About This Project

Due to the recent decline in Florida manatee populations, accurate population assessments are crucial. Current aerial surveys overlook undetectable manatees and don't track movements between wintering sites. Prior studies suggest manatees might have unique vocalizations, so we'll use data from rehabilitation centers to determine if we can reliably identify and count manatees based on their calls.

Ask the Scientists

What is the context of this research?

In past years, the Florida manatee population has declined, and it is expected to continue as manatees face a number of anthropogenic and environmental challenges.Manatee survival depends on wintering sites, so it's important to understand both the number of manatees using these sites and the manner in which they use them. Currently, manatee populations are estimated based on aerial surveys that tend to undercount manatees and do not provide information about movement between the sites. Since previous research indicated that manatees might have individually distinct calls, we propose to use these calls as evidence of the presence of a particular manatee at the site. We plan to train a deep learning model to identify and match these distinct calls both at the site and between sites.

What is the significance of this project?

This project seeks to establish an alternative way to identify individual manatee presence at the wintering site and to learn more about the potential utilization of different wintering sites by individual manatees. This is a non-invasive way to collect data and could be useful as an alternative (or complementary) method of counting manatees because it doesn't rely on visual observations, which can be challenging in several wintering sites with very turbid waters.
Moving between wintering sites might also mean that manatees are particularly vulnerable during transit.Recognizing this could help establish additional conservation measures to protect them, as manatees rely on wintering sites for their survival.

What are the goals of the project?

The project's goal is to develop a probabilistic model capable of counting manatees based on their distinct calls. This is grounded in the hypothesis that distinct calls can reliably indicate individual manatees, thus aiding in their counting and tracking. The model will have two components: a deep learning model to match calls within our dataset, and another model that links distinct call types to the number of manatees using Bayes' Theorem. We will validate the developed model on previously unseen acoustic data, asking it to estimate the number of manatees present in the pool at the time of recording (this value is known and can be compared to the model's estimations).