Heather Haas & Joshua Hatch
Project Title: Dive Profile Classification for Sea Turtle Ecology and Conservation
NOAA mission goal: Healthy Oceans
The Sea Turtle Ecology and Population Dynamics Group of the Northeast Fisheries Science Center (NEFSC) has worked collaboratively with federal, state, and non-federal partners to place hundreds of satellite tags on loggerhead sea turtles in the Greater Atlantic and Southeast Regions of the United States. The advancement of biologging technologies, such as satellite tags, has greatly enhanced the spatial and temporal resolution of sea turtle ecology and conservation research. Tags deployed on sea turtles not only allow for remote observations of long-distance movements (e.g., migrations), but also offer glimpses into the environmental conditions that tagged individuals encounter as they move throughout the water column. These vertical movements, often referred to as dives, mark the departure from and the eventual return to the surface as tagged sea turtles need to breathe and thermoregulate. Recorded dives from a tag consists of a series of depth and time values that allow vertical movements to be represented as two-dimensional shapes. Given that a single tag can stay on an individual for well over a year, tens of thousands of dives could be obtained. However, only a random sample of those dives ever get transmitted via satellite to conserve battery life and extend the life of the tag – except in the rare instance when a tag is recovered and the entire time-depth recording can be downloaded. Still, hundreds to thousands of dives per tag are transmitted across hundreds of individuals consisting of a very large and rich dataset. An automated method, then, is needed to objectively categorize dives into meaningful groupings by making use of the information inherent in the two-dimensional shapes of the time-depth values.
The objective of this project is to classify dive profiles obtained from satellite tags into meaningful categories using appropriate quantitative methods (e.g., unsupervised statistical learning algorithms). Once dives are satisfactorily classified by shape, a secondary objective will be to better understand the behavioral states of turtles that gave rise to those dive groupings.
Duties and responsibilities:
- Research data tables and attributes of Oracle database that house the satellite tag data
- Write custom SQL queries to pull dive data from the Oracle database
- Use a programming language (e.g., R or python) to read in dive data, explore dive data through graphics or other visualizations, and prepare dive data so that it can be ingested by an unsupervised statistical learning algorithm
- Scope statistical learning algorithms to cluster dive profiles into meaningful groupings
- Apply the scoped statistical learning algorithm to the dive data to develop meaningful groupings of dive shapes
- Work together with Center scientists to assess groupings of dive shapes to ensure they are scientifically reasonable
Special skills/training required:
- Familiarity with a programming language, such as R or python
- Develop an understanding of sea turtle diving behavior
- Develop an understanding of satellite tag data, as it pertains to sea turtles
- Develop an understanding of relational databases
- Develop an understanding of unsupervised statistical learning algorithms or other appropriate quantitative methods for clustering
- Create a dive shape attribute for the existing dive profile data table
- Potentially identify behavioral states of sea turtles that drive variability in dive groupings
Guidance and supervision:
Joshua Hatch from the Protected Species Branch, NEFSC, NOAA Fisheries will guide the intern in data acquisition, analysis, and communication of results in the form of a presentation or poster. The intern will also be co-mentored by Heather Haas, PhD.