Preparation of a tool for automatic detection of dead trees in selected spruce stands in Luxembourg forests using remote sensing. Technical and legal aspects of flying a drone (UAV).

The deteriorating condition of urban and rural forests caused by climate changes or biotic factors has become a burning topic for decision-makers and politicians, as evidenced by a parliamentary question (No 6806/27.09.2022) on the extent of their occurrence and prevention options in the Grand Duchy. The mass occurrence of trees infested by pests results in extensive economic losses through trees dying, leading to the depreciation of wood material, increased management costs, posing a threat to public safety on hiking trails and indirectly affecting human health. The challenges are the rapid detection localization of dangerous sites, their security, proper management, and recommendations for the ecological conversion of forests. Traditional monitoring methods based on in-situ field inventories often prove ineffective due to their labor- and cost-intensive nature. Thanks to the dynamic development of remote sensing (RS) techniques, efficient collection of information about the features of the earth's surface from satellite- or aircraft-based platforms, and unmanned aerial vehicles (UAV) combined with interactive humane-machine processing have revolutionized traditional surveys.
The aim of the project was to develop and test methods and tools for the automatic detection of dead and diseased trees in spruce stands based on remote sensing data for a selected municipality in Luxembourg. We tested geo-information technologies and methods to collect data and generate new information, which can contribute to effective forest and urban forest health monitoring. For this purpose, we used Geographic Object-Based Image Analysis (GEOBIA) based on the latest available RGB/CIR (colour-infrared) orthophotos and LiDAR point cloud. We developed a tool (a rule-set based on deep learning algorithms within the eCognition Developer software) for a selected municipality, where the problem of dead trees is very prominent, in order to investigate whether it will be possible in the future to carry out automated analyses for the whole country on available open data. On selected sample plots in the field, we acquired experimental ground photos with a camera with built-in GPS and aerial photos using a UAV (drone) to refine our GEOBIA algorithm and estimate classification accuracy. The prepared script/tool, which is the result of this pilot project, can be quickly adapted to the new data with minor modifications to perform monitoring on a cyclical basis. During this presentation, we will mainly focus on explaining the methodology, technical aspects and challenges and presenting the results.