Natural disturbances including severe storms and bleaching events have devastating impacts on the Great Barrier Reef's health. Unfortunately, the increasing pressures associated with climate change are causing these disturbances to occur more frequently, for a longer duration and with more intensity.
It's essential to understand the recovery dynamics between major disturbances so we can manage the health of the Great Barrier Reef under increased environmental pressures. Many studies modelling reef recovery assume a specific form for the growth dynamics. However, the broad applicability of some of these assumptions has been questioned. While alternative models for reef recovery are being proposed, it's unclear to what extent standard reef survey protocols can distinguish between these models.
This project will explore and optimise the experimental design of reef monitoring for the purpose of discriminating between candidate reef recovery models.
You will develop a Bayesian experimental design problem to quantify the optimal duration between visits to the same reef. This will help discriminate between pre-specified coral recovery models. A key component of this task will include developing a utility function that relates the timing of observations to the information gained by the observation.
Once the problem has been defined, the design is optimised with respect to the utility function. This task requires the evaluation of potential, computationally-expensive expectations over the design space.
This project will provide guidelines for research groups to optimally schedule their reef revisits to better understand the dynamics of reef recovery. This deeper insight into reef recovery will better inform how we manage the Great Barrier Reef's ecosystems to mitigate the negative effects of climate change.
Skills and experience
To be considered for this project, you should have an understanding of statistical inference and model selection from a Bayesian perspective. We'd also prefer some level of programming proficiency in R or a similar language.
Contact supervisor Dr David Warne for more information