One team counted 172 sangai. Another counted 64. Same park.

In February 2026 the Manipur forest department shut Keibul Lamjao National Park for a week. Enumerators climbed bamboo platforms at first light, waited for the deer to come out and graze, and wrote down what they saw. The report came out in July at a Van Mahotsav function in Imphal: 172 sangai in the park's 40 square kilometres, and 375 hog deer sharing it with them.

Three years earlier, researchers from the Wildlife Institute of India who monitor the same population had arrived at a mean estimate of 64.

Neither figure is fabricated. Both come from people who went out to the same 40 square kilometres and tried to count the same animal. The gap between them is not really about arithmetic. It is about what the sangai stands on.

The ground is not ground

Keibul Lamjao is the only national park in the world that floats. Its surface is phumdi, a mat of roots, humus and dead vegetation with living reeds growing out of it, lying on the water of Loktak lake. About one fifth of the mat sits above the surface. The other four fifths hang below it.

The naturalist E.P. Gee, who surveyed the place for the International Union for Conservation of Nature in 1959 and 1960, put the problem plainly in his report.

It must be explained at the outset that this is no ordinary swamp. It is a floating swamp.

He measured phumdi from six inches to about five feet thick. Where it is thick you sink to your knees. Where it is thin you go straight through into the lake. Reeds grow to fifteen feet, so a person standing on the mat can see nothing but the grass in front of their face.

The deer is built for it. The pasterns of the Manipur subspecies, the joints just above the hoof, are hairless, hard and horny, which spreads the load and stops the animal punching through. Gee noted that the Burmese and Thai subspecies, which live on dry undulating country, have hairy pasterns instead. The adaptation that makes this deer distinctive is in its feet.

Its walk is distinctive too. Gee quoted an earlier sportsman, Lieutenant Colonel H.S. Wood, describing it:

No other animal could exist in such swamps. The action of this deer is very peculiar. Instead of bounding with fore- and hind-feet coming on to the ground at the same time, it seems to be proceeding on its hind legs, the body being held almost vertical. This action prevents it from sinking in the quagmire.

That high, hesitant gait is where the English nickname comes from. Manipuri dance is often said to borrow from it. The Meitei name has a gentler sense: Gee recorded sangai as meaning "the animal that looks at you".

The letter that killed the sanctuary

In December 1951 the Government of Manipur wrote to Gee, who was then honorary secretary for the eastern region of the Indian Board for Wild Life. He quoted the letter in full in his report, and it is worth reading twice.

Enquiries have been made by the Forest Department and it has been found that this deer has become totally extinct and no alive specimen is now available.... In view of what is stated above there is no need to establish a Wild Life Sanctuary at present.

The extinction was the argument against protection. Gee passed the report on to the IUCN, and the Manipur brow-antlered deer went onto the international record as gone.

It was not gone. Gee wrote that he, Lieutenant Colonel R.W. Burton and P.D. Stracey, then senior conservator of forests in Assam, "continued their efforts to obtain further information and to protect the deer in the event of their being found". They were found. In October 1953 the whole of Loktak was closed to shooting and declared a sanctuary. That was quickly judged excessive, since it deprived duck shooters of their sport, so in July 1954 the lake was reopened except for the southern portion where the deer lived, about 20 square miles. By the time it was surveyed properly in 1959 the sanctuary had shrunk to 10.75 square miles.

You will often read that Gee rediscovered the sangai in 1953. His own published report does not say that. He was the man told it was extinct, he relayed the news, and he did not set foot in the place until October 1959. The "six deer" figure that gets attached to the rediscovery is not in his report either.

Where the first number came from

What is in his report is the first attempt at a count, and it explains a great deal about everything that followed.

In March 1960, at the driest time of year, Gee arranged for twenty Manipuri villagers to drive a section of the sanctuary towards the hillock he was watching from. The men stripped to the hips and waded in with sticks to probe for soft patches. They covered about a quarter of a square mile in two hours, pushed out four brow-antlered deer and ten pigs, and then declined to do it again.

From that, Gee did the arithmetic. Total sanctuary 10.75 square miles, minus 2.5 grazed by buffalo, minus a square mile of cultivation, minus a quarter of hill, leaves seven. At sixteen deer per square mile that gives 112, "say 100".

This is not a criticism of Gee. He was honest about the method, he wrote down exactly what he did, and his recommendations, permanent staff, an ecological study, an end to cultivation inside the sanctuary, were sensible. He also noticed something most later accounts skip: the deer may be on the phumdi by necessity rather than by choice, pushed off the grassy land at the foot of the hills as rice cultivation spread. Its famous habitat may be a refuge, not a preference.

Sixty years of numbers that refuse to line up

The number everybody knows is fourteen. In 1975 M.K. Ranjitsinh, then a deputy secretary in the forests department, flew the first aerial census over Keibul Lamjao and counted fourteen animals. It is the figure every recovery story starts from, and it did its job: the park was notified in March 1977, land inside it was bought out, and the counts went up.

The full official series, compiled by Manipur's chief wildlife warden for a 1992 viability workshop, is more interesting than the headline. It comes with a remarks column, and the remarks are where the honesty lives.

