Bayesian Modelling and Software Development for the POST Project
The Pacific Ocean Shelf Tracking (POST) project is part of the Census of Marine Life Study. In this project, acoustic transmitters are surgically implanted into salmon and the salmon are tracked during their migration over a series of listening lines placed along the ocean floor.
At the moment, researchers observe the simple descriptive statistics at different locations based on the actual number of radio detections. However, these methods are not sufficient to study their movement patterns and we need to employ advanced mark-recapture models for better understanding of the movement patterns. Estimating survival probabilities of animals at different locations is a key component in mark-recapture studies. In the POST project, detection probabilities at listening lines are also important.
In our project, we develop a Bayesian model for estimating detection probabilities, survival probabilities, log travel times and the correlation structure that is well suited for the POST project. Previous mark-recapture models do not make any adjustments in survival probabilities between listening lines for their travel times whereas our model treats survival probabilities as a function of travel times. This plays a key role when distances between listening lines vary greatly.
The model is implemented via Markov Chain Monto Carlo (MCMC) using WinBUGS. Simulation results indicate that the model is well behaved in estimating parameters.
This type of interdisciplinary work is a hallmark of our program in Applied Statistics at Simon Fraser University. For more information, please contact Saman Muthukumarana (firstname.lastname@example.org) or his supervisor Tim Swartz (email@example.com), Department of Statistics and Actuarial Science.