Substrate and species constraints on the use of track incidences to estimate African large carnivore abundance

1. What they did

Managers of lions and other large carnivores need to know how many animals there are and whether numbers are going up or down, but counting them directly is slow, costly and hard to repeat. Track counts are a cheaper alternative: the more animals there are, the more tracks people should find. Earlier work (Stander 1998) warned that such indices might only hold for the species and site where they were calibrated, so the authors tested whether one relationship could work across many places, species and soils.

• Track surveys. At 18 study sites in seven study areas in Namibia, South Africa and Botswana, Zimbabwe, Kenya and Tanzania (including published Namibian data from Stander 1998), teams drove sections of road or off-road routes (transects) on soils fine enough to show tracks. They left out routes with more than 50% vegetation cover. A tracker sat on the bonnet of a four-wheel-drive vehicle moving at about 10 to 20 km per hour and scanned for tracks ahead. When tracks were found, the vehicle stopped and the trackers identified the species, leaving out any they could not identify reliably. Only tracks from the previous 24 hours were counted; roads were swept with a brush drag when transects were driven on consecutive days. In total they drove 15 306.4 km on 339 transects. The Serengeti sites were surveyed in both wet and dry seasons.
• Known densities. They compared track counts with carnivore densities from long-term monitoring of known individuals (radio collars, coat or spot patterns, whisker spot patterns, brands) or from published studies.
• Models. They fitted relationships between track density (tracks per 100 km) and carnivore density (animals per 100 km²) by species and by soil type, then worked out how far a survey team needs to drive to get a precise estimate, how to put confidence intervals on the estimate, and how often surveys need to be repeated to detect a 10% change per year.

2. What they found

Track density and true density

• One relationship for all cats regardless of soil did not work well: it explained only 21% of the variation.
• Sandy soils: lion density explained 97% of the variation in lion track density. Cheetah, leopard and hyaena tracks fitted the lion model just as well, so one combined model for all species on sandy soils explained 96% of the variation (about 3.15 tracks per 100 km for each animal per 100 km², plus 0.40).
• Clay soils: lions and cheetahs did not differ, and a combined cat model explained 71% of the variation. Spotted hyaenas on clay, at very high densities in the Serengeti, were the only case where the index levelled off (saturated) as density rose.
• The sandy and clay models differed, so soil type matters.

How much survey effort is needed

• Track density fell steeply as the distance between track finds grew.
• On average, finding 19 sets of tracks gave a coefficient of variation (a measure of precision) under 20% for the distance between finds, but 30 sets of tracks most often ensured it. This held across densities and soils.
• The distance needed to find 30 sets of tracks on sandy soils dropped sharply as lion density increased. For the one clay-soil case, the sandy-soil model predicted 22.5 km, but 274.1 km had been surveyed in the Serengeti long-grass wet season.

Lion estimates and trends

• Lion densities estimated from tracks matched the other estimates inconsistently. The authors report that the confidence interval included the other estimate in three of nine sites (Table 3 shows two), but low track-based estimates went with sites that also had low densities by other methods.
• Most sites were precise enough that a 10% change could be detected with two surveys one year apart. Less precise sites needed up to three surveys up to three years apart, with up to six years of not knowing the trend, by which time numbers could have changed by up to 77%.

3. What they concluded

• Indices based on track counts can give useful estimates of large carnivore abundance, once soil type is taken into account.
• On sandy soils a single model works for lions, leopards, cheetahs and hyaenas, despite their different behaviour and ecology. Stander’s advice to treat indices as site-specific is a sensible starting point, but these results show helpful generalization is possible.
• Saturation was a problem only for spotted hyaenas at very high density.
• A particular strength of the method is that it gives confidence intervals; only a third of the lion estimates from other methods did.
• Two conditions are needed for success: suitable substrate with sufficiently skilled trackers, and, especially where the substrate is less suitable, calibration of track density against known local carnivore density.
• They offer a flow diagram (Figure 3) for designing a track survey, estimating density with confidence intervals, and planning repeat surveys.

4. How tracking was or was not used

Tracking is the core method. Trackers seated on the vehicle bonnet spotted fresh carnivore tracks (footprints) on roads and routes, identified the species, and left out any tracks they could not identify reliably. Each set of tracks was counted once per day; if trackers found two similar sets within 500 m and could not tell them apart as individuals, the second was not counted. No trailing was reported. Radio-tracking (telemetry flights or ground-tracking) was used separately to estimate the known densities.

5. Tracker or observer certification

Reported but not validated. Where available, expert indigenous local trackers were used, and every survey team included “a competent track interpreter”. Different trackers worked at different study areas, each site directed by its lead researcher. The authors were confident this avoided observer bias (citing Stander et al. 1997), but the trackers’ skill was not tested or certified in the study, and they are not named.

6. Other technologies, skills or innovations

• Handheld geographical positioning systems (GPS) and digital road maps to log routes and measure transect lengths, and to map vegetation cover along them.
• Brush drags towed behind the vehicle to sweep roads clean of old tracks.
• Simulation (10 000 random draws) to put confidence intervals on track-based density estimates.
• A flow diagram guiding survey design, density estimation and trend monitoring.

7. Implications for conservation (from the paper)

• Track surveys are cost effective and repeatable, which matters because changes in numbers are often more important to managers than absolute numbers.
• The method allows confidence intervals to be calculated, so managers can weigh trade-offs between survey effort, survey interval and the risk of missing a decline.
• Where soils are less suitable, local calibration against known densities remains a sensible precaution.