engleski

Estimation of Forest Stand Attributes Using Image- Based Canopy Height Model

The aim of this research was to investigate the applicability of canopy height models (CHMs) derived from digital aerial images for estimation of the main forest stand attributes (mean diameter, mean height, stand density, basal area, volume) of even-aged pedunculate oak forests. The research was conducted in the management unit Topolovac in Spačva basin (eastern Croatia), one of the largest coherent complex of lowland pedunculate oak forests in Europe. A CHM of the research area was generated by subtracting ground from surface elevations, i.e., by subtracting a digital terrain model (DTM) from corresponding digital surface model (DSM). For the DTM generation, a national digital terrain data were used. First, a triangular irregular network was created from the digital terrain data, which was then converted through linear interpolation into a raster DTM of 1 m grid size. For DSM generation, the colour infrared digital stereo aerial images collected as a part of the regular national survey in 2011 were used. A DSM of 1m grid size was generated using Dense DTM image matching algorithm (PHOTOMOD 5.24 software). In order