Statistical methods for identifying unobserved structure in complex ecological and environmental data

点击翻译按钮获取中文摘要

Due to the advancement of sensor technology, we can now remotely monitor ecological and environmental systems at fine temporal scales (e.g. multiple times a second to every few hours) with ease, over extended periods of time and space. This advancement comes at a crucial point as our environment undergoes massive disruptions due to climate change and anthropogenic influences. However, the technology and data that can be collected has outpaced the development of statistical methods designed to analyze it. A common goal in the analysis of sensor data is the identification of ecologically and environmentally relevant latent structures, for which the classes of Markov-switching, hidden Markov models, state-space models, and spatial models provide rich frameworks. In this proposal, I aim to (i) advance Bayesian Markov-switching models for ecological and environmental data, (ii) develop generative models for ecological processes over time and space, and (iii) develop statistical learning approaches for classification of video and movement data collected from animals. The HQP involved in the three objectives will form part of my research group, "Bayesian Ecological and Environmental Statistics (B.E.E.S.)", where the focus will be on statistical development of methodology that tackles pressing ecological and environmental problems and will form part of an encouraging and supportive community composed of myself and their peers. I am fully committed to building and empowering a cross- and interdisciplinary research group that advocates for the success of all, in particular by providing opportunities for those who are visible minorities, Indigenous or part of another historically underrepresented group.