Effectiveness and Accuracy of Track Tubes for Detecting Small-Mammal Species Occupancy in Southeastern Herbaceous Wetlands and Meadows

1. What they did

Live traps are the usual way to survey small mammals, but they can injure or kill animals, and in the southeastern United States invasive red imported fire ants can attack animals held in traps. Track tubes are a non-invasive alternative: an animal walks through a tube, over soot, and leaves its footprints on sticky paper. But small mammals’ tracks can look very similar, so identifications can be uncertain. The authors asked how well track tubes compare with live traps for working out which sites a species occupies, and how to handle uncertain track identifications.

• Study sites. 37 grassy and wet meadow sites in eastern Alabama and at Arnold Air Force Base in Tennessee, sampled in 2015 and 2016 as part of a survey for the meadow jumping mouse.
• Traps and tubes. Each site had 24 Sherman live traps and 24 track tubes, alternating along a grid with 10 m between stations, all baited with black oil sunflower seeds. Each tube was two 30 cm sections of vinyl rain gutter, with one end capped against rain, a strip of metal flashing sooted with a kerosene lantern, and a sheet of sticky shelf-liner paper to catch the tracks. Sites were sampled for 4 nights and checked every morning.
• Identifying tracks. Tracks were identified to species, or to the lowest level possible, and given high, medium or low confidence, using a reference library of known tracks from captured animals and tracking guides. One main observer with 1 year of experience identified all tracks, trained by a certified tracker who acted as second observer. Both observers identified a subset of 58 track plates independently to check agreement.
• Analysis. They compared detection by the two methods for deermice, marsh rice rats and hispid cotton rats, handling uncertain tracks in different ways: keeping all of them, dropping low-confidence ones, dropping low- and medium-confidence ones, or modelling the chance that uncertain detections were false (false-positive occupancy models).

2. What they found

Effort and species

• Total effort was 7008 trap nights, half with each method.
• They detected six rodent species and one shrew. No meadow jumping mice were detected.
• On average, live traps and track tubes detected about the same number of species per site.
• With all confidence levels counted, track tubes recorded more identifications than live traps caught (265 versus 189). With only high-confidence identifications, track tubes recorded fewer marsh rice rats and deermice than live traps.

Agreement between the two observers

• 35 of the 58 plates checked had small-mammal tracks.
• The observers agreed 100% to order and family, on 34 of 35 plates to subfamily, and 94% to genus.
• To species, they agreed on 26 of 35 plates (12 with the same confidence, 14 with different confidence). In 7 of the 9 disagreements, one observer gave a species and the other said the plate did not have enough information.

Detection

• Live traps detected cotton rats best (median 0.55 per night), compared with rice rats (0.19) and deermice (0.27).
• When all tracks were kept, or uncertainty was modelled, track tubes and live traps detected the animals about equally well.
• When low- or medium-confidence tracks were dropped, track tubes’ odds of detection were on average 55% lower than live traps’. This was strongest for rice rats, which were detected only by live traps at 6 of the 19 sites where they were found.
• Deermice were found at 22 sites: only by track tubes at 10 sites, and only by live traps at 3.
• Cotton rats were found at 20 sites: by both methods at 12, only by live traps at 4, and only by track tubes at 4.

False positives

• The chance of a false-positive detection was about 0.11 for deermice and 0.05 for cotton rats per survey. The data could not tell this for rice rats.
• The chance that a true track detection was identified with high confidence was highest for cotton rats (0.82), then deermice (0.73), and lowest for rice rats (0.52), even though rice rat and cotton rat tracks are generally similar.

Practical points

• Track tubes cost about US$5 each in supplies and building time, against about US$20 for a Sherman trap.
• In dry weather, sooted tubes worked for up to 6 nights before the paper needed replacing, but in heavy rain the plates had to be replaced daily.

3. What they concluded

• Track tubes are a useful non-invasive way to study which sites small mammals occupy in the Southeast, but only if tracks can be identified to the level the study needs and the uncertainty is handled properly.
• Ignoring uncertain identifications would bias occupancy estimates, because false positives did occur. Simply dropping uncertain tracks lowers detection and would mean more survey nights. False-positive models kept the uncertain tracks and lost no effectiveness compared with live traps.
• Cotton rats were easier to identify with confidence than rice rats, perhaps because they were more abundant and have prominent paired pads on the hind foot.
• Identification should be done with the training and oversight of an experienced tracker, with regular checks of observer accuracy. Studies should report the experience of key staff, ideally as tracker certification levels, how much identification uncertainty there was, and how it was handled.
• Where tubes are always wet, they may not be useful.

4. How tracking was or was not used

The study is about tracks and signs identification of small-mammal footprints on track plates. Each track was identified from diagnostic features of the feet (based on a reference library of known tracks and tracking guides) and given a confidence level. The agreement between two observers was tested on a subset of plates, and uncertainty was built into the analysis. No trailing was involved.

5. Tracker or observer certification

One primary observer (D. Duffie, 1 year of track identification experience) identified all tracks. He was trained by N. Sharp, described as an experienced tracker with 6.5 years of experience and certified Level III in the CyberTracker evaluation system, who served as the second observer. Their agreement was tested on 35 plates with small-mammal tracks.

6. Other technologies, skills or innovations

• A modified track tube design: two 30 cm sections of vinyl rain gutter taped together, with one end capped with metal flashing and duct tape against rain splatter and feeding debris, soot from a kerosene lantern, and sticky shelf-liner paper. Plates with tracks were kept in clear plastic sleeves for later analysis.
• A three-level confidence score (high, medium, low) for each identification.
• Sevin Dust under traps and tubes to reduce fire ant problems.
• False-positive occupancy models that treat live-trap captures and high-confidence tracks as certain and other tracks as uncertain.

7. Implications for conservation (from the paper)

• Track tubes carry no meaningful risk to animals, a major advantage for protected species and where fire ants threaten trapped animals.
• They are cheaper than live traps and allow more flexible checking.
• Their value for monitoring depends on how reliably tracks are identified and how identification uncertainty is handled.