Why Counting Elephants Is Far Harder Than Counting Tigers

A tiger patrols a defined home range and leaves behind a signature: paw prints, scrape marks, scent sprays, and above all, a face that camera traps read with the same logic as a fingerprint. India's camera-trap grid now covers more than 141 tiger reserves and produces individual ID matches at scale. The 2022 census counted 3,167 tigers with a confidence interval narrow enough to call it a real number.

Elephants do not work this way. An adult bull in musth can cover 300 kilometres in a week, crossing reserve boundaries, state lines, and international borders with the same indifference. A family herd moves seasonally across a landscape that no single forest department administers. Camera traps cannot tell one elephant from another with the reliability they manage for tigers, ear tears, tail shape, and tusk asymmetry help, but only when the animal is cooperative enough to stand in frame at the right angle.


Project Elephant's primary field method is the synchronised count: trained enumerators spread across a landscape on the same day, at the same hour, and record every elephant they directly observe. The logic is that simultaneous counting eliminates the risk of counting the same herd twice as it moves between zones.


The problem is that an elephant in dense sal forest in Uttarakhand or in the Nilgiri shola is under a canopy that aerial surveys cannot penetrate and ground teams cannot fully comb. Studies comparing direct counts with DNA-based population estimates, in which dung samples are collected and individual animals identified through genetic profiling, have found that direct counts can miss between 30 and 60 percent of animals in heavily forested terrain. The synchronised count is not wrong; it is a lower bound dressed up as a total.


Dung DNA census methods, piloted in parts of the Western Ghats, are more accurate but prohibitively expensive at the scale of a national count. India has roughly 30,000 elephants across five broad population clusters, the Northeast, Eastern India, Central India, the Northwest, and the South. Running a statistically robust DNA survey across all of them simultaneously remains beyond current operational capacity.

A population number drives everything downstream: how much corridor land needs legal protection, how many conflict mitigation teams a state deploys, how compensation frameworks are calibrated when an elephant kills a farmer's crop or, far worse, a farmer. When the number is uncertain by a margin of several thousand animals, those decisions rest on softer ground than anyone in the system likes to acknowledge.

The elephant-human conflict figures are not uncertain in the same way. The Ministry of Environment, Forest and Climate Change reported that elephants killed more than 500 people annually across conflict-prone states in recent years, with Odisha, West Bengal, Jharkhand, and Assam consistently at the top. The animals most likely to be involved are bulls in musth and family groups whose traditional movement corridors have been severed by railway lines, highways, and plantation agriculture. Knowing the population size more precisely would help identify which corridors are genuinely critical and which protection investments would reduce that toll.

Elephants are cognitively complex enough to learn from experience with humans. In areas of persistent conflict, they become nocturnal, shifting their movement to hours when camera traps are less likely to be checked and human enumerators are not in the field. Research from the Anamalai Hills and from Assam's Kaziranga fringe has documented herds that altered their routes within a single season after an encounter with a census team.

Their social structure adds another layer. A herd is not a fixed unit. Adult males are largely solitary, joining herds briefly. Young males get pushed out of natal herds and form loose bachelor groups that drift across enormous ranges. A count that captures the matriarch-led family unit may entirely miss the bulls, who are also the animals most frequently in conflict with people and most frequently killed in retaliation.

Elephants also live long enough, 60 to 70 years in the wild, that a population's trajectory only becomes legible over decades. A count that is off by 20 percent in a given year may not register as a crisis until the error compounds across three or four census cycles.

Acoustic monitoring offers one partial answer. Elephants produce infrasound calls below 20 Hz, inaudible to humans, that travel several kilometres through forest. Arrays of buried sensors can detect and triangulate these calls, giving a presence-absence signal across a landscape even when no animal is visible. The Wildlife Institute of India has been developing and testing this approach in several reserves.

Satellite-collar data, combined with movement modelling, can estimate how many animals are likely to be in an uncollared population based on the ranging patterns of collared individuals. Neither method gives a clean headcount. Both give better information than a single-day direct observation in dense forest.

The honest version of where Project Elephant stands is this: the count is improving, the methods are diversifying, and the number the government publishes every five years is more useful as a trend indicator than as a precise census. For a species that crosses three countries, ignores administrative boundaries, and learns to avoid the people trying to count it, that may be the most that ground truth allows.