Forestcare: Monitoring Forest Quality Using Satellite and Drone Data

Forest quality declines through climate-changed induced droughts and bark beetle infestations. In order to quickly and appropriately respond to these challenges, forest monitoring with a high spatial resolution is necessary. Hence, we propose an HPC-based monitoring system that combines different sensor data to predict the vitality of single trees. In this project, we will first extract predictive attributes of tree healthiness. Beyond manual classification to anchor the ground truth, we have and will collect multispectral images and LiDAR data obtained by unmanned aerial vehicles. Furthermore, chemical information emitted by trees when stressed will be collected via an electronic nose on the drone. The processing of this data is computationally intensive and thus requires the usage of HPC. In particular, we will use sensor data fusion to combine all features for an accurate single-tree-based vitality evaluation. We will use deep learning, specifically convolutional neural networks on the image data, to generate this classification. Lastly, we will integrate satellite data (SAR, multispectral images) to test the validity of our vitality tree predictions. With our HPC-approach to forest monitoring, we will build a base for tailoring policies efficiently towards quickly fighting bark beetle infestations and towards fast responding to first signs of droughts on a larger scale.

---

Contributors:

• Dorothea Müller (Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen (GWDG))