Japanese scientists invent AI tool to predict bear encounters as attacks rise

Researchers at a Japanese university have developed a system that uses artificial intelligence to determine the possibility of encountering a bear within 1 sq km (0.4 square miles), a critical new tool as the number of people killed or injured by bears is expected to hit a record high. Lead developer Yusuke Fukazawa is an associate professor of applied machine learning at Sophia University who enjoys orienteering and hiking and hopes his studies will help to create an application that makes such outdoor pursuits safer. He points out, however, that encounters between bears and humans are increasingly occurring in urban areas across Japan. “I have focused my research on Akita prefecture and it was clear that there was a surge in bear encounters in 2023, rising to 3,910 incidents from just 800 cases the year before,” Fukazawa said. Those encounters led to a number of people sustaining injuries, with 193 bear attacks across the country over the course of 2023 leaving 212 people injured and six fatalities. “In previous years in Akita, there were very few human-bear encounters in urban areas, but that changed in 2023,” he told This Week in Asia. Wildlife experts say there are a number of reasons why confrontations between bears have been increasing, but most point to the poor crop of acorns and beech nuts in the bears’ mountain habitats as the primary reason. Hungry and preparing to go into hibernation as autumn approached, many became increasingly desperate for food and started to extend their foraging range into towns and even the suburban areas of major cities. Fukazawa’s AI-based system creates a 1km-by-1km grid across all of Akita prefecture and applies a range of data points to determine the likelihood of a person encountering a bear. The input information includes previous sightings in the area, the local environment, which can range from mountain forest to city centre, weather conditions, elevation, population index and the presence of roads. The AI tool predicted encounters with a precision of 63.5 per cent, Fukazawa said, although he admitted he had hoped to reach 75 per cent accuracy. “We want to use more real-time data, such as weather forecasts and actual human-bear encounters, to improve the accuracy,” he said. Fukazawa is presently developing an easy-to-use application based on the system, giving each 1km block across the entire prefecture a colour corresponding to the likelihood of encountering a bear. People entering a red square, for example, are urged to take particular caution. He intends to extend the project to cover all parts of Japan that have bear populations. Beware of bears The roll-out of the system is timely. An elderly man and his wife were attacked when they went for an evening walk in the town of Minakami, Gunma prefecture, on Monday. The 76-year-old man sustained bite wounds to his hand and his wife, 69, was bitten on the arm and leg. Police believe the same bear attacked people at a nearby supermarket last week, injuring two people. On Sunday, a bear jumped onto the back of an 83-year-old woman as she prepared to feed cattle in Yamagata prefecture, causing minor injuries, while residents of Mikasa in Hokkaido and Nagaoka City in Niigata prefecture were warned to be alert after bears were seen in both communities. Tourists have also been attacked, with a Spanish man sustaining injuries in an incident earlier this month in the World Heritage mountain town of Shirakawa in central Japan. According to government statistics, there have been six deaths so far this year, equalling the record high set in 2023, but that figure is likely to rise as the bears forage ahead of going into hibernation. “A lack of beech nuts and acorns in the forests is a major reason why bears have been going into areas that they would previously have avoided, but it is not the only reason,” said Kevin Short, a naturalist and former professor of cultural anthropology at Tokyo University of Information Sciences. “Rural depopulation means there are just fewer people in mountain villages, so bears are exploring, and they are becoming less afraid of humans because there has been a sharp decline in the number of hunters,” he said. To keep bears at bay, Short suggested rewilding areas that had been turned into single-species plantations that did not provide the food and environment that bears needed. “More areas should be returned to native forest land and there will then naturally be fewer bears coming into towns.” He carries a walking stick with a bell when he is walking in areas that are known to have a bear population, as experts suggest that bears will avoid humans if they hear them coming. And while an app predicting the percentage likelihood of an encounter would be useful to some, Short said he preferred to use common sense. “If you are in a mountainous area of central or northern Japan or anywhere in Hokkaido, assume that bears are going to be nearby and take precautions,” he said.

Japanese scientists invent AI tool to predict bear encounters as attacks rise

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}The scientists are developing an app to colour 1km blocks across Akita according to the likelihood of meeting a bear

The scientists are developing an app to colour 1km blocks across Akita according to the likelihood of meeting a bear