Monitoring wildlife abundance through track surveys: A capture-mark-recapture inspired approach to assess track detection by certified trackers in the Kalahari, Botswana

1. What they did

The Formozov–Malyshev–Pereleshin (FMP) formula, which turns track counts into density, assumes that trackers detect every track on the transect. That may hold on foot, but large surveys are usually driven, and tracks can be missed. The authors measured how likely trackers are to detect a track that is there, what affects it, and how often two tracker pairs disagree on species and track age.

• Design, inspired by capture-mark-recapture: two teams drove each transect one after the other. Each team had a vehicle driven by a recorder and a pair of certified trackers on seats on the bull bar. The first team "captured and marked" the tracks; the second "recaptured" them. The teams kept a gap between vehicles so the second team could not easily see where the first had stopped.
• The tracker pairs: the NG pair was led by a Master Tracker, with a Tracker III. The XZ pair was led by a Tracker III, with a tracker certified in Track & Sign I and Trailing III. The two pairs alternated between the first and second vehicle.
• Survey: seven days in May 2022 on 40 transect segments (175 km). Driving was capped at 10 km per hour and slowed where visibility was poor. Surveys avoided the midday hours, so shadows made tracks stand out. Trackers stopped the vehicle, and stepped down when needed, to confirm identification and age.
• Recording: for each track, location, species, number of individuals and age (from 24 hours to over a month) were recorded in CyberTracker on a smartphone. A trail crossing the transect counted as one track interception, and every crossing was counted, whoever made it. Identifications were not checked by an independent assessor; the aim was to measure disagreements in real conditions, where the truth is unknown.
• Factors tested: tracker pair; substrate (sand cutline, sand road, gravel road, off-road); livestock track density; track age (24 hours or less versus older); and whether a species tends to cross roads (the five antelopes) or follow them (the six carnivores and ostrich).
• Matching: tracks detected by both teams were matched by location, species and age, then analysed with logistic regression.

2. What they found

Detection is high but not perfect

• The teams recorded 1612 tracks: 1251 by the first team, 1127 by the second, and 766 by both.
• The probability of detecting a track was 0.68 for the first team and 0.61 for the second, rising to 0.77 and 0.75 for fresh tracks (24 hours or less).
• They estimated there were about 1841 tracks in total, so 229 were missed by both teams.

Track age matters most

• Fresh tracks were more detectable (0.77) than older tracks (0.61).
• Older tracks of followers (carnivores and ostrich) were the hardest to detect (0.55, against 0.67 for crossers). The authors suggest antelope cross in groups, leaving more obvious track sets than solitary carnivores.

Other conditions mattered little

• Detection was slightly higher on clear sand (0.73) than on gravel roads and off-road transects (0.65), a marginal difference.
• The pair led by the Master Tracker detected slightly more tracks, but only in the hardest conditions: fresh and antelope tracks off-road, and older tracks on gravel roads. The difference was very small (5%).
• Detection by both teams was marginally higher when the NG pair went first (0.52 versus 0.44), which may mean slight information transfer from the first vehicle.

Agreement on species

• The two pairs agreed on the species for 81% of the 766 tracks both detected, and 92% for fresh tracks.
• Agreement was 95% for carnivores and ostrich but 77% for antelope, mainly because gemsbok and red hartebeest, which leave very similar tracks, were confused in 29% of cases (119 of 412).
• The only carnivore mismatches were 4 leopard–cheetah, 1 lion–leopard and 2 spotted hyena–brown hyena.

Agreement on age

• The pairs agreed on whether a track was fresh or older for 88% of tracks. Disagreement was higher on gravel roads (0.23), whose fine sand and dust erode quickly.

No overall bias

• Disagreements took the form of two-way confusion rather than one pair consistently favouring one species or age class, so errors largely cancelled out in the total counts.
• One pair recorded proportionally more spotted hyena and blue wildebeest tracks, but the counts were too small to conclude a bias.

3. What they concluded

• The main finding: track detection was high compared with many other wildlife survey methods, and remarkably constant across tracker pairs and sampling conditions. Fresh tracks were slightly more detectable and more reliably identified than older ones.
• A first: to their knowledge, this is the first published study to use a capture-mark-recapture approach to measure track detection in real field conditions.
• For the FMP formula: detection of fresh tracks was below 1 (0.77), so absolute densities from driven surveys will be underestimated to some extent. Because detection was largely constant, the formula can at least give reliable relative densities. Practitioners must decide whether measuring detection to correct absolute estimates is worth the effort.
• Survey design: to keep detection high in driven surveys, survey where and when disturbance is mild, in daytime with good track visibility, at low speed set by the observers, and with observers whose skills have been assessed and maintained to an excellent level. Clearing transects by dragging 24 hours before may not be worth the cost, given that age confusions cancel out.
• Skill: track ageing is one of the hardest parts of tracking. The results emphasize training to reach and maintain excellent skills. Standardizing and quantifying tracking skills, for example through CyberTracker Conservation's Tracker Certification System, is necessary to ensure reliable data.
• Limitations: only two tracker pairs were compared, so formal proof that pairs do not differ would need more pairs. In walking surveys, detection is likely close to 1, but this was not tested.
• Communities: tracking knowledge is eroding. Involving communities in track-based work can maintain tracking skills, value cultural identity and Indigenous knowledge, and provide much-needed employment.

4. How tracking was or was not used

• Track counts: two pairs of certified trackers, driving at up to 10 km per hour, detected, identified and aged every track of 12 large species crossing the transect.
• Rules: a single trail crossing counted as one interception, an animal walking along the road counted once, and every separate crossing was counted, regardless of individual.
• Checking: trackers stopped the vehicle, and stepped down when needed, for a closer look.
• No trailing was done.

5. Tracker or observer certification

• Certified trackers. The NG pair was led by a Master Tracker, assisted by a Tracker III. The XZ pair was led by a Tracker III, assisted by a tracker with Track & Sign I and Trailing III certifications. The certifications follow the CyberTracker Tracker Certification System (Liebenberg et al. 2018).
• The four Indigenous tracker co-authors took part in the fieldwork. The paper states that they did not take part in the data analysis and did not influence its conclusions about their tracking skills.

6. Other technologies, skills or innovations

• Capture-mark-recapture applied to tracks: two teams driving the same transect one after the other, with matching by location, species and age instead of physically marking tracks.
• CyberTracker: a smartphone app used to record location, species, number of individuals and track age.
• Timing and speed: avoiding midday so shadows show the tracks, and capping speed at 10 km per hour.

7. Implications for conservation (from the paper)

• Value of track surveys: where the substrate is suitable, track surveys are cost-effective. They can cover many species at once, including elusive large carnivores, over large areas.
• Further uses: once both parts of detectability are measured or controlled, track counts can also help identify key habitats, distribution and landscape connectivity.
• Communities: involving local communities in track-based work can bring them benefits from wildlife conservation, for example through Community Based Natural Resource Management and Payments for Ecosystem Services.
• Recognition: certified skills could be formally recognized by governments and the private sector.