An evaluation of structured snow-track surveys to monitor Eurasian lynx Lynx lynx populations

1. What they did

In Norway lynx are heavily hunted, and the main way of monitoring them is a minimum count of family groups (adult females with dependent kittens) from tracks in the snow. The authors wanted to know whether an extra index, counts of 3 km transect lines crossed by lynx tracks in the snow, could reliably detect changes in population size.

• Simulating the survey. Using a Geographic Information System (GIS), they laid out virtual 3 km transects over a 10,173 km² study area in Hedmark, south-central Norway. One network of 264 lines was placed “deliberately”, using their experience of where lynx travel: mostly from the valley bottom up the valley sides, or near steep terrain lynx favour for day beds. The other network of 259 lines was placed at random, at about the same density (about one line per 38 km²).
• Using real lynx movements. They overlaid the daily locations of 31 radio-collared lynx, collected in winter between 1995 and 2001, and counted how often each lynx crossed a line over one, two, three or five nights. These periods stand for the number of nights since the last snowfall, during which tracks can build up.
• Testing the design. For each scenario, and with 100%, 75%, 50% and 25% of the lines, they estimated the chance of detecting a lynx that was present, and the statistical power to detect a population decline from one survey to the next, starting from 30 lynx. They also simulated how many years of annual surveys would be needed to detect a steady long-term decline.
• Field trials. They compared the simulations with two large one-day surveys organised by the local wildlife management agency in January 1999 and January 2001. Hundreds of volunteers, mainly local lynx hunters, skied 3 km lines placed by local experts, and the researchers checked their reports of lynx tracks.

2. What they found

Detecting lynx that are present

• Deliberately placed lines detected lynx significantly more often than random lines, for every track accumulation period.
• With all deliberate lines, detection rose from 35% after one night to 57% after two, 80% after three and 91% after five nights. Random lines gave 17%, 30%, 52% and 68%.
• Fewer lines meant lower detection: three nights after snowfall, a lynx would be detected in 72% of cases with 75% of the lines, and only 38% with 25% of the lines.
• Males and females with kittens appeared easier to detect than females without kittens, though this was hard to test properly.

Detecting a decline between two surveys

• Deliberate lines again performed better than random lines, and power fell as line density fell.
• With all lines surveyed three nights after snowfall, a drop from 30 to 19 lynx would be detected eight times out of ten; two nights after snowfall, a drop from 30 to 17.
• A drop from 30 to 20 lynx would be detected in 69% of cases with all lines, 54% with 75%, 43% with 50% and only 30% with 25% of the lines (three nights).

Detecting a long-term decline

• With annual surveys of deliberate lines two days after snowfall, the smallest decline detectable with 80% power went from 30 to 5 lynx over seven years, to 30 to 13 lynx over 20 years.

Field trials

• 1999: 692 volunteers covered 396 lines. Lynx tracks were on 13% of lines; 10 of 12 radio-collared lynx and all 4 family groups present were detected.
• 2001: 500 volunteers covered 286 lines, only about one and a half nights after snowfall. Lynx tracks were on 6% of lines; 2 of 4 radio-collared lynx and 1 of 2 family groups were detected on the lines. A member of the public found one more collared lynx, bringing detection to both family groups and three of four lynx.
• Most volunteers identified lynx tracks correctly, but 16% (1999) and 6% (2001) of the checked lynx track reports were misidentified. The tracks confused with lynx were those of wolverine, red fox, mountain hare and domestic dog.

3. What they concluded

• A structured snow-track survey can give a fairly effective index for monitoring large changes in lynx numbers, and it also secures a minimum, constant search effort for finding family groups for the annual minimum counts, as long as enough volunteers take part.
• Only large changes (25 to 50%) are likely to be detected, as in other track-count studies, so the method’s power should not be overestimated. Detecting a change between two surveys is much easier than detecting a slow trend.
• Cutting the lines from 100% to 75% (about 15 to 20 lines per family group home range) made little difference, but cutting to 50% clearly reduced effectiveness.
• Deliberately placing lines where lynx are likely to cross should be standard practice. Random lines may also be impossible to survey in rough terrain because of avalanche risk or cliffs.
• They scored each line as crossed or not crossed, because counting individuals from tracks is fraught with difficulty: snow changes track size and shape, and lynx tracks show very little difference by age or sex. In good, stable snow, nearly every lynx that crosses a line leaves tracks.
• The results come from a saturated population with very large home ranges. Lower-density populations, shorter nightly movements, different kill rates, or gentler landscapes would need different line densities and may show less difference between deliberate and random lines. Comparing densities between areas from an index should be done with caution.
• Volunteer hunters are key to keeping the survey cost-effective, and some misidentification shows a need for more education and field checks of reported tracks. Involving local people may also reduce conflict over population numbers. Tracks of other rare species, such as wolves and wolverines, add value.
• Several mutually supportive methods, along with other data such as livestock losses and the age and sex of harvested lynx, should be used when adjusting hunting quotas.

4. How tracking was or was not used

Tracking is the subject of the study. The simulated survey is a count of transect lines crossed by lynx tracks in the snow, scored as present or absent rather than by number of individuals. In the two field trials, volunteers skied 3 km lines one to three nights after snowfall, identified and recorded lynx tracks, and many continued along unstructured routes; the researchers checked more than 90% of the lynx track reports for correct identification. Lynx were also captured for radio-collaring partly by treeing with trained dogs and by snares placed at kills.

5. Tracker or observer certification

No certification is reported. The volunteers were mainly local lynx hunters, and the transect lines were placed by local experts. Their identification skill was tested in the study itself: the researchers checked over 90% of the reported lynx tracks and found 16% misidentified in 1999 and 6% in 2001. The authors call for further education and field checks of reported tracks.

6. Other technologies, skills or innovations

• Radio-collared lynx (including implanted transmitters in kittens), captured with snares at kills, walk-through box traps, darting from helicopters and treeing with trained dogs.
• GIS simulation that overlays real animal movements on virtual transect networks to test a track survey design before running it in the field.
• Resampling simulations to estimate statistical power.

7. Implications for conservation (from the paper)

• Lynx in Norway are hunted at 20 to 30% of the population a year, and there are no unhunted refuges to buffer overharvest, so good data on numbers and trends are needed to set annual quotas.
• Surveys that depend on chance reports from the public cannot control search effort, and the public tends to report less once lynx become familiar. A structured survey gives a constant minimum effort.
• The findings for lynx should apply broadly to large carnivores moving in landscapes where a suitable track surface occurs.