Integrating individual behaviour into simulations of disease management in Ethiopian wolves

Domestic animals are one of the primary global sources of disease due to their position at the interface between humans and wild environments. Wildlife that lives on this interface is extremely vulnerable to infectious disease. Disease has been a particularly prominent threat to wild carnivores that are already threatened through habitat loss or poaching. Outbreaks of disease have caused population declines in lions (Panthera leo), African wild dogs (Lycaon pictus), and Ethiopian wolves (Canis simensis). A specialised social canid, the Ethiopian wolf is restricted to the Afroalpine ecosystem of Ethiopia with no more than 500 individuals surviving; the species is ultimately threatened by a shrinking range, but the most immediate threats are rabies and canine distemper virus (CDV) transmitted by sympatric free-ranging domestic dogs (C. familiaris). A series of sequential and even concurrent outbreaks of these pathogens have led to severe declines in the largest population in the Bale Mountains; and an isolated small population to near extinction. The impact of disease on the last remaining Ethiopian wolf populations is currently mitigated by dog vaccination campaigns and reactive vaccination of the wolves. Successful and ongoing oral vaccination trials aim to use preventive wolf vaccinations to reduce the frequency and impact of outbreaks, but a better understanding is required of how to best apply different vaccination regimes (both of dogs and wolves) and dog management or exclusion strategies, proposed as part of an integrated approach in the National Action Plan for the conservation of the species. With this purpose in mind, custom-built monitoring devices were constructed to study the movement and behaviour of domestic dogs in sympatry with wolves using high resolution GPS and accelerometer data. These devices were deployed on 26 domesticdogs in the Web Valley, Bale Mountains National Park, over multiple weeks in two seasons in 2018. Data (during day time and from when dogs left the village area) were analysed to determine dogs’ extent of movement based on inter-point distances, their habitat preferences from Z tests of proportions of available to used habitat, and home ranges using 90% autocorrelated kernel density estimate. Dogs ranged out of villages to a mean distance of 727m (SD ±79.4m) away from their respective village within a home range of 3.0km2 (SD ±0.63km2) and showed a preference for grassland (z = 192.35, df = 1, p = < 0.00001), and drainage lines (z = 54.396, df = 1, p = < 0.00001). Deep learning was used on the accelerometer data to determine patterns in time allocation of activities throughout the day. Dogs exhibited two activity peaks: 07:00h to 10:00h and 14:00h to 18:00h, when they spent the majority of their time foraging or travelling. From long-term observations of wolf packs (focal pack observations and individual behavioural follows) equivalent home ranges were calculated as 10km2 (SD ± 0.975km2, n = 253, between 2007 and 2018) as well as movement and behavioural budget breakdown. A measure of overlap between the two species was developed to understand when and where the probability of disease transmission might be highest, calculating as overlap of function graphical area using integration. Dogs and wolves overlapped by 75% in active behaviours across the day and on average 48% of a dog’s home range overlapped with wolf range. An epidemiological Agent-Based model including the two species, rabies and CDV, was parameterised with data on their movement, behaviour and demography, and published data on the dynamics of the diseases. When validated using historical disease outbreak data from the Bale Mountains population, the model produced a low root mean squared error of 3.74 for predicted population size and had a low linear sensitivity (less than 0.1) to input parameters. The validated model was used to simulate a variety of disease and management scenarios in the Bale wolf population. Under current disease management interventions (domestic dog vaccinations and reactive wolf vaccination in response to outbreaks) the population extinctionrisk was predicted to be 17% (in 30 years), compared to 58% without management while the number of epizootics doubled without management. The model was run without reactive wolf vaccination which increased extinction risk by 10% with no effect on the number of epizootics compared to current management. Restricting dog movement to close to villages, or fewer dogs going into wolf habitat, removed extinction risk to 0% and reduced the incidence of epizootics by 50% in models. A parameter sweep of combinations of levels of vaccination coverage predicted that preventive wolf vaccination will be more effective and cost-effective at reducing epizootics in the wolves than dog vaccination at all levels of coverage. This is because the large number and turnover of dogs reduce effectiveness of dog vaccination in preventing outbreaks in the wolves. The most cost-effective strategy predicted was 70% of vaccination of the wolves with 0% dog vaccination. Until CDV vaccination is trialled and approved for wolves, it is better to retain high levels of dog vaccination to protect wolves from CDV, while using oral rabies vaccination in wolves to provide higher levels of rabies protection. This thesis showcases how custom-built monitoring devices, tailored to individual projects, have the potential to open up a wealth of new data sources and to improve the quantification of overlap between hosts for the modelling of transmission dynamics. The agent-based modelling framework developed could be expanded to numerous systems, particularly ones where domestic dogs are the primary host of disease. The model outputs provide guidelines for the implementation of preventive vaccination in wolves based both in its effectiveness as well as the long-term costs. The model also supports behavioural changes on the part of dog owners living next to the wolves as an effective measure that would require cooperation from all stake-holders. The model framework could be used globally to understand the impacts of behaviour on disease transmission and demonstrate the benefits of wildlife vaccination for species threatened by disease