Brown hyaenas on roads: Estimating carnivore occupancy and abundance using spatially auto-correlated sign survey replicates

1. What they did

Carnivores are elusive, so surveys often miss them where they are present. Surveys that properly account for this are seldom practical across whole landscapes. The authors adapted a newly developed method, first used for tigers (“tigers on trails”, Hines et al. 2010), to survey brown hyaenas across the agricultural land of South Africa’s North West Province.

Survey design

• They laid a grid of cells about 20 km across over the farmland, sized to exceed the expected maximum home range of most brown hyaenas, and randomly selected 25 cells (“sites”). Sites were separated by at least one cell to make it unlikely that the same animals moved between them.
• In each site, they drove un-tarred farm roads chosen to give the best chance of detecting sign. Each route was split into consecutive 1 km segments that served as repeated samples (“replicates”). So each site needed only one visit, with no extra observers or methods.
• Survey length per site (11 to 25 km) matched the amount of agricultural land in the site.
• All surveys were done over 5 months of the dry season, when sign lasts longest.

The sign survey

• Two experienced observers drove at 10–15 km per hour: one in a spotter’s chair above the front bumper, the other looking from inside the vehicle.
• They searched for brown hyaena sign on the road and in a 2 m strip on either side.
• They identified sign to species by its colour, dimensions, position and any accompanying signs.
• Each segment was coded for substrate quality: good (for example fine, shallow soil) or poor, where scat and particularly tracks were hard to detect (for example deep sand or rocky areas).

Interviews

• At each surveyed property, the owner or manager was interviewed.
• They estimated how much of the property was used for wildlife farming, and rated nine statements about carnivores on a five-point scale, giving each person an attitude score.

Analysis

• Consecutive segments are not independent: sign in one segment makes sign in the next more likely. The authors measured this dependence and joined segments into 4 km units, which came close to independence.
• Occupancy models then estimated the proportion of sites used by brown hyaenas, corrected for imperfect detection, and tested the effects of attitude, wildlife farming and substrate.
• An abundance model (the Royle–Nichols model) estimated density.
• Simulations tested how many sites and replicates a future monitoring program would need, and how much power it would have to detect changes.

2. What they found

The survey

• The team surveyed 577 km between 13 May and 10 October 2008. Brown hyaena sign was found in 85 segments at 18 of the 25 sites, a “naive” occupancy of 0.72.
• Dependence between segments: with 1 km segments, there was strong dependence between neighbouring segments. At 4 km, the dependence was close enough to independence for standard models.
• Occupancy: the model-averaged estimate was 0.748 (± SE 0.1), slightly above the naive estimate. The species may have been present but undetected at one site.
• Detection probability: 0.477 per 4 km segment. The odds of finding sign were 2.5 times higher on good substrate than on poor substrate.

Attitudes and wildlife farming

• Both attitude score and wildlife farm area had strong positive effects on occupancy.
• Sites where landowners’ attitudes to carnivores were above average had 89% higher estimated occupancy than sites below average (0.991 versus 0.523).
• Sites with wildlife farms had 57% higher estimated occupancy than sites without (0.906 versus 0.577). Having wildlife farms seemed to matter more than their area. Among sites with wildlife farms, those with an above-average farmed area had 15% higher occupancy.

Abundance and density

• The abundance model gave about 1.87 brown hyaenas per site, an average density of 0.45 per 100 km² in the surveyed cells.
• Assuming each brown hyaena used an average of three grid cells, based on home ranges in Botswana, the overall density was 0.15 per 100 km² (95% confidence interval 0.07–0.23). This is at least an order of magnitude lower than published estimates from protected areas.

Planning future surveys (simulations)

• Reliable occupancy estimates needed at least 6 replicates per site, and reliable abundance estimates at least 12.
• None of the designs tested had enough power to detect a 30% change in occupancy.
• With 12 replicates a year at 65 or more sites, a 50% decline in occupancy could be detected over 5 years, with the error rates other carnivore studies have accepted.

3. What they concluded

• Sign surveys along consecutive road segments, analysed with occupancy models that account for imperfect detection and dependence between segments, could be used to monitor carnivore populations at large, possibly even country-wide, scales.
• Combining interview data with biological data in one occupancy framework was an efficient way to bring social and biological factors into the same analysis.
• Caution on the density estimate: it rests on untested assumptions, including how many grid cells each hyaena uses and that home ranges match those in Botswana. The authors call it speculative.
• Why attitudes and wildlife farms matter: the authors suggest that positive attitudes may reduce persecution, and the deaths of brown hyaenas from lethal control aimed at other carnivores. Wildlife farms may allow natural vegetation to return, improve habitat and possibly prey, and offer refuge for dispersing animals. Further surveys are needed to test these ideas.
• Planning a monitoring program: six replicates per site are adequate for occupancy, but abundance needs 12; adding replicates is cheap once observers are at a site. Waiting 5 years to detect a 50% decline in a low-density species is probably too risky, so sampling more sites would be a safer strategy. Smaller grid cells could suit species with smaller home ranges, and large volunteer programs, with help from landowners, could make surveys of more than 200 cells possible.

4. How tracking was or was not used

• The survey was a sign survey from a slowly moving vehicle. Observers looked for brown hyaena sign on and beside farm roads, and identified it to species by colour, dimensions, position and accompanying signs, using a field guide to tracks and signs.
• The paper does not report which kinds of sign were found. The substrate notes refer to scat and tracks.
• Only whether sign was present in each segment was recorded. Individuals were not identified, and there was no trailing.

5. Tracker or observer certification

• Reported but not validated. The paper says “two experienced observers conducted all surveys”, but their experience was not validated: no certification or test of their skill is reported.
• The occupancy models correct for missed sign (false absences), but they assume the sign that was found was identified correctly. So the results depend on the observers’ identifications.

6. Other technologies, skills or innovations

• The method gets repeated samples from consecutive road segments in a single visit. That makes landscape-scale occupancy surveys more practical than repeat visits or extra observers would.
• Interview data on landowner attitudes, rated on a scale, were built into the occupancy models as factors.
• Simulation software (GENPRES) and program MARK were used to plan sampling effort and test statistical power.

7. Implications for conservation (from the paper)

• Status: brown hyaenas are classified as near threatened, with fewer than 10,000 mature individuals worldwide. A 10% decline over three generations would see them reclassified as vulnerable. They are often persecuted for suspected livestock predation, although they are mainly scavengers, and their restricted range makes climate change a potential threat too.
• What could help: education and outreach that promote positive attitudes are likely to increase the area brown hyaenas occupy, and may reduce declines caused by persecution. Encouraging wildlife farming is likely to improve connectivity between subpopulations.
• A risk: if carnivore numbers grow, livestock losses and persecution may rise too, so support for non-lethal ways to protect livestock may become increasingly relevant.
• Volunteers: large-scale monitoring with volunteers and landowners could also raise public awareness of carnivore ecology and possibly improve attitudes.