Science deficiency in conservation practice: the monitoring of tiger populations in India

1. What they did

For about 30 years, Indian wildlife managers counted tigers with one method, the “pugmark census”. During a 1 to 2 week period, thousands of forestry department staff spread out across India to find tiger tracks and take plaster casts or tracings of the left hind paw print of nearly every tiger in the country. Officials then compared the prints to tell individual tigers apart, using shape, measurements and local knowledge, and added them up into national tiger numbers.

The authors, a group of tiger biologists and conservationists, asked whether this method holds up under modern ideas of how to sample animal populations. They:

• Described a general framework for counting animals, built on two problems: most surveys cannot cover the whole area (spatial sampling), and they cannot detect every animal or sign in the area they do cover (observability).
• Compared the biological ideas behind the pugmark census with what field studies have shown about how tigers use space.
• Set out the statistical assumptions the census depends on, and used published studies, tests and their own field observations to show where these fail.
• Proposed three monitoring goals and alternative methods to meet them.

The paper looks only at the pugmark census, not at tiger track surveys in general.

2. What they found

The census assumes it sees everything

The method treats its counts as complete: all of the area searched, and all tigers found. It does not record search effort or estimate how much was missed. The authors list five assumptions it must meet: (1) all tiger habitat in India is covered; (2) all four paw prints of every tiger are found; (3) the same hind print of each tiger is lifted from good, comparable ground or standard track-plots; (4) thousands of staff record each print’s shape without distortion; and (5) supervisors correctly sort the prints into individual tigers. If the first two fail, tigers are missed; if the third and fourth fail, tigers may be counted twice; if the fifth fails, the count can go either way. The final count may then have no predictable link to the real number of tigers.

The biology behind the census does not match tiger ecology

• The census assumes that adult tigers keep others of the same sex out of their areas, so the same tiger’s tracks are unlikely to turn up in different counting units. Earlier census writers even suggested tigers live in pairs.
• Field studies show tigers do not form pairs. Neighbours use the same travel routes, ranges overlap a lot in some areas, an adult male’s range overlaps those of two to six breeding females, and transient “floaters” of both sexes move across territories. Denser populations have more transients, not fewer.
• Earlier work by Karanth found census results did not follow prey densities and showed unrealistic long-term growth rates.

Where the assumptions break down in the field

• Coverage: India has over 300,000 km² of potential tiger habitat, but only an unknown part is searched well. In the 1996 census, about 300 km² of Namdapha Tiger Reserve’s 3000 km² was covered. In Nagarahole and Bandipur, teams of three or four people walking 15 to 20 km a day are assumed to search blocks of 10 to 15 km² fully. Logistics, security and staff shortages limit coverage.
• Finding every tiger’s tracks: this is unlikely except in a few reserves with many roads and good ground. In Nagarahole, with 12 tigers per 100 km², tracks could not be found on many tiger travel routes because of the ground, although radio-tracking and camera traps showed heavy tiger use of them.
• Ground for good prints: clear prints need firm ground with a layer of dust or sand, which is missing in most areas, so staff lift prints from poor ground. Artificial track-plots (“Pug Impression Pads”) often fail: no suitable soil nearby, soil hard to carry in, and rain, wind or animals spoiling them. In Namdapha in 1996 the plots were made of the wrong soil, and in the Sundarban delta, which lacks suitable ground, plots are not used at all.
• The same paw: without clear prints of all four paws, staff cannot pick the same hind print for each tiger, and often lift prints of different paws of the same tiger in different places.
• Recording and sorting prints: different people tracing the same print can produce quite different shapes. In a blind test by Karanth (1987) with prints of captive tigers from different ground, even experienced census staff failed to sort the prints into the right individuals. Later captive tests on standard ground (Riordan 1998; Sharma 2001) suggested that experts, or statistical methods, do somewhat better when fewer than about 12 tigers are compared. The authors stress that such controlled tests do not solve the wider problems of a nationwide field census.

Consequences

• Because sign locations were not mapped and effort was not recorded, there are still no country-wide maps of where tigers live, and the old counts cannot even be used as simple indices of change.
• In 1994, managers reported an increase in tigers despite growing evidence of poor protection and more poaching. Project Tiger then cut the tally by 750 tigers to create a sense of crisis. Recent censuses again report rising numbers.
• The authors estimate that in 1999 alone, Indian governments spent about 10 million US dollars on tiger conservation, with another 1.75 million from non-governmental donors, but without reliable monitoring no one can judge whether this money worked.

3. What they concluded

• The pugmark census rests on complex, unstated assumptions that are not met, and after three decades and large investment, tiger monitoring in India has basically failed. Invalid data can do more harm than having no data.
• The census aims straight at the hardest goal, counting every tiger in the country, and ignores two simpler goals that would meet most management needs.
• They propose three goals, each with suitable methods:
• Where tigers are (national level): map tiger presence or absence every year, by turning the census into surveys that simply record tiger tracks and other signs, with a proper sampling design, recorded effort and map locations, and no attempt to identify every tiger from its prints.
• Trends (each reserve): yearly indices of relative abundance from sign encounter rates, such as tiger track sets or scats found per 10 km walked, the share of 1 km trail sections with tiger tracks, or fresh tracks crossing a boat or vehicle path. With standard methods and enough effort, these can follow changes in tiger numbers, are not subjective, and can be repeated.
• Absolute numbers (a few priority sites): skilled staff and advanced methods, such as camera-trap capture-recapture for tigers and line transect surveys for prey.
• The central challenge is to decide what level of detail is needed and then use the people and equipment available within a sound sampling framework.

4. How tracking was or was not used

The paper is about a tracking-based method. It reviews the pugmark census, in which forestry staff searched for tiger tracks and lifted plaster casts or tracings of hind paw prints, and supervisors tried to identify individual tigers from footprint shape, measurements and local knowledge. The authors do not criticize track surveys as such: their proposed alternatives for mapping distribution and following trends rely on recording tiger tracks and other signs, such as scats, within a sampling design that records effort and location.

5. Tracker or observer certification

Census work is done by thousands of forestry staff “with varying levels of skills”; no certification is reported. The paper cites a blind test (Karanth 1987) in which even experienced census staff failed to sort captive tigers’ footprints into the right individuals, and captive tests in which trained people or “experts” did somewhat better with small numbers of tigers on standard ground. The authors note that differences in how people trace the same print can be reduced with well-trained people in controlled trials, but not in large field censuses.

6. Other technologies, skills or innovations

• Plaster casts and tracings of footprints, and artificial track-plots (Pug Impression Pads).
• Multivariate statistical methods to tell individual tracks apart (cited).
• Radio-telemetry and camera traps, which showed tigers using travel routes where no tracks could be found.
• Camera-trap capture-recapture and line transect surveys, proposed for estimating tiger and prey densities at priority sites.
• Geo-referencing of sign records, proposed for mapping distribution.

7. Implications for conservation (from the paper)

• Tiger monitoring methods must be practical, but also scientifically defensible.
• Without reliable monitoring, it is impossible to judge whether large investments in tiger conservation are working.
• The results of the pugmark census led to poor conservation practice, such as reporting increases while protection was weakening and poaching was growing.
• Methods should be matched to the question: sign surveys for distribution, sign encounter indices for trends in reserves, and camera trapping with prey surveys for densities at a few key sites.
• The pace at which conservation practice takes up new science may be too slow to help species recover.