Digital forest inventory by drone: Mapping DBH in UAV data from tree stem shadow features

Forest inventories are traditionally conducted at regular time intervals during field campaigns measuring trees manually. In general, a small subset of trees is selected for data collection, information about the whole forest stand is obtained by extrapolation. These methods are time-consuming and tend to be accompanied by great uncertainties as the sample might not be representative for a heterogenous forest stand. At the same time, developments within the German forestry sector reinforce the need for an accurate and up-to-date forest data base even more: a) large-scale forest damage due to climate change related drought and heat events b) lack of personnel in parts of the forestry sector c) tendencies of a forest conversion from monoculture stand to more mixed forest areas d) an increasing digitalization strategy being pursued. To address this need, structure from motion (SfM) data products acquired using unmanned aerial vehicles (UAV) are a cost-effective method to derive forest parameters. In the "Shadow" project spectral and geometric information from UAV data is being used to develop methods for automated derivation of forest parameters such as diameter at breast height (DBH), tree stem positions, individual tree crown delineation, coarse wood debris, etc. as well as secondary parameter like timber stock. One focus of the method development within the project is its orientation to user-friendly tools and practical needs of potential users of digital inventory methods. So far, for example stem coordinates and crown delineation could be generated using a combination of leaf-off and leaf-on UAV data sets and point cloud-based algorithms. Also, the DBH as an important forest parameter, e.g. for estimating wood supply, biomass and stem growth rates, has been derived using classified cast shadows of tree trunks. Leaf-off data has been acquired over a deciduous forest stand near Jena and in the Hainich National Park, Germany, during sunny weather conditions. Using SfM a point cloud has been generated from the UAV images and normalized with respect to the relief. Points belonging to the tree canopy and stems are removed resulting in an orthomosaic image containing only ground information. In a second step, deep learning models have been tested to achieve an automatic detection and delineation of cast shadows. As the shape of the cast shadow and of the stem are correlated, parameters such as DBH can be derived from the detected shadows