Background

​

Traditional methods to survey marine megafauna can be costly and logistically challenging. The National Oceanic and Atmospheric Administration (NOAA) has identified satellite-based monitoring efforts as priority area to help address these limitations. Translating satellite images into reliable measures of megafauna abundance and distribution requires developing two key areas: (1) consistent detection across large image archives, and (2) a clear understanding of how environmental conditions and animal behavior shape detection. Meeting these needs demands an integrated approach that combines computer vision with empirical behavioral data and ground-truth measurements of oceanographic and atmospheric conditions.

This project advances these efforts through broad collaboration with federal agencies, NGOs, and academic partners. We are developing automated detection methods across multiple satellite platforms and sensor types. With a team of undergraduate interns, we are systematically searching for humpback whales in the Maui Nui region and testing AI-based detection using these growing labeled datasets. In parallel, we are exploring synthetic aperture radar to enable weather-independent counts of amphibious megafauna such as seals and turtles. To translate detections into functional monitoring programs, we are leveraging the Marine Mammal Research Program’s extensive tag dataset to quantify how movement and dive behavior influence satellite detection, while designing ground-truth studies to assess how visibility and other environmental factors affect detection performance. Together, these efforts will improve the accuracy and applicability of satellite-derived abundance and distribution models, strengthening our ability to track ecosystem change and inform megafauna management.