Abstract:
|
Our data consist of position observations of 22 small sharks in a wetlands tidal basin covering 0.57 square miles. The sharks were followed over a 17-month period and more than 68,000 observations were recorded. We infer, based solely on the observations, the existence of two types of movement behavior, namely feeding and transiting. We use a sequential Bayesian particle filter to model the sharks' movement patterns, based on our paradigm of two behavior types. Our model includes parameters that are updated through sequential predictions of shark locations. Using these parameters we will answer biologically-relevant questions about the sharks' behavior: (1) Can we accurately identify the two inferred behaviors from the shark movements and model the parameters for the behaviors? (2) Is feeding behavior more likely in certain parts of the wetland? We will show that locations where movement patterns indicate feeding behavior is most common correspond roughly to actual locations of the shark prey. (3) Do sharks near each other tend to have the same behavior, beyond the regional dependency? A poster presentation will be accompanied by a computer simulation display on a laptop.
|