In addition, there is an increasing interest in observing temporally current broad forest areas to determine the consequences of climate change on forest as soon as possible.
This information allows the responsible parties to initiate countermeasures early enough in order to preserve and expand the natural carbon sink function of the forest.
Measurements based on on-site surveys or airborne sensors capture the forest stand only in small areas at a single time and also involve high costs and labor. For these reasons, the collaborative research project between the Karuna Technology UG and the Department of Remote Sensing and Image Analysis attempts to provide an effective, low-cost and AI-based alternative to current forest mapping methods.
The goal of the project is the investigation of a method, which enables the simultaneous detection and spatial delineation of individual tree crowns and tree crown areas from satellite images. In addition, other characteristics such as tree species and canopy health should be assigned to each individual tree in the study area as well as a method for estimating the biomass on an individual tree level will be explored.
By using multisensory satellite data and developing a deep-learning architecture, these parameters will be recorded simultaneously in large-scale and ecologically varying forest areas in a precise and temporally up-to-date manner.
By doing so, the method attempts to overcome the disadvantage of existing methods that limit their procedure to the determination of very few forest parameters or cover only small-scale forest areas.
In the end, the essential single tree parameters of an individually selected and large-scale forest area should be provided to the user via a user interface.
The project is supported by the LOEWE funding line 3 “KMU-Verbundvorhaben”.