1. 1975First aerial census. 14 animals. No breakdown by sex recorded.
2. 1986Ground count reaches 95, the highest so far.
3. 1987Aerial count gives 35. Remark: "Poor ground visibility, population under-estimated."
4. March 1992Ground count, 104.
5. April 1992Aerial count three weeks later, 62. Remark: "Population is underestimated due to poor visibility."
6. 1994Twenty blocks, twenty machans, two mornings. Total count 137. The sampling count run alongside it gives 140.

Look at 1992. The same park, the same season, the same deer, counted twice within a month, and the answers differ by forty two animals. In 1987 the aerial figure came in at barely a third of the previous year's ground figure and the department wrote down why: from a helicopter over fifteen-foot reeds, you cannot see.

So when a 2026 count and a 2023 estimate disagree, that is not a new scandal. It is the oldest fact in the file.

The two numbers we are living with

The March 2016 census put the population at 260, up from 204 in 2013. That 260 is the number that escaped into the world. It is on tourism pages, in press releases, in international summaries, and it has been repeated for a decade as the current state of the sangai.

The Wildlife Institute of India has been producing a different series. Presenting at WII's annual research seminar, Mirza Ghazanfarullah Ghazi gave mean estimates of 90 animals in 2006, 88 in 2007, 92 in 2008, 76 in 2019 and 64 in 2023. The rate of decline had gone from about 1 per cent a year in 2019 to about 1.9 per cent a year by 2023. Then the February 2026 state census came in at 172, well below the 260 that had been circulating but nearly three times the researchers' figure.

The genetics is not ambiguous in the same way. In 2017 a team from WII, led by Sangeeta Angom, sequenced the mitochondrial control region of 30 sangai. They found exactly one haplotype. Not low diversity. None. The related Burmese and Thai subspecies showed plenty. The authors concluded the population had passed through a bottleneck severe enough to leave it exposed to inbreeding depression.

The other finding in the 2026 census was reported as news: hog deer now outnumber sangai in the park, 375 to 172, and the report recommends moving some of the hog deer elsewhere. It is worth knowing how old that problem is. Gee saw no hog deer at all on his two visits in 1959 and 1960 and guessed there might be twenty five in the sanctuary. By 1992 the forest department's own habitat notes described the hog deer flatly as the cohabitant "whose number far exceeds that of Sangai". Thirty four years later it was announced from a podium.

What is happening under their feet

Phumdi is not permanent scenery. It works on a cycle. In the dry season the water drops, the mats settle onto the lake bed, their roots take up nutrients from the sediment, and when the monsoon comes they lift off again thicker than before. Take away the fall in water level and you take away the feeding step.

That is what the Ithai barrage did. Built on the Manipur river as part of the Loktak hydroelectric project and commissioned in 1983, it holds the lake at a working level for a 105 MW station and for irrigation. Water fluctuation across the year fell from around 3.1 metres to about 1.4 metres. The mats stopped grounding.

Nobody had to wait for modern research to notice. The 1992 workshop report, written by a serving Manipur forest officer, says it in one line: the project "has broken the cyclical settling and re-floating of phumdi in the National Park".

The measurements since have put numbers on it. WII researchers have recorded phumdi at a maximum thickness of 163 cm and a minimum of 47.44 cm, degrading at around 9 per cent a year. An adult sangai weighs 90 to 115 kg and needs roughly 0.8 to 1 metre of mat under it to be held up. Thick phumdi cover across the wetland has fallen from 9.82 per cent in 1977 to 6.33 per cent in 2023. Loktak was listed as a Ramsar site in March 1990 and put on the Montreux Record, the list of wetlands whose ecological character is changing under human pressure, in June 1993. It is still there.

During the 2023 survey, the team recovered the carcasses of two male sangai and four hog deer. The likely cause was drowning.

That is the sentence to sit with. This is a deer that can go through the floor of its own national park.

The second home, and the reason the count matters

The obvious insurance is a second population somewhere else, and the site has been chosen: Pumlen Pat with the adjoining Thongam Mondum reserve forest in Thoubal district, about 63 square kilometres of wetland and hillocks, assessed as the closest match to Keibul Lamjao. A conservation breeding facility has been set up near Iroisemba in Imphal West. Local communities have objected, on the grounds that turning their lake into a conservation reserve will shut them out of the fishing they live on, and they are not wrong to expect that, because it is what happened elsewhere.

Which brings the counting back into focus. The gap between 172 and 64 is not a technical squabble between survey methods. It decides whether the sangai is a recovering species that can be managed where it is, or a population small enough that the entire subspecies could be lost to one bad flood, one disease outbreak, one more decade of thinning phumdi.

The barrage is the same question in a different currency: 105 MW and irrigation for 24,000 hectares on one side, and on the other the seasonal rise and fall that used to build the only ground this animal can stand on. Manipur's chief minister asked the union government to review removing the barrage in 2017. Nothing came of it. No Indian dam has been decommissioned for ecological reasons.

Nobody has to weigh any of that seriously while the number in circulation is comfortable. The most useful column in the whole sixty-year record is the one the forest department has been filling in since the 1980s, the one that says, in the department's own hand, that the population is underestimated and the visibility was poor.