# Camera trap, line transect census and track surveys: a comparative evaluation

> In the grasslands of Brazil’s Emas National Park, researchers ran camera traps, roadside sighting counts and a daily track census side by side. Tracks found the most mammal species, and found them fastest, with the full list in just 12 days. Yet the authors still favoured camera traps for the long run, because track surveys depend on the ground and on experienced trackers.

Summary by Kersey Lawrence, Original Wisdom. Updated 9 November 2026. Page: https://www.originalwisdom.com/tracking-literature/2003-silveira-jacomo-camera-trap-line-transect/

## Citation and access

- **Authors:** Leandro Silveira, Anah T. A. Jácomo and José Alexandre F. Diniz-Filho
- **Year published:** 2003
- **Journal or source:** Biological Conservation
- **Full citation:** Silveira, L., Jácomo, A. T. A. and Diniz-Filho, J. A. F. 2003. Camera trap, line transect census and track surveys: a comparative evaluation. Biological Conservation 114(3): 351–355.
- **DOI:** 10.1016/S0006-3207(03)00063-6
- **Access status:** Publisher link only
- **Where to find it:** https://doi.org/10.1016/S0006-3207(03)00063-6

## Tags

- **Species:** Mammals > Giant anteater; Mammals > Giant armadillo; Mammals > Hoary fox; Mammals > Jaguar; Mammals > Lowland tapir; Mammals > Maned wolf; Mammals > Mountain lion; Mammals > Ocelot; Mammals > Pampas cat; Mammals > Pampas deer; Mammals > Six-banded armadillo; Mammals > Southern tamandua; Mammals > Striped hog-nosed skunk; Mammals > White-lipped peccary
- **Region:** South America > Brazil
- **Tracking:** Sign surveys; Track counts; Tracking; Tracks; Tracks and signs identification
- **Methods and technology:** Camera traps; Detectability; Direct observation; Linear regression; Moran’s I; Pearson correlation; Survey costs
- **Data type:** New field data
- **Observer qualification:** Not reported
- **Discipline:** Wildlife tracking
- **Year published:** 2003

## Detailed keywords

- **Species:** Studied: the medium to large terrestrial mammals of the park’s grasslands (28 species are known there; the paper does not list them all). The 14 species detected by both camera traps and tracks, as named in the paper (common names added here): jaguar (Panthera onca), puma (Puma concolor), pampas cat (Oncifelis colocolo), ocelot (Leopardus pardalis), maned wolf (Chrysocyon brachyurus), hoary fox (Dusicyon vetulus), striped hog-nosed skunk (Conepatus semistriatus), lowland tapir (Tapirus terrestris), pampas deer (written Ozotocerus bezoarticus), white-lipped peccary (Tayassu pecari), giant armadillo (Priodontes maximus), six-banded armadillo (Euphractus sexcinctus), giant anteater (written Myrmecophaga tetradactyla) and southern tamandua (written Tamandua tridactyla). Inconsistency: the paper appears to swap the anteater species names (the giant anteater is usually Myrmecophaga tridactyla and the southern tamandua Tamandua tetradactyla); the body masses given (30 kg and 5.2 kg) match the giant anteater and the tamandua.
- **Location and habitat:** Emas National Park (18°19′S, 52°45′W), central-western Brazilian plateau, Goiás, Brazil. 132,000 ha, 98% grassland on flat land; Cerrado. Sampling along a segment of dirt road in the north-central part of the park; the park has 440 km of dirt roads.
- **Topics:** rapid faunal assessment; comparing survey methods; species richness; relative abundance; species accumulation over time; camera trapping; road track census with a palm-leaf brush drag; line transect (direct sighting) census from a vehicle; spatial autocorrelation; agreement between methods and body size; double counting of the same animal; survey costs; open versus closed habitats.
- **Statistics or models:** Species accumulation curves; photographic and track rates (records per day); Pearson’s correlation between methods across 29 sampling units; spatial correlograms with Moran’s I; Dutilleul’s (1993) correction of degrees of freedom for spatial autocorrelation; linear regression of photographic rate on track rate across species, and of between-method correlation on log body mass. Inconsistencies: (1) the text says camera traps detected 17 of 28 species (64%), but 17 of 28 is 61%; (2) Moran’s I was said to be estimated at five distance classes, but only four upper limits are listed (5, 10, 15 and 20 km); (3) the text gives the slope of photographic on track rate as b = 0.22, but the Figure 2 caption gives the equation as Y = 0.00478 + 0.798X (0.798 is the correlation r); (4) the text gives r = 0.792 for between-method correlation against body mass, but Figure 3 shows r = 0.74, and its caption equation has slope 0.792; (5) the effort figures (24,840 hours of camera trapping, 30,600 hours of track census, 28,050 hours of line transect census) are not explained and do not obviously match 29 cameras over 44 days or driving 1.5 km units at 20 km/h.
- **Data type:** New field data: camera trapping, daily track census and vehicle sighting census during January and March 2002 (the text reads “During January and March 2002”); track census run for 44 days.

## Synopsis

### 1. What they did

Conservation planning often needs a quick picture of which mammals live in an area and how common they are. The authors compared three ways of doing this in the grasslands of Emas National Park, central Brazil:

- Camera traps. During January and March 2002 they placed 29 automatic cameras (Cam Trakker) beside a dirt road, 1.5 km apart. Each camera was treated as one sampling unit covering a radius of 750 m.

- Track census. Each camera unit was also a 1.5 km track sampling unit. Every day for 44 days, a vehicle drove the road at 20 km/h, dragging a brush of palm leaves behind it to wipe out old tracks, so that only fresh tracks were counted on each pass.

- Line transect (sighting) census. Two observers rode on the back of the pick-up during the track census and recorded every mammal they saw. Further sighting surveys, by day and night, were run on other park roads to match the effort of the other two methods, so this method could not be compared with the others site by site.

They looked at how quickly each method built up the species list, and calculated photographic and track rates (records per day) for each species. They then compared camera and track results across the 29 units, allowing for the fact that nearby units tend to give similar results (spatial autocorrelation), and asked whether the two methods agreed better for larger species.

### 2. What they found

Species richness

- Of the 28 medium to large terrestrial mammal species known from the park’s grasslands, track surveys detected 19 (68%), camera traps 17 (reported as 64%) and direct sightings 16 (57%).

- Track census found species fastest, reaching its full count after only 12 days. All three methods arrived at similar species totals after about 30 days.

- 14 species were recorded by both camera traps and tracks; these are mainly species of open habitats.

Comparing camera traps and tracks across the 29 units

- Species richness and total abundance were not significantly correlated between the two methods (richness r = 0.312, P = 0.167; abundance r = 0.422, P = 0.078).

- For some species the two methods did agree significantly from place to place: white-lipped peccary (r = 0.703), giant armadillo (r = 0.230) and giant anteater (r = 0.403). For others, such as maned wolf (r = 0.006), they did not.

- Track rates were usually higher than photographic rates. For example, the maned wolf had a track rate of 0.154 records per day against 0.052 for photographs, and the tapir 0.121 against 0.024. The jaguar was similar by both methods (0.022 and 0.024).

- Across species, photographic and track rates were strongly correlated (r = 0.798), but photographic rates were consistently lower.

- The two methods agreed better for larger animals: agreement rose with body mass (r = 0.792 in the text, 0.74 in the figure).

### 3. What they concluded

- Track counts were the most efficient quick method for detecting species and relative abundance in Emas Park, but they need at least two people (a driver and an observer) and a vehicle, and the field staff need considerable experience with the signs of the local animals.

- Weather and ground conditions limit track surveys: ground that is too wet or too dry affects whether tracks can be seen and identified, and can make a survey valid or invalid. Game trails are good places for tracks but may not represent the wider area.

- Both track and photographic rates can overestimate abundance because the same animal can be counted more than once. Tracks are more prone to this, since photographs taken at very short intervals (probably of the same animal) can be discarded. The authors suggest this may partly explain the higher track rates.

- Direct counts along transects were the most limited method: they depend on good field conditions and well trained observers, and can favour large species active by day.

- Despite high start-up costs, the authors judge camera trapping the most appropriate method for mammal inventories in all conditions. It is cheaper in the long run, identifies species accurately, disturbs the environment little, detects night and day species equally (at least compared with direct counts), can be run by untrained staff, covers large areas at once, and can also show activity patterns and support population studies.

### 4. How tracking was or was not used

Tracks were counted daily on a dirt road divided into 29 units of 1.5 km, from a vehicle driving at 20 km/h. A palm-leaf brush dragged behind the vehicle wiped the road so that each day’s count held only fresh tracks. Tracks were identified to species and turned into track rates (records per day). Tracking was one of the three methods being compared.

### 5. Tracker or observer certification

Not reported. The paper does not say who identified the tracks or how experienced they were, although the authors state that track surveys need field staff with considerable experience of local signs. Several people and “all the volunteers” are thanked for help with data collection.

### 6. Other technologies, skills or innovations

- Cam Trakker automatic cameras (CamTrak South, Watkinsville, Georgia, USA).

- A brush of palm leaves dragged behind the vehicle to wipe old tracks off the road before each count.

- Statistical correction for spatial autocorrelation (Dutilleul’s method) when comparing methods across nearby sites.

### 7. Implications for conservation (from the paper)

- Fast, reliable ways to measure species richness and abundance are crucial for setting conservation priorities, especially given urgent conservation needs worldwide.

- Camera trapping can estimate both richness and abundance in a short time, and long-term monitoring could track changes in abundance, activity, breeding and habitat use. These data can serve as a rapid assessment of wildlife conservation status and guide conservation priorities and management.

### 8. Implications for tracking (Original Wisdom's observations)

_These are Original Wisdom's observations, not the authors'._

- Track counts found more species, and found them faster, than either camera traps or sightings: 19 species, with the full list in 12 days.

- The authors’ main reasons for preferring camera traps (dependence on ground conditions and on experienced staff) point to the value of trained and evaluated trackers: the skill of the observer, which the paper does not report, is one of the things that decides how reliable a track survey is.

## Cite this summary

Lawrence, K. 2026. Summary of: Camera trap, line transect census and track surveys: a comparative evaluation (Silveira and Jácomo et al., 2003). Original Wisdom, Tracking Literature. https://www.originalwisdom.com/tracking-literature/2003-silveira-jacomo-camera-trap-line-transect/