3 D FOREST STRUCTURE ANALYSIS FROM OPTICAL AND LIDAR DATA

In Austria about half of the entire area (46 %) is covered by forests. The majority of these are highly managed and controlled in growth. But besides timber production forest ecosystems play a multifunctional role including climate control, habitat provision and, especially in Austria, protection of settlements. The interrelationships among climatic, ecological, social and economic dimensions of forests require technologies for monitoring both the state and the development of forests. This comprises forest structure, species and age composition and, forest integrity in general. Assessing forest structure for example enables forest managers and natural risk engineers to evaluate whether a forest can fulfil its protective function or not. Traditional methods for assessing forest structure like field inventories and aerial photo interpretation are intrinsically limited in providing spatially continuous information over a large area. The Centre for Geoinformatics (Z_GIS) in collaboration with the National Park Bayerischer Wald, Germany and the Stand Montafon, Austria, has tested and applied advanced approaches of integrating multispectral optical data and airborne laser scanning (ALS) data for (1) forest stand delineation, (2) single tree detection and (3) forest structure analysis. As optical data we used RGBI line scanner data and CIR air-photos. ALS data were raw point data (10 pulses per sqm) and normalised crown models (nCM) at 0.5 m and 1 m resolution. (1) Automated stand delineation was done by (a) translating a key for manual mapping of forest development phases into a rule-based system via objectrelationship modelling (ORM); and (b) by performing multi-resolution segmentation and GIS analysis. (2) Strategies for single tree detection using raw ALS data included (a) GIS modelling based on a region-growth local maxima algorithm and (b) object-based image analysis using super class information class-specific rule sets. (3) Vertical forest structure has been assessed statistically by (a) applying basic statistics (like mean, standard deviation, and variation coefficient) on the raw data using a moving window approach; and (b) by applying landscape metrics (Shannon Evenness Index, SHEI, and division index, DIVI) for different strata extracted from the nCM.