Protecting forests with visual computing

Climate change is causing extreme climate events, windstorms, wildfires, and pests to become increasingly frequent, threating forest ecosystems. Action is urgently needed to protect the multi-functional role of forests—to maintain water and soil quality, reduce air pollution, and store carbon.

Fraunhofer IGD is looking at ways to support policymakers and forest managers in developing and maintaining sustainable and adaptive forestry practices. For example, researchers are using AI-enabled visual computing to predict forest fires and analyze huge volumes of geoinformation more precisely and more rapidly.

Predicting forest fires around the globe

Forest fires can now be predicted with great accuracy around the world, and not just for isolated regions. Researchers from Fraunhofer IGD collaborated with start-up Karuna Technology to develop a solution that leverages multispectral data from the Sentinel-2 Earth observation satellite, launched within the scope of the European Union’s Copernicus program. These data are fed into an artificial neural network, together with forest inventory, location, and meteorological data, including temperature, wind speed, and wind direction. The self-learning algorithm then processes this information to determine the probability of forest fires up to two months in advance. In particular, this enables Fraunhofer to assist countries that do not have the data or infrastructure required to conduct these risk assessments themselves.

Making forests more resilient with AI

Fraunhofer IGD is working with the Fraunhofer Austria Center for Data Driven Design to develop a further solution for climate-adaptive forest management. The premise is the same: For certain geographical areas, the detailed data needed to make forests more resilient are often not available. To overcome this issue, artificial intelligence takes known decision-making patterns and applies them to regions where data are lacking. This results in wide-scale environmental monitoring based on high-resolution Earth observation data. Various tools can then be employed to answer a wide range of questions, including: What are the most important tree species? What condition are the trees in? Have they been affected by pests or disease? What tree species are best suited to a particular location? And what is the carbon sequestration potential of a given forest? In the future, these tools will be combined with a shared data repository on a scalable cloud-based platform, enabling interdisciplinary collaboration.