TAIWAN-WILDLIFE-CONSERVATION-ANIMAL
The Taiwan Roadkill Observation Network proposes using artificial intelligence to predict high-risk roadkill sections for wild animals and combining it with sensors and road warning devices to reduce the risk of animals being hit by vehicles and related traffic accidents. The image illustrates wildlife road safety and ecological conservation.

【TAIPEI, Oct. 26】The Taiwan Roadkill Observation Network has proposed using artificial intelligence to predict road sections where wildlife is most likely to be struck by vehicles and establish a real-time warning system, in a bid to identify high-risk areas before accidents occur while reducing both traffic incidents and wildlife deaths.

According to a Focus Taiwan report on Oct. 26, the Taiwan Roadkill Observation Network (TaiRON) said an average of about 3,000 traffic accidents a year over the past eight years were caused by animals suddenly entering roads. Motorcycles accounted for 95% of those incidents, which resulted in a total of 25,345 injuries during the period.

Citing government statistics, the organization said the direct economic losses from such accidents are estimated at NT$800 million to NT$6.4 billion annually. However, existing statistics mainly cover more serious accidents, meaning the actual number of incidents may be higher.

The project will integrate multiple data sources, including wildlife mortality records, data from the Taiwan Biodiversity Information Alliance, public reports, traffic accident records, road networks, speed limits, traffic volumes, microclimate data and Google Street View images. AI will then be used to identify environmental conditions associated with a higher risk of wildlife roadkill.

The system is intended not only to map known accident locations, but also to identify high-risk road sections that have not previously been recognized. Pangolins will be the main test species in the initial phase, after which the team will assess whether to expand the system to other wildlife, including leopard cats, land crabs and frogs that gather in large numbers during breeding season.

In addition to the prediction system, the team has proposed combining variable-message road signs with real-time warning devices. Some of the devices use infrared sensors to detect warm-blooded animals approaching a road, then use light sensors to identify vehicles and alert drivers in low-light conditions.

The warning equipment will first be tested on roads managed by the New Taipei City Government before any wider deployment is considered. The proposal has advanced to the final round of the 2026 Presidential Hackathon, but its implementation will depend on the test results, cross-agency data integration and funding.

Since its establishment in 2011, the Taiwan Roadkill Observation Network has amassed a large number of public reports on wildlife deaths. If AI forecasting can effectively identify high-risk locations, it could help authorities more precisely allocate warning equipment, road improvement projects and wildlife conservation measures in the future.