Snow-tracking versus radiotelemetry for predicting wolf-environment relationships in the Rocky Mountains of Canada

1. What they did

Radio-collaring is one of the most common ways to learn where animals live, but it means capturing and handling them, and it can be costly, slow and prone to location error. Snow-tracking avoids this. The authors asked whether a model built only from wolf tracks in snow could predict where wolves occur as well as radiotelemetry, the standard method. They knew of no earlier study that had compared the two for a large carnivore.

• Tracks in snow (winters 1997 to 2000). About 24 hours after every snowfall, they drove Highway 40 and the Smith-Dorrien Trail in Kananaskis Country at 15 to 20 km per hour and recorded every place wolves had crossed the road, then repeated the survey every 3 to 4 days until the next snowfall. They also walked 1 km transects running out at right angles from the roads, 24 to 120 hours after snowfall. Locations were recorded with Global Positioning System (GPS) units.
• Building the model. They compared wolf track locations with randomly placed “pseudo-absence” points (places where wolves were not known to be present), using maps of terrain (elevation, ruggedness, aspect), vegetation from satellite images (greenness, wetness, forest cover) and elk and deer track density. They chose the best model with logistic regression and Akaike’s Information Criterion (AIC), and turned it into a map of the probability of wolf presence across Kananaskis.
• Testing the model. They checked the map against 331 radiotelemetry locations of 2 radio-collared wolf packs whose home ranges overlapped the tracking area, collected from the ground and air in the early 1990s.

2. What they found

• The best model: wolves were more likely to be found in flatter, less rugged terrain, in places with less forest cover within 500 m, in places with a higher “wetness” value (which the authors take to mean older, more complex forest), and where elk and deer tracks were denser.
• This model was 2.4 times better supported than the next best, which also included aspect (wolves favouring more southern slopes). The third model was the same as the best, but without deer.
• Against telemetry: the track-based map matched the radiotelemetry locations well. The area under the curve was 0.78, meaning the model told wolf locations from non-wolf locations correctly 78% of the time, which the authors call acceptable for this kind of spatial modelling. Telemetry locations also fell in places the map rated as high probability.
• Wolf movements almost always crossed transects at right angles; wolves rarely wandered along or back and forth across a transect.

3. What they concluded

• Track data give a reliable, cost-effective way to map wolf distribution and predict wolf-environment relationships in mountainous regions.
• Wolves may use flat, more open forest to save energy while travelling, while staying close to elk and deer in valley bottoms. Avoiding rugged terrain may help wolves share the landscape with cougars, which favour rugged ground.
• The snow-tracking data only cover winter, while telemetry covers the whole year. The authors argue this matters less than it seems: wolves move down to lower ground in winter and use the same valley-bottom paths in all seasons, so winter routes may be especially important to them.
• Snow-tracking cannot replace telemetry, which gives information on dispersal, individual identity and social ties, and telemetry cannot fully replace snow-tracking. Which to use depends on the species, the question, the place, the funding and logistics. The two can be combined to offset each other’s weaknesses.
• The model may be extended to other parts of the Rocky Mountains, within reason, but not to less rugged terrain.

4. How tracking was or was not used

Tracking was the core of the study. Teams identified wolf tracks in snow at road crossings (from a slow-moving vehicle) and on foot along transects, and recorded elk and deer tracks to map prey density. The radiotelemetry data were also checked with tracking: every telemetry location was confirmed by finding wolf tracks on the ground. No trailing is described.

5. Tracker or observer certification

Not reported. The paper does not say who identified the tracks or how their skill was checked; field assistants are thanked. The author notes say D. Joanne Saher has extensive experience in large-carnivore telemetry and backtracking research. Because the whole model rests on correct identification of wolf, elk and deer tracks, observer skill matters to the findings.

6. Other technologies, skills or innovations

• Handheld Garmin GPS (location error 50 to 130 m) for road crossings, and a backpack Trimble Pathfinder GPS, corrected to within 1 m, for transects.
• Geographic Information System (GIS) mapping in ArcView, with Landsat 7 satellite images processed into greenness, wetness and a Normalised Difference Vegetation Index (NDVI).
• Pseudo-absence points drawn only from the area actually surveyed (within 130 m of roads and 60 m of transects), to reduce bias.
• Verification against an independent telemetry dataset with a Receiver Operating Characteristic (ROC) curve.

7. Implications for conservation (from the paper)

• For some research goals, such as modelling wolf-environment relationships in mountains, location data from non-invasive snow-tracking could reliably be used instead of radiotelemetry data.
• Funding for large-carnivore research is increasingly hard to get while the need is greater than ever. Tracking is a cost-effective, reliable way to run long-term surveys of wolf distribution and support management and conservation decisions.
• Many remaining wolf populations live in rugged mountain terrain, where track-based models may be a highly reliable inventory method when funding is limited.