingo ecological parameters-such as home range size and density-derived from NPA field studies. Rabies spread between dingo packs in nearly 60% of simulations. In such situations rabies would affect a median of 22 dingoes (approximately 14% of the population; 2.5-97.5 percentiles: 2-101 dingoes) within the study area which covered 1,131 km2, and spread 0.52 km/week for 191 days. Larger outbreaks occurred in scenarios in which an incursion was introduced during the dry season (vs. wet season), and close to communities (vs. areas with high risk of interaction between dingoes and hunting community dogs). Sensitivity analyses revealed that home range size and duration of infectious clinical period contributed most to the variance of outputs. Although conditions in the NPA would most likely not support a sustained propagation of the disease in the dingo population, due to the predicted number of infected dingoes following a rabies incursion and the proximity of Indigenous communities to dingo habitat, we conclude that the risk for human transmission could be substantial.

Fig 1. Location of the study area in the Northern Peninsula Area (NPA) of Queensland, Australia.

Location of the study area and unsuitable habitat types for dingoes’ long-term survival and for movement of dingoes during the dry and wet seasons used in a spatial rabies spread model within the dingo population in the NPA.

Fig 2. Proposed daily probability of contact of two dingoes from different packs according to distance.

A) Influence of the home range size and number of home range days on the daily probability of contact between a pair of non-clinical dingoes of different packs. For illustrative purposes, the two dingoes in each pair (i.e. each curve) occupy home range areas of equal size, either small (most likely value from the second half of the wet season) or large (most likely value from the second half of the dry season). B) Influence of rabies-induced home range size changes on the probability of contact between a pair of dingoes of different packs. In this example, the healthy dingoes (grey curve) both have a standard deviation, σ, equal to 1.0 km, with an average number of home range days, n_ij, equal to 3. The pink and blue curves represent the revised probability of contact if one of the individuals becomes clinically infectious with the furious and dumb forms, respectively, applying the maximum and minimum inflation factors on standard deviations (i.e. factor of 2 and 0.5, respectively), with the number of home range days unchanged.

Fig 3. Example of a simulated spatial spread of rabies in a dingo population in northern Australia.

Results of a model simulation at time (A) 0, (B) 50 days, (C) 100 days and (D) 167 days (end of the outbreak). The simulation was initiated by the infection of a single primary case (A; red) and terminated once there were no longer any infected dingoes. (A-D) Each dot represents the home range centroid of a pack of dingoes, which can either be categorized as “Never infected”, “Previously infected” (at least one dingo from the pack, whether it has died or moved, was previously infected), “Actively infected” (at least one dingo from the pack is latently infected or infectious). The grey zones represent unsuitable habitat for movement of dingoes during the dry season and the dark-beige zone in (D) represents the area of infection. For this particular simulation, the density of dingoes for the dry season was set at 0.135, the primary case occurred at the beginning of the dry season (Dry 1 scenario) and the “Hunter” scenario was selected for the location of the primary case. All other parameters were randomly chosen, using the specified distributions in Table 1.

Fig 4. Outputs obtained from 200,000 simulations of a spatial rabies spread model within a dingo population.

Outcome measures include (A) number of infected dingoes, (B) proportion of infected packs, (C) R₀ at the dingo level, (D) R₀ at the pack level, (E) area of infection, (F) maximum distance of infection, (G) speed of disease spread and (H) proportion of simulations for which infection has spread to more than one pack and more than 10% of the initial dingo population. For (A-G), the distribution of outputs reported are based on simulations in which disease has spread to at least one other pack, whereas proportions in (H) are based on all simulations. Triangles represent median values, diamonds represent the proportion of simulations with spread of disease to more than one pack and squares represent the proportion of simulations with spread of disease to at least 10% of the averaged initial dingo population. The vertical bars correspond to the range between the 2.5^th and 97.5^th percentiles of values distribution.

Fig 5. Global sensitivity indices for input parameters of a rabies spread model within a dingo population.

First order (cyan) and total effect (pink) Sobol’ sensitivity indices for each input parameter on (A) the number of infected dingoes, (B) the area of infection and (C) the binary outcome indicating whether the infection has spread to at least one other pack, following the incursion of rabies within the dingo population. Error bars represent the 95% confidence interval for each Sobol’ index. The abbreviation “Prob.” stands for “probability”, “HR” for “Home range”, “Cx” for “clinical”, “D” for “dumb form” and “F” for “furious form”.