Spatially Discrete Probability Maps for Anti-Poaching Efforts
Alta de Waal, Department of Statistics, University of Pretoria, email@example.com
Recorded at the 5th Annual BayesiaLab Conference in Paris on September 29, 2017.
In the past, spatial statistics applications on wildlife has focused on counting and modelling, in order to get the best possible estimate out of the collected data. A common approach in particular is point patterns and their modelling.
The objective of this study, however, is to generate spatially discrete probability maps that can be used for anti-poaching efforts. These maps should identify areas with high poaching risk for the next period of time.
The principle of the modelling effort is to calculate relative probabilities for cells in a spatial map, whilst incorporating causal aspects that relate to these probabilities. A probabilistic graphical model (PGM) is used to define conditional dependencies between variables. The PGM approach is particularly suitable to this complex problem. Its mathematical formulation allows for the inclusion of additional information apart from historical data such as context variables, latent variables and explicit conditional independencies.
The method is as follows: For each cell, all the relevant covariates are calculated. Then, an instance of the model is created and a Bayesian inference process is executed using these sets of covariate values. The output of the model is the probability of a poaching event, and is calculated for each cell. These probabilities are re-normalised to sum to unity over the whole map, thus resulting in a spatial discretised probability heatmap.
The case study to be presented is rhino poaching in the Kruger National Park in South Africa.
Alta currently holds a senior lecturer position in the Department of Statistics, University of Pretoria, South Africa. Previously, she was employed at the Council for Scientific and Industrial Research (CSIR). She received her PhD in Engineering Science from the North West University in South Africa. Alta’s research focus is on probabilistic graphical models with a special interest in spatial applications. Current research projects include wildlife protection and green security games.