PREDICTING HABITAT SUITABILITY OF MIGRATORY SHARKS USING MACHINE LEARNING METHODS
Whale shark (Rhincodon typus) populations have declined significantly over the last century due to anthropogenic mortality. Concerns about the sustainability of known populations and their interactions with humans have generated a high level of interest in the movement and migration patterns of the...
| Main Author: | |
|---|---|
| Format: | Text |
| Language: | unknown |
| Published: |
DigitalCommons@URI
2023
|
| Subjects: | |
| Online Access: | https://digitalcommons.uri.edu/theses/2312 https://digitalcommons.uri.edu/context/theses/article/3282/viewcontent/Daye_uri_0186M_13062.pdf |
| Summary: | Whale shark (Rhincodon typus) populations have declined significantly over the last century due to anthropogenic mortality. Concerns about the sustainability of known populations and their interactions with humans have generated a high level of interest in the movement and migration patterns of the ocean's largest fish. Despite their seasonal aggregation at locations across the globe, little is known about whale shark movements and habitat use away from these locations. We tracked 26 whale sharks from the male-dominated aggregation near Isla Mujeres, Mexico using SPOT (Smart Position and Temperature) tags. One mature female – Rio Lady – generated location transmissions for nearly 1,500 days, over a distance of more than 40,000 km, revealing consistent seasonal migrations within three regions of the Gulf of Mexico (GOM) across four years. Tracks of 26 predominantly male sharks revealed three distinct behavioral phases of movement and habitat use similar to those of Rio Lady. State-space modeling (SSM) and move persistence modeling (MPM) were used to generate continuous tracks and to identify areas of concentrated movement. Movement data was combined with environmental data to construct habitat suitability models using machine learning (ML), which predicted areas of high use throughout the GOM, Caribbean, and Western North Atlantic, based on observed whale shark move persistence values and their associated environmental conditions. The combination of these techniques with Argos-derived location data has provided substantial insight into the long-term movement patterns of whale sharks and shows promise for identifying other areas of high use away from known aggregation sites. |
|---|