Power of Sign Surveys to Monitor Population Trends

1. What they did

Grizzly bears in the lower 48 states of the United States live at low density in remote country, wander widely and are rarely seen, which makes their numbers very hard to follow. At the time, trends were followed either by expensive radio-telemetry studies or by chance sightings of mothers with cubs. The authors asked how much survey effort would be needed to pick up a fall in numbers from the signs bears leave on trails.

• Field surveys (1984 to 1988). Each autumn, three groups of three to five observers walked trails in Glacier National Park, Montana, recording all bear signs. Observers who were new to the work had four hours of classroom and four to eight hours of field instruction in recognising bear sign. Feeding signs (digs, stripped trees) and marking signs (rubbing trees, worn bear trails) were rarely found, so the analysis used only scats and tracks.
• Recording signs. Grizzly and black bear scats could not be told apart, so all bear scats were counted together. Only front-foot tracks were recorded, because hind tracks are hard to assign to species. A track was counted as a new entry when its front pad differed by more than 1 cm in length or width from a track already recorded on that stretch, when tracks of similar size were going in different directions, or when it was at least 1.6 km from the last track of the same size.
• Setting up the data. Trails were divided into 1.6 km (1 mile) segments, and each segment was scored for whether it had sign or not.
• Modelling and simulation. They fitted a statistical model to how often sign turned up on trail segments, then used a computer to simulate thousands of “before and after” surveys in which the amount of sign dropped by 20%. This showed how often each survey design (5 to 50 trails, each 5 to 20 segments long) would detect the drop.

The authors state that they did not study the biology or population status of the park’s bears.

2. What they found

Sign on trails

• The surveys covered 131 to 150 segments a year on 17 to 20 trails.
• Scats were by far the most common sign: on average 0.43 scats per segment, and 27% of segments had at least one scat.
• Tracks were much rarer: 0.06 grizzly bear and 0.05 black bear tracks per segment on average.
• Sign on one segment did not predict sign on the next, and sign on a trail one year only weakly predicted sign the next year.
• Counts differed little between years, so the five years were pooled.
• In spring (May 1987) there was more sign than in autumn: 0.56 scats and 0.14 grizzly bear tracks per segment, against 0.43 and 0.06.

How much effort is needed

• With scats (the more common sign), the largest designs reached 50% to 90% power, meaning they would detect a 20% drop that often.
• With grizzly bear tracks (the rarer sign), power stayed at or below 60% even when 1000 segments (50 trails of 20 segments) were surveyed.
• Spreading effort over more trails worked better than making trails longer: 100 segments as 20 trails of 5 segments gave 1.24 times the power of 5 trails of 20 segments.
• Power rose slowly with effort: a 16-fold increase (from 50 to 800 segments) only doubled it.
• Averaging three years of data raised power by about 30% to 50% for scats. Even so, at least 200 segments were needed to reach 70% power when trails varied independently from year to year.
• More variation between trails lowered power.

3. What they concluded

• Sign surveys on trails are cheap and easy to repeat, but they will not detect small yearly changes. At best they reliably detect only large declines, and then only with large samples, fairly abundant sign, and some false alarms.
• A 20% decline is already a serious one, yet detecting even that needs large samples, so the authors doubt that much more sensitive monitoring over large areas could be done at realistic cost.
• Managers should decide the power they want, the significance level and the largest decline they can accept going unnoticed.
• Design suggestions: survey as many trails as possible, even if they are shorter; pool several years or repeat surveys within a year; survey when sign is densest (spring for grizzly bears); do not place routes only where animals are most common, because routes must represent the whole area; use the shortest segment length that keeps segments independent; run a pilot survey where sign density is unknown.
• A feasible grizzly bear survey would need about 500 to 1000 segments, or about 0.5 sign per segment, for example by repeating surveys from spring through autumn and pooling them.
• Scat-based monitoring of grizzly bears needs scats identified to species. The authors recommend DNA analysis as a research priority and mention thin-layer chromatography as another promising method.
• A decline in sign is taken as a warning about the population, but linking sign trends to population trends is a matter of biological judgement.

4. How tracking was or was not used

Tracks and signs identification was the basis of the field data. Observers recorded all bear signs along trails, including sightings, tracks, scats, digs, stripped trees, rubbing trees and bear trails. They told grizzly from black bear tracks using front-foot tracks only, and used front pad size (a difference of more than 1 cm) and direction of travel to decide whether tracks came from different bears. Scats were counted but not identified to species. No trailing was reported.

5. Tracker or observer certification

Not reported. The surveys were carried out by students of the Wildlands Research Institute (San Francisco State University) and their instructor. Observers new to the work received four hours of classroom and four to eight hours of field training in bear sign recognition, covering species recognition for sightings and tracks, feeding signs and marking signs. Their skill was not tested. The authors note that presence or absence data are robust to differences in how observers interpret sign, which was one reason they used them.

6. Other technologies, skills or innovations

• Using a beta-binomial model to describe how often sign turns up on trail segments, and computer simulation to work out the power of different survey designs before running them.
• Power isopleths (contour maps of power against number of trails and segments per trail) that managers can read to choose a design.

7. Implications for conservation (from the paper)

• Sign surveys may reveal long-term trends or coming disaster earlier and at lower cost than other tools, but only for large declines.
• The authors believe a large-scale grizzly bear trend-monitoring programme in the Rocky Mountains could be based on scat occurrence, once scats can be identified to species.
• With 80% power, a real decline would be missed one time in five; with a significance level of 0.2, a false decline would be signalled one time in five when nothing had changed.
• Declines may show up last in the best habitat, so concentrating surveys there would seriously weaken the design.