How Do You Count Ghost Cats?

Nobody Knew

A single male can range over more than a thousand square kilometres in the course of his life. An animal like this is rarely seen and counted with difficulty. For most of the history of trying to count this species, the honest answer has been that nobody knew. That matters more than it might seem. Conservation decisions like which landscapes to protect, which populations are in trouble, whether a species is recovering or slipping away, rest on knowing roughly how many animals there are and where. When the snow leopard was moved on the IUCN Red List from Endangered to Vulnerable in 2017, the decision was contested by scientists and range-country governments, in part because of how thin the underlying evidence was. Of the sixty-nine samples that fed into the global assessment, sixty-three were based on expert opinion and sign surveys rather than direct counts. A change in a species’ official status was resting largely on estimates that could not be confirmed.

The reason good estimates were so scarce comes down to two persistent problems. Study areas were often small, which overestimates density at the site. Researchers also tended to survey the best habitats, the well-protected valleys with the most prey and cover, and those numbers were then applied across the whole landscape.

Pakistan is a great example. Early assessments, drawn from interviews and sign surveys, placed the national population between 300 and 420 animals, and the higher figure was widely repeated. Later camera-trap and genetic work in the core habitats suggested the actual number was closer to 80 to 120. The earlier estimates had extrapolated limited data across large areas, counted the same cats more than once as they moved between sites, and assumed the animals were present in places they were not.

Camera trap footage of snow leopards at night. [SLF-Kyrgyzstan]

Ghost in the Data

Even the method now considered the gold standard carries a hidden version of the same problem. Camera-trap surveys identify individual snow leopards from the spot patterns on their coats, then use how often each one is photographed to estimate the wider population. The whole approach depends on the assumption that individuals can be positively identified. For a long time that assumption went untested. 

When Snow Leopard Trust scientists later tested the method, photographing sixteen captive snow leopards on forty occasions and asking eight experienced observers to identify them, the observers misclassified one in eight capture events, and the resulting population estimates were inflated by an average of roughly one-third.

Statisticians call this a ghost: when one cat’s photos or scat samples get mistakenly split into two, the survey counts an animal that doesn’t exist. In one Nepal study, this kind of error turned 34 individuals identified from DNA into an extrapolated estimate of 144. Small identity errors, multiplied across a landscape, disproportionately inflate the headline number.

Counting Consistently

The counting has to be done carefully and consistently everywhere, so that a figure from Pakistan means the same as one from Mongolia. That is the problem the range countries set out to solve after 2017. The twelve snow leopard range states work together through the Global Snow Leopard and Ecosystem Protection Program (GSLEP). Within that framework, Snow Leopard Trust helped launch a shared approach to population assessment known as PAWS, the Population Assessment of the World’s Snow Leopards.

PAWS works in two stages. First, interviews, sign surveys, and cameras are used to map the presence of snow leopards across large areas. Then, intensive camera-trap or genetic sampling estimates abundance, with sites chosen to cover the full range of habitat quality rather than only the best. The design is overseen by a panel of statisticians and field scientists from across the range and beyond, and tested in the field. The first large-scale trial, across the entire snow leopard habitat of Himachal Pradesh in India, an area larger than every previous snow leopard population study combined, confirmed that the earlier opinion-based figure for the state had been substantially overstated.

Counting accurately is slow and expensive. A single assessment can take the better part of a year, and several rounds of training. The statistical tools are still being refined.  One recent method for identifying the snow leopards before they enter the count was developed by a group of scientists, including a scientist from the Snow Leopard Trust.

Earlier this year, the Snow Leopard Network, a community of more than 800 researchers and practitioners, published status assessments for all twelve range countries that underpin the ongoing IUCN Red List review. Six of those countries have now completed PAWS-standardized assessments, and China, which holds around 60% of the world’s snow leopard habitat, produced its first standardized national estimate.

The picture these assessments describe is uneven, with clearer knowledge in some places and evidence of local decline in others. What the assessments offer is a picture built on consistent, cross-border measurement, one that lets range countries speak about the state of the species with a confidence they didn’t have before.

Why It Matters

For Snow Leopard Trust, monitoring a population the same way over time shows whether it is stable, growing, or in decline, and whether the protections already in place are working. It flags emerging threats early and identifies the critical habitats and movement corridors that should stay connected for these wide-ranging cats to survive. Those findings feed into the work Snow Leopard Trust supports through GSLEP, the alliance of all twelve range-country governments that runs the PAWS assessment and provides robust and reliable evidence on which landscapes are designated for conservation.

When a count from Pakistan is built the same way as a count from Mongolia, the two can be compared, added, and tracked across borders. Keeping that measurement comparable across range countries is the role of the Snow Leopard Network, which published this year’s status assessments for all twelve range countries in Snow Leopard Reports. Together they give the range countries a shared, reliable basis for deciding where protection is needed most, and for confirming, over time, that it is making a difference.

Sign up for our monthly E-News to keep up with all the cats and our conservation efforts to protect them.

Photo/video credits: Snow Leopard Foundation [SLF] – Kyrgyzstan

Acknowledgements:

Ale, S. B. & Mishra, C. (2018). The snow leopard’s questionable comeback. Science, 359(6380), 1110.

Suryawanshi, K. R., Khanyari, M., Sharma, K., Lkhagvajav, P. & Mishra, C. (2019). Sampling bias in snow leopard population estimation studies. Population Ecology, 61(3), 268–276.

Johansson, Ö., Samelius, G., Wikberg, E., Chapron, G., Mishra, C. & Low, M. (2020). Identification errors in camera-trap studies result in systematic population overestimation. Scientific Reports, 10, 6393.

Chetri, M., Odden, M., Sharma, K., Flagstad, Ø. & Wegge, P. (2019). Estimating snow leopard density using fecal DNA in a large landscape in north-central Nepal. Global Ecology and Conservation, 17, e00548.

Suryawanshi, K. R. et al. (2021). Estimating snow leopard and prey populations at large spatial scales. Ecological Solutions and Evidence, 2, e12115.

Kodi, A. R., Howard, J., Borchers, D. L., Worthington, H., Johansson, Ö., Samelius, G., Low, M. & Sharma, K. (2024). Ghostbusting: Reducing bias due to identification errors in spatial capture-recapture histories. Methods in Ecology and Evolution, 15(6).

PAWS (Population Assessment of the World’s Snow Leopards) guidelines and process outline. Global Snow Leopard and Ecosystem Protection Program (GSLEP) Secretariat.

Snow Leopard Reports, Vol. 5 (2026). The Status of Snow Leopards Across High Asia. Snow Leopard Network and Swedish University of Agricultural Sciences.

Related