Bears in Western Trentino
The violence and danger represented by bears epitomizes human fear of nature. Yet despite their threatening appearance, bears play an important function within ecosystems by increasing the complexity and interdependence of the ecological community. Plantigrades are engineer species, that is, capable of modifying the environment. With their foraging habits, bears spread plant and berry seeds in their droppings and spread marine-derived nitrogen in forests surrounding streams with salmon. In addition, they represent important connections within food webs, and their predatory activity is a control mechanism for deer and other prey populations. Awareness of the importance of bears’ ecological function and their protection gave rise to the Life Ursus project, sponsored by the Adamello Brenta Nature Park and funded by the European Union.
The project, initiated in 1996, was based on the reintroduction of nine individuals to recreate in 20-40 years a bear population of 40-50 individuals. A key component of the program was the distribution of bear populations between “suitable areas” in western Trentino and the provinces of Bolzano, Brescia, Sondrio and Verona. Concluded in 2004, the project was deemed a success, but recently the positive results have been questioned following the Caldes tragedy, in which the tragic death of runner Andrea Papi turned the spotlight on plantigrade management in Trentino. Many observers have denounced the excessive presence of bears in this area and the failure of the project. Although this tragic episode is the first in Italy in the last 150 years, Legambiente points out that the coexistence and management of problematic situations has been obstructed by the absence of a communication and information campaign in Trentino-Alto Adige in the last 10-15 years.
Good bear, bad bear
Increased awareness among local communities in order to improve coexistence with plantigrades could be fostered by new technology resulting from a collaboration between ecologists at the University of Victoria and programmers in Silicon Valley that has led to the creation of artificial intelligence-based software that can identify the faces of brown bear individuals. The technology was called BearID and is based on Deep Learning software tested on thousands of photographs of 132 brown bear individuals from Alaska and British Columbia.
The program has achieved an accuracy rate approaching 84% by classifying each bear, that is, distinguishing individual from individual, while the facial recognition rate is up to 98%. The artificial intelligence decides for itself the facial data relevant to the identification of each bear, including the distance between different points on the face. This technology has multiple advantages being a non-invasive conservation mode for identification and study of plantigrade behavior and ecology. In combination with the use of photo traps, BearID can also be used for bear population estimates.
The technology is easily accessible and is offered as a substitute for expensive GPS collars. In addition to price, the most critical elements of the collar in the management of risk situations include battery life, which depends on the number of localizations made, and the delay in receiving data on the animal’s location as well as the need to capture the animal to implement the collar. In addition, the Trentino territory is characterized by alpine and prealpine environments whose geomorphological and vegetation peculiarities compromise the quality of satellite data. In addition, BearID can be a preventive tool to resolve conflicts between humans and bears.
In the presence of repeated risk situations, the software can help experts assign responsibility to one or more problematic individuals. Plantigrade population management activities may differ based on this information, providing a more specific picture of what is happening within the population or with an individual bear.
BearID’s main challenge is the absence of distinguishing features in the bear’s physiognomy, making it lacking individual identification within the population inventory. A valuable expedient that can bring technical improvements to the technology is the use of photo traps, which allow data to be collected in the field without human presence ranking as one of the most versatile and least invasive devices for biological conservation. Researchers at the University of Victoria are experimenting with new photo trap placements to obtain an image dataset to refine the methodology of using the software by improving its accuracy by making it accessible for other plantigrade species as well.

A graduate in Chinese Studies (Unibg and Unimc), he is currently pursuing a degree in Environmental Humanities at Ca’ Foscari University. He is a teacher of Italian in a public school in Beijing. He believes in the importance of environmental justice.

