News Release

AI-based satellite count of migrating wildebeest

Peer-Reviewed Publication

PNAS Nexus

Wildebeest 1

image: 

Wildebeest comprise 3–4 pixels in length and 1–3 in width.

view more 

Credit: Duporge et al. Satellite imagery from Maxar Technologies.

An AI-powered satellite counting effort conducted over two years concludes that less than 600,000 wildebeest migrate across the Serengeti-Mara ecosystem each year—half of previous estimates from manned aircraft surveys. Lions, hyenas, crocodiles, and tourism professionals all rely on the annual migration of wildebeest (Connochaetes taurinus) which transit through Kenya and Tanzania as the herds follow the seasonal growth of grass. Prior estimates of the migratory wildebeest population that make the trek each year were reached by extrapolating counts from aerial photos in the south of the ecosystem. Isla Duporge and colleagues use satellite images from 2022 and 2023, at 33–60 cm resolution, which were then fed to UNet, a pixel-based segmentation model, and YOLO, an object-based detection model. Each wildebeest is represented by just 6 to 12 pixels. According to the authors, the study demonstrates the feasibility of conducting population counts of terrestrial mammal populations using satellite imagery and machine learning.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.