On tracks: A spoor-based occupancy survey of lion Panthera leo distribution in Kafue National Park, Zambia

1. What they did

Africa’s lions have lost most of their historic range, and Zambia is one of only nine countries thought to hold more than 1000 wild lions. Kafue National Park is a key stronghold, but there were no lion data for its northern half. Lions there are killed in wire snares set by illegal bushmeat hunters, and Zambia banned lion trophy hunting in January 2013. The authors wanted to know where lions are in northern Kafue and what shapes their distribution.

• Survey grid. Using home ranges from nine GPS-collared lions, they chose a grid cell size of 200 km² and laid 73 cells (14,600 km²) over northern Kafue. 41 cells had unpaved roads they could survey.
• Spoor surveys (1 July to 15 October 2012). They drove 17 road transects of 17 to 30 km, two to four times each, at least a week apart. Surveys started within 30 minutes of sunrise, when tracks are most visible, at about 10 km per hour. One observer sat on the front left corner of the vehicle and one drove; both scanned the road for lion tracks. Each transect was split into 1 km segments, recorded as lion detected or not.
• Occupancy analysis. An occupancy model uses repeat surveys to estimate how often lions were missed where they were present, and corrects the simple “tracks found” picture for these false absences. They used a version designed for signs found along consecutive stretches of road or trail.
• Testing what drives distribution. They predicted lions would use areas with more prey and areas farther from people. They tested prey biomass, habitat type, fire, water, law enforcement effort, and distance to the park boundary, human activity, tourist camps, safari hunting camps and wildlife authority posts. They also tested whether road surface quality and vehicle traffic affected detection.

2. What they found

The survey

• 46 surveys covered 1010 one-km segments. Lion tracks were found on 147 segments (14.6%) and in 24 of the 41 surveyed cells.
• Road surface was good or very good on about a third of segments, moderate on half, and poor or very poor on 15.6%. Most segments (94.3%) showed no impact from other vehicles.
• Neither road surface quality nor vehicle impact had a clear effect on detecting lion tracks.

How much of the area lions use

• Counting only cells with tracks, lions used about 58.5% of the grid (8541 km²).
• After correcting for missed detections, the estimate was 72.1% (about 10,527 km²), 23.3% higher.

What drives where lions are

• Munga and termitaria woodland was the strongest predictor: the more of this habitat in a cell, the more likely lions used it.
• The top-ranked model added proximity to water, but water’s effect was not clear, so the authors used the simpler habitat model for their map.
• Prey biomass had little support, ranking 13th of 14 candidate models in the first round.
• Nothing linked to people (distance to the park boundary, human settlements, safari hunting camps, wildlife authority posts, tourist camps, or patrol effort) fitted the data well.
• Predicted probability of lion use per cell ranged from 0.760 to 0.952. 19 cells were below 0.800, 29 were between 0.800 and 0.899, and 25 were above 0.900.

3. What they concluded

• This was the first lion survey using track-based occupancy, and the first use of the Hines et al. (2010) spatial correlation model for an African cat. It worked well over a very large area with limited access and resources.
• Habitat was the main driver of lion use in northern Kafue. The authors suggest the patchy munga and termitaria woodland suits lions’ stalk-and-pounce hunting better than dense miombo woodland or open scrub and grassland.
• They were surprised that human disturbance did not matter. They believe illegal bushmeat hunting is so widespread in Kafue that it no longer shows as an edge effect, so even the interiors of Africa’s largest protected areas are not safe. This does not rule out human effects on lion numbers.
• Results reflect dry season use only. The prey layer came from a single 2011 aerial survey, and prey might have ranked higher if measured during the survey.
• Track surveys need roads with suitable substrate and staff with tracking skills. They may work less well outside protected areas, where carnivores are warier of people and avoid roads; there, audio lures or camera traps may suit better.

4. How tracking was or was not used

Lion presence was detected entirely from tracks on unpaved roads. Two observers scanned the road from a slow-moving vehicle at sunrise and identified lion tracks by size and shape. Tracks they were unsure of were left out, because the model assumes there are no false detections. No trailing was reported. The authors chose roads because off-road animal trails were mostly on hard clay where tracks did not show.

5. Tracker or observer certification

Reported but not validated. The two observers, one of them the lead author, are described as “competent at detecting and identifying tracks”; no certification or test of their skill is reported. Five people are thanked for help in the field.

6. Other technologies, skills or innovations

• GPS collars (Vectronic Aerospace) on nine lions to set the survey grid cell size.
• Google Earth and ArcGIS to choose transects and map covariates and results.
• The occupancy software PRESENCE, using the model of Hines et al. (2010) that allows for detections along consecutive road segments being linked.
• Road surface quality (scale of 1 to 5) and vehicle impact (scale of 0 to 2) recorded for each segment.

7. Implications for conservation (from the paper)

• The results give the Zambia Wildlife Authority a survey method and focus areas for further lion research in Kafue, and can inform the debate on whether to reopen lion trophy hunting.
• Illegal hunting appears to reach deep into the park, and the authors highlight an urgent need for more effective law enforcement.
• Track-based occupancy surveys can show where large carnivores are in large, previously unsurveyed African protected areas. Multi-season surveys could act as a proxy for population trends where numbers are unknown.