A new study led by researchers at the University of Oxford has used artificial intelligence and satellite imagery to produce revised estimates of migratory wildebeest populations in the Serengeti-Mara ecosystem. The findings challenge long-standing figures, which have been based on manned aerial surveys since the 1970s.
Dr Isla Duporge, lead researcher from Oxford's Wildlife Conservation Research Unit, said: "The field of wildlife conservation relies on having accurate data on wildlife population numbers. Combining earth observation satellite data with deep learning, this study has revolutionised our understanding of migratory wildebeest numbers, and could open the floodgates for surveying other species using this method."
Traditionally, estimates relied on aerial surveys that photograph herds along specific flight paths and use statistical models to extrapolate overall numbers. This approach can introduce errors if animals move between survey lines or are unevenly distributed.
Satellite-based surveys offer broader coverage without disturbing wildlife or risking human safety. However, counting animals in these images manually is not practical due to the large volume of data.
In their research, Dr Duporge’s team trained two deep-learning models—U-Net and YOLOv8—using a dataset containing more than 70,000 manually labeled wildebeest. The models achieved F1 scores up to 0.83 when tested for accuracy.
The AI was then applied to over 4,000 square kilometers of high-resolution satellite imagery covering Tanzania’s and Kenya’s Masai Mara National Reserve. These images were taken in August 2022 and 2023 by Maxar Technologies Worldview-2 & 3 satellites at altitudes between 617 and 770 kilometers above Earth.
Results from both AI models indicated populations ranging from about 324,000 to nearly 338,000 in 2022, and from about 503,000 to over 533,000 in 2023. These counts are roughly half a million fewer than previous estimates of around 1.3 million wildebeest—a figure that has remained largely unchanged for decades.
Dr Duporge commented on these results: "The sheer difference between traditional estimates and our new results raises questions about where the ‘missing’ wildebeest might be. Based on data from GPS tracking surveys, we are confident that most of the herd were contained within the surveyed area. And whilst some individuals may have been obscured by tree cover, it seems unlikely that such a large number—on the order of half a million—would have been concealed in this way."
The researchers noted that lower counts do not necessarily indicate a recent population collapse; migration routes may have shifted instead. Nevertheless, threats such as habitat fragmentation caused by agriculture and infrastructure development—as well as climate change affecting rainfall patterns—are putting pressure on wildebeest populations. Accurate numbers are considered important for guiding conservation strategies.
Professor David Macdonald (founder of the Wildlife Conservation Research Unit at Oxford) said: "The most basic fact to know as a foundation for conserving any species is how many of them there are. The technological breakthrough of our study - satellite-based wildlife monitoring, powered by AI - potentially revolutionises the answer for wildebeest, besides opening up incredible possibilities for monitoring other large species."
This work builds upon earlier efforts where similar technology was used to count elephants via satellite images but marks its first application for surveying widely dispersed mammal populations rather than isolated groups. The research team suggests this technique could also help monitor other herd animals such as reindeer or camels; they are now working on adapting it for African rhinos.
Researchers involved in this project came not only from Oxford but also from Princeton University, The University of Hong Kong, University of Twente, University of British Columbia, Army Research Office (Durham NC), and Sun Yat-sen University.
The code developed for this study has been made available online. The full findings appear in PNAS Nexus under the title “AI-based satellite survey offers independent assessment of migratory wildebeest numbers in the Serengeti.”