Road Ecology Center

Animal detection systems use high-tech equipment to detect large animals when they approach the road. Once a large animal is detected, warning signs are activated urging drivers to reduce their vehicle speed, be more alert, or both. Lower vehicle speed and increased alertness may then lead to fewer and less severe collisions with, for example, deer (Odocoileus sp.), elk (Cervus elaphus), or moose (Alces alces)). For this study, we investigated the reliability of the animal detection system installed along US Hwy 191 in Yellowstone National Park, Montana, USA. The system was designed to detect elk and stored all detection data, including the detection zone in which the detection occurred, and a date and time stamp. Interpretation of the detection data suggested that at least 47 percent of all detections were related to animals crossing the road. However, animals walking in the right-of-way or medium-sized mammals (e.g., coyotes, Canis latrans) do not generate a clear detection pattern, and were, therefore, classified as “unclear.” Therefore, the 47 percent should be regarded as a minimum estimate. The timing and direction of travel of crossing events, indicated by detections on opposite sides of the road, matched local knowledge about the behavior of the elk, suggesting that the system was able to detect large animals, specifically elk, and that the data were interpreted correctly. We also compared the spatial distribution of the crossing events with snow tracking data. The spatial distribution of the crossing events and elk tracks showed a close match, again suggesting that the system was able to detect elk, and that the data were interpreted correctly. Almost 87 percent of all elk crossings recorded through snow tracking could be linked to a crossing event detected by the system. However, medium-sized mammal species, such as coyotes and wolves (Canis lupus), were not or rarely detected. Furthermore, we identified the presence and location of blind spots (potentially 17.8% of the total length covered by the sensors). Blind spots were defined as locations where the system failed to detect a human crossing between the sensors. Most of the blind spots were due to curves and slopes that caused the detection beam to shoot too high above the ground. The total time for which the flashing warning lights would have been activated was estimated at one hour and 13 minutes per day, a marked difference compared to permanently activated warning signs. Most crossing events (72.6%) were completed within three minutes, and the median duration of a crossing event was one minute and 29 seconds. If the warning signs would be activated for three minutes after the last detection, the signs would have been continuously activated for 88.1 percent of all detection intervals (i.e., time between consecutive detections) during crossing events. Similarly, 78.1 percent of all crossing events would have had the warning signs continuously activated while the crossing was in process. We conclude that the system reliably detects large animals, especially elk, but the system does not detect all elk that cross the road, e.g., because of blind spots. In addition, a three-minute activation period for the warning signs appears to be a good balance between keeping the signs turned on while elk are in the process of crossing the road, and not presenting drivers with activated warning signs longer than necessary.