engleski

Modelling and estimating height of lowland oak forests using various 3D remote sensing data

Forests are the most widely distributed terrestrial ecosystem on the earth, and they provide many direct and indirect benefits for human society. Sustainable management of forests requires spatially explicit information about their state and development. These information are usually acquired through field-based forest inventories which can be very time-consuming and labour intensive. Remote sensing (RS) data present alternative which can reduce field work and improve the efficiency, but the accuracy of obtained results has to be carefully tested and evaluated. The main goal of this study is to investigate the capability of various remote sensing data for use in forest inventory, with a special focus on estimation of plot-level mean tree height. The study was conducted in the pedunculate oak forests of Pokupsko Basin complex located in Central Croatia. Ground-truth data were collected by field measurements of diameters at breast height and tree heights from a total of 105 systematically sampled circular plots with radii of 8 or 15 m. The mean tree height (Lorey’s mean height) for each plot was calculated. RS estimates of plot-level mean tree height were obtained from four different datasets: (i) Airborne Laser Scanning (ALS), (ii) stereo WorldView-3 satellite images, (iii) stereo aerial images, and (iv) stereo Unmanned Aerial System (UAS) images. Namely, point clouds were generated for each RS dataset and normalized with ALS digital terrain model. From each normalized point clouds and for each plot, various height and density metrics were extracted and calculated. These metrics were then further considered in the statistical modelling of Lorey’s mean height as potential independent variables. RS estimates were evaluated with ground-truth data using leave-one-out cross- validation method. As expected, ALS provided the highest accuracy, but all other RS datasets (satellite, aerial, and UAS images) confirmed the great potential for plot-level forest inventory.

3D remote sensing data ; Airborne Laser Scanning (ALS) ; stereo WorldView-3 satellite data ; aerial photogrammetry data ; Unmanned Aerial System (UAS) ; plot-level tree height

nije evidentirano