Home to the largest overland migration on Earth — and, according to the first AI-assisted satellite count of it, a migration roughly half the size everyone has been citing since the 1970s.
The Serengeti–Mara ecosystem spans the Serengeti National Park in Tanzania (14,763 km², a UNESCO World Heritage Site since 1981) and the adjoining Maasai Mara National Reserve in Kenya, together with the neighboring Ngorongoro Conservation Area (8,300 km²) — one of the largest and most studied wildlife landscapes on Earth, and the stage for the annual wildebeest migration.
A 2025 study led by the University of Oxford applied AI-assisted satellite image analysis to count the migration for the first time using this method, and found fewer than 600,000 wildebeest — roughly half the widely cited figure of around 1.3 million that had remained largely unchanged in public reporting since the 1970s.
Source: University of Oxford, 2025; Mongabay, 2025The Serengeti–Mara's ecological engine is movement: wildebeest, zebra and gazelle follow seasonal rainfall across the open plains, and predators, scavengers and vegetation cycles are all structured around that movement rather than any fixed location. That makes the ecosystem unusually sensitive to anything that blocks or narrows migratory routes — a herd that can no longer reach water or calving grounds at the right time doesn't just move less, it can decline even where the broader landscape still looks intact.
It's worth being precise about which number applies to what: the "roughly half of 1.3 million" revision is a correction to a decades-old ecosystem-wide estimate using better counting methodology, not necessarily evidence of a sudden ecosystem-wide crash — populations may have already been smaller than believed for years. The Mara-specific collapse from 100,000 to under 25,000, by contrast, is a documented decline over a known cause (fencing and land-use change) in a specific, smaller area. Both are real and both matter, but they answer different questions, and conflating them risks either understating the Mara's local crisis or overstating the ecosystem-wide one.
The Maasai Mara Conservancies model is one of the more genuinely successful community-conservation approaches documented anywhere in this series: Maasai landowners pool land into conservancies under roughly 15-year lease agreements with safari operators, receiving direct, guaranteed lease payments — currently supporting more than 14,528 landowners and totaling over $4.9 million annually, plus further income from ranger employment. Wildlife populations within these conservancies have reportedly increased across nearly all species monitored, and the model is increasingly cited as a template for community-based conservation in working landscapes beyond formal protected areas, rather than relying solely on park boundaries and enforcement.
This report carries a "Moderate, uneven by topic" Evidence Level rating because its component facts sit at genuinely different levels of certainty. The Mara-specific wildebeest collapse, the lion count and the conservancy lease data are drawn from focused, relatively recent, well-documented sources. The revised ecosystem-wide wildebeest count is new — a single 2025 study using a novel AI-assisted counting method — and while methodologically significant, it has not yet been replicated or fully reconciled with decades of prior ground-survey estimates, so it should be read as a serious, credible revision still under scientific discussion rather than a settled final number. The mid-century poaching projection is a forward-looking model, not an observed trend, and depends on assumptions about human population growth that may or may not hold.