Abstract
Estimating biomass and canopy structure is important to characterize forest ecosystems. The characteristics of vegetation at local level affect both ecological and hydrological processes. The increased availability of Airborne Laser Scanning (ALS) surveys and ALS-derived datasets in the last decade pushed forest scientists to find the way to properly exploit the detail information contained to extract forest characteristics. The topic of the present research is the development of a methodology for extracting the position and main characteristics of single trees and the volume of the aboveground biomass (AGB) from LiDAR data of Alpine forests. In the framework of the present research project two new single- tree extraction algorithms from LiDAR have been developed and integrated in an automatic calibration system based on Particle Swarming (PS). These tools have been tested and compared with a widely used single-tree extraction tool on two test areas, one of conifers and the other of broadleaves. The first study area covers approximately 10 km2 of a coniferous forest with different forest structures and it is located in the Aurina Valley (Italy). Adetailed LiDAR dataset was made available with a resolution of 10point/m2 and a vertical precision of 5-20 cm. A field survey has been carried out by the University of Bolzano on 12 circular plots of 15 m radius, where position, height, diameter and species of each single tree inside the plots have been recorded. The second study area covers approximately 15 km2 in Altipiano di Asiago (Veneto – Italy) of a mixed broadleaves forest with different forest structure. The data have been collected within the NewFor project and consist of a LiDAR survey with a resolution of 11 pt/m2 and a field survey on 5 circular plots of 20 m radius, where position, height, diameter and species of each single tree inside the area of the plots has been recorded. The validation procedure consisted in finding the matching between field and LiDAR-derived measurements at single-tree and plot level. The second research objective was to integrate the results of a single-tree approach in a GIS-based tool to predict the magnitude of Large Wood (LW) transport during flooding events. Therefore, resulting LiDAR-based vegetation was used for modeling large wood (LW) recruitment and transportation during flooding events and estimate the amount of LW potentially contributing to each section within a river basin. The PS has been proved to be a powerful method to parametrize in order to properly detect treetops of coniferous and broadleaves stands with complex forest structures. The optimization procedure together with the use of point cloud data allow to obtain high detection rates (matching rate 0.88 for coniferous and 0.53 for mixed broadleaves forests) and estimation accuracy of forest volume also in comparison to the most recent available literature data. The differences in performance between the methods were found to be higher for tree detection than for height and volume estimation. The single-tree extraction algorithms and LW model have been integrated in the Open Source GIS gvSIG.