Population and disease modelling in the Tasmanian devil
thesis
posted on 2023-05-26, 02:23authored byBeeton, NJ
Infectious disease threatens many wildlife populations, but managing disease in free ranging populations is difficult. Resources, including time, are almost always limited; both for collecting data, and for being able to make effective decisions using these data. Modelling is an increasingly important and widespread tool in the arsenal of the wildlife ecologist, and particularly in dealing with threatened species. It represents a low-cost method for extrapolating empirical findings to a wider context which can often be performed quickly, as compared to experimentation which can take up valuable resources or may be ethically controversial. Predictive modelling is a vital part of an adaptive management strategy, both taking from and feeding into the process of active management and passive experimentation to help enable timely and effective conservation in challenging circumstances. This study examined the case of the Tasmanian devil (Sarcophilus harrisii) and its disease, known as Devil Facial Tumour Disease (DFTD), from a modelling perspective. The devil provides a classic example of a threatened species for which conservation is urgent and the consequences of the threat are potentially devastating to both its survival and the health of the wider ecosystem. In such a scenario, effective allocation of resources into the management strategies with the best chance of success is vital.Modelling was undertaken in two sections, the first studying population dynamics on a local, closed-population scale and the second looking at spatial dynamics across the devil's range, namely the main island of Tasmania. First, a compartmental ODE model was developed and then mathematically analysed in detail using a Dynamical Systems approach. The steady states of the system were calculated and their stability analysed. Mathematical descriptions of the bifurcation points between these steady states were found based on the bifurcation parameter ˜ìvÖ, the measure of removal rate. The model was also studied in relation to an unfolding parameter k, the measure of the disease latent period. The model's behaviour was found to be biologically reasonable. Findings indicated the removal effort theoretically required for successful disease suppression, as well as the range of values for latent period whereby host extinction would not occur, given model assumptions. These values appeared not to be realistic for devils, suggesting that as modelled, DFTD is capable of threatening the Tasmanian devil with extinction. A suite of compartmental models based on this work was then developed and used to analyse the disease suppression strategy that had been trialled on the Forestier Peninsula in Tasmania's south-east. Predictions from the model demonstrated that removal of infected animals, while more successful in suppressing disease when performed regularly, was unlikely to be effective in the long term under current practical constraints. The second section of the study began with the use of statistical modelling techniques such as Boosted Regression Trees and Monte Carlo analysis to estimate the mean abundance of the Tasmanian devil prior to the emergence of DFTD. From this analysis, a map of devil abundance across Tasmania and the first published estimate of overall pre-disease abundance were generated. The estimate was significantly lower than previous informal estimates. This information was then used to generate a spatial model of host-disease dynamics using a reaction-diffusion model. A Bayesian Markov Chain Monte Carlo (MCMC) analysis was run on longitudinal data from populations where data were collected both before and during disease to estimate the value of model parameters at each site, and thus determine which parameters are likely to be spatially heterogeneous. The reaction-diffusion model was then fitted to data in order to provide an estimate of the pattern of the disease's spread. Though results using only trapping or only spotlighting data were unrealistic, results incorporating both datasets together looked more reasonable. No conclusive evidence was shown to point to the location of the disease origin, which remains an open question. The addition of abundance and prevalence data from different sites in future work may help the model to better fit the true pattern of disease spread. This study has demonstrated, using both novel and established techniques, that effective and informative modelling is possible using limited or disparate data, by applying these methods to the case of the Tasmanian devil and DFTD. These techniques and future work will hopefully aid in conservation efforts for this and other species.