Interpreting Carnivore Scent-Station Surveys

1. What they did

For decades, wildlife managers have counted the tracks left at scented baits (scent stations) to follow trends in carnivore numbers. Carnivores are secretive and usually live at low densities, so counting them directly is hard and costly. Scent stations seemed a cheap answer, but attempts to check whether they really track abundance had given mixed results.

The authors used data from one of the most intensive long-term scent-station surveys: Minnesota, 1986 to 1993. Each station was a circle of smoothed earth 0.9 m across with a fatty-acid scent tablet in the middle. Ten stations, 480 m apart along unpaved roads, made one line, and 441 lines were spread across the state. Each line ran for one night a year, and the next day the presence or absence of each species’ tracks was recorded at every station.

They asked:

• How do scent-station data behave statistically? Are visits to nearby stations independent, or does one animal visit several stations in a row?
• How should the data be analysed? They compared a station index (share of stations visited) with a line index (share of lines with at least one visit), and tested for trends over the years in each part of the state.
• Do scent stations reflect abundance? They reanalysed two of the most rigorous earlier validation studies: raccoons on Davies Island, Tennessee, and bobcats introduced to Cumberland Island, Georgia.

2. What they found

Visits to nearby stations are linked

• For every species except raccoons, lines had more stations visited, and fewer lines were visited, than expected if each visit were independent.
• Results from stations closer than about 2 to 2.5 km were similar. The authors put this down mainly to single animals visiting several stations in succession. The effect was strongest for wide-ranging, territorial species (wolves, coyotes, red foxes, bobcats) and weakest for skunks and raccoons.
• Stations at opposite ends of a line also gave surprisingly similar results, which the authors could not explain.
• Counting each line as simply visited or not visited removed the problem: the number of lines visited fitted a binomial distribution for all species.

Trends in Minnesota, 1986 to 1993

• Statewide station indices rose for wolves, coyotes, red foxes, raccoons and bobcats, and fell for skunks. Line indices showed the same directions.
• Statewide figures hid regional differences. Coyote station indices fell in the West Superior and Peatland sections but rose elsewhere. Raccoon indices fell only in the Peatland section. Skunk line indices rose in three sections and fell in the rest.

The validation studies

• Raccoons, Davies Island: even if raccoon visits truly rose in line with numbers, the original tests would have detected it in only 18.3% (rank correlation) and 12.6% (logistic analysis) of 10,000 simulations. The original study’s failure to find a link was most simply explained by low statistical power.
• Bobcats, Cumberland Island: abundance explained 73% of the variation in visit rates. But within each of the two time periods, visits did not rise with numbers, and a model that only separated the two periods explained 79%. Other factors that differed between periods could explain the result as well as abundance could.

3. What they concluded

• Visits to stations closer than about 2 km are not independent, because one animal can visit several. Treating stations as independent makes tests find differences too often and lets a few busy lines dominate.
• Lines are the right unit. Scoring each line as visited or not gives well-behaved data and gives individual animals fairer weight.
• Testing for trends over several years is more reliable than comparing two years, because many factors other than abundance change visit rates from year to year.
• Regions should be analysed separately, because trends differed between them.
• The validity of scent-station indices has been neither proven nor seriously disproven. Earlier validation studies were limited by low statistical power and confounding.
• Long-term trends in visit rates probably reflect real population changes. But the method has poor spatial and time resolution, is easily confounded, and has low statistical power. It should supplement other information, not replace it.

4. How tracking was or was not used

Scent stations are a tracks and signs method: each station is a patch of smoothed earth where visiting animals leave footprints, and observers recorded which species’ tracks were present the day after the station was set. All analyses rest on these track identifications. No trailing was reported. The authors list observer bias and observer error among possible sources of error, but did not test identification accuracy.

5. Tracker or observer certification

Not reported. The paper does not say who checked the stations or how their track identification skills were assessed. The authors name observer bias and observer error as possible influences on the results, and note that their simulations did not include observer error.

6. Other technologies, skills or innovations

• Indicator variograms, a geostatistical tool, used to show how similarity between stations changes with distance.
• Computer simulation of 10,000 datasets to estimate the statistical power of an earlier study’s tests.
• A simple trend test (regression of ranked index values on year) that the authors found easy to apply and interpret.

7. Implications for conservation (from the paper)

• Scent stations are best used for broad, long-term trends at a regional scale, not for rare species or for local monitoring of wide-ranging carnivores.
• Reliable results may need hundreds of lines of many stations, so the method may be less cost-effective than often believed.
• Scent-station indices cannot yet be converted to numbers of animals, so it is hard to know from them alone when management action is needed.