Monitoring Carnivore Populations at the Landscape Scale: Occupancy Modelling of Tigers from Sign Surveys

1. What they did

Tigers have lost about 93% of their range in two centuries, and their future depends on about 40 “source populations” where they breed above replacement level. Inside reserves, researchers can count tigers with camera traps or DNA from scats, but across whole landscapes they usually rely on finding tracks and scats. The problem is that not finding signs does not prove tigers are absent, so simple presence-versus-absence surveys underestimate where tigers live.

The authors set out to design a practical sign survey that corrects for missed signs, and to test what decides whether tigers are present: they expected prey numbers and protection from human disturbance to matter most.

• The study area. The Malenad-Mysore Tiger Landscape in Karnataka, India: about 38,000 km² of forest mixed with farms and plantations, with 21,167 km² of potential tiger habitat. It was thought to hold about 20% of India’s wild tigers.
• The grid. They divided the landscape into 205 square cells of 188 km² each, larger than the largest expected tiger home range (non-resident males, about 80 to 90 km²), and surveyed every cell.
• The walks. Teams of 2 to 3 skilled trackers walked forest trails from morning to dusk in the dry season, up to 40 km per cell (less where there was less habitat). Tigers often travel 1 to 20 km a day along forest trails, leaving tracks and the occasional scat. The teams photographed, located and recorded signs of tigers, leopards, dholes, prey species and livestock, recording only signs they could identify without doubt.
• The analysis. Each 1 km of trail became one “spatial replicate” (a repeat sample). Because a tiger may walk more than 1 km along a trail, finding signs on one kilometre makes signs on the next more likely. The authors compared a standard occupancy model with a newer one that allows for this, and tested the effect of prey signs, livestock signs (a measure of human disturbance) and the amount of habitat in each cell.

2. What they found

Survey effort

• 2,021 person-days, 4,174 km walked, and 403 tiger signs found: 223 scats and 180 sets of tracks.

Which model worked

• Signs on neighbouring kilometres were clearly not independent. The model that allows for this fitted far better than the standard model (a difference in AIC of 184.9).

Detection and occupancy

• The chance of finding tiger signs on a 1 km section, where tigers were present, was only 0.17 (17%).
• Tigers occupied an estimated 66% of the potential habitat: 14,076 km² (standard error 1,814 km²) of 21,167 km².
• The simple presence-versus-absence approach gave only 7,537 km², underestimating occupancy by about 47%.

What drives tiger presence

• More prey (all prey, and large prey: gaur, sambar, chital and wild pig) and more available habitat went with a higher chance of tigers being present.
• More livestock signs, the measure of human disturbance, went with a lower chance.
• These covariates had a strong effect on how easily signs were found, but the evidence for their effect on occupancy itself was weaker. All effects were in the predicted direction.
• Maps of occupancy probability picked out the source populations at higher tiger densities and matched known differences in tiger density.

3. What they concluded

• Landscape-scale sign surveys can efficiently show where large carnivores live and what governs their presence, as long as both the ecology and the way signs are detected are modelled.
• Ignoring missed signs can badly underestimate tiger distribution. The authors found it puzzling that other recent reports suggested sign surveys can estimate tiger distribution reliably without this correction.
• Designing surveys around the animal’s biology (home range size, movement along trails) and field logistics (distance walked, coverage of each cell) paid off. Three covariates (habitat, prey and human disturbance) were enough to explain the pattern of tiger distribution.
• The results support earlier findings that prey depletion and human disturbance drive local tiger extinctions. Tigers persisted better in this densely populated landscape than in larger, more remote forests elsewhere in India, apparently thanks to a few well-protected source populations.
• The method could be combined with camera-trap density estimates to estimate tiger numbers across large landscapes, and could suit many other wide-ranging carnivores whose signs are fairly easy to find.

4. How tracking was or was not used

The whole survey rested on tracks and signs identification. Teams walked forest trails and recorded fresh tiger tracks and scats, plus signs of leopards, dholes, five prey species and livestock. Only signs identified without doubt were recorded, and each was photographed and its location recorded. Each type of sign was logged once per 100 m of trail. No trailing was reported: teams followed trails, not individual animals. The design used what is known of tiger movement: tigers travel along forest trails and leave tracks and occasional scats as they go.

5. Tracker or observer certification

Reported but not validated. The survey teams are described as “skilled trackers” and “trained observers”, but no certification or test of their identification skill is reported. The field surveys were led by named team leaders, and many field assistants and volunteers took part. All results rest on these sign identifications.

6. Other technologies, skills or innovations

• A spatially replicated survey design: 1 km sections of a continuous trail used as repeat samples, in place of repeat visits over time, chosen mainly for practical reasons. Each cell was surveyed within about 12 to 36 hours.
• The Hines et al. (2010) occupancy model, which allows for the fact that signs on one stretch of trail make signs on the next more likely, run in program PRESENCE.
• Livestock signs used as an easily detected measure of human disturbance and poor protection.
• Maps of the probability that each cell is occupied by tigers.

7. Implications for conservation (from the paper)

• About 70% of wild tigers now survive in source populations on just 6% of their habitat, and these sources have so far been identified from anecdotal evidence. This survey approach can identify them quickly and objectively, so conservation effort can be targeted.
• Tiger presence depends strongly on prey density and protection, which cannot be seen by satellite, so landscape analyses without field data have limits.
• Combining intensive camera-trap surveys inside reserves with cheaper landscape-wide sign surveys offers a way to monitor tiger metapopulations. Repeated over time, the surveys could track range contractions or expansions.