A pioneering study from the University of Bari Aldo Moro introduces DIAMANTE, a cutting-edge technique using satellite imagery to detect forest dieback due to bark beetles, vastly improving forest management and conservation efforts.
In a significant advancement for environmental science and forest conservation, researchers at the University of Bari Aldo Moro in Italy have unveiled a pioneering method to map forest dieback through satellite imaging. Their study, published in the Journal of Intelligent Information Systems, presents a novel approach named DIAMANTE, which leverages data-centric semantic segmentation to detect tree dieback events triggered by bark beetle infestations.
Forests play an essential role in maintaining environmental balance, covering one-third of the Earth’s surface. They are crucial for carbon sequestration, water regulation, timber production, soil protection and biodiversity conservation. However, with accelerated climate change, forests face numerous disturbances. Insect infestations and diseases can drive large-scale tree dieback, disrupting ecosystem dynamics and services.
Monitoring and assessing such events has traditionally relied on time-consuming and labor-intensive field surveys, limiting the scope of surveillance. In contrast, remote sensing through Earth observation missions offers a transformative opportunity to scale up surveillance across expansive areas.
Recognizing this potential, researchers developed DIAMANTE (Data-centrIc semAntic segMentation to mAp iNfestations in saTellite imagEs). The approach utilizes a U-Net-like model trained on a labeled dataset drawn from both Copernicus Sentinel-1’s Synthetic Aperture Radar (SAR) data and Sentinel-2’s multi-spectral optical data.
The effectiveness of DIAMANTE was evaluated against real inventory data of forest scenes from northeastern France, collected in October 2018. These areas were hotspots of bark beetle infestation, which had surged due to mass reproduction that year.
The study’s results highlighted the advantages of using multisensor data. While using Sentinel-1 alone proved inadequate, Sentinel-2 alone achieved satisfactory results. However, the combined use of both significantly reduced false alarms and enhanced delineation of infested areas. Notably, signs of bark beetle attacks could be identified with reasonable accuracy up to a month before ground truth data acquisition.
Despite these promising results, early stages of the dieback remain weakly detectable via satellite imagery. This gap underscores the need for further research to leverage historical data for more robust temporal and spatial transferability. The ultimate goal is to enable direct deployment of a trained model across different regions and time periods efficiently.
This groundbreaking research is part of the SWIFTT project, which strives to equip forest managers with accessible remote sensing tools supported by Copernicus satellite imagery and powerful machine learning models. These tools aim to detect and map various forest threats, including insect outbreaks, wildfires and windthrow.
The findings of this study could revolutionize forest management practices, allowing for early detection and better management of forest health. By providing a scalable solution to monitor vast forested areas, it offers a significant step forward in forest conservation efforts worldwide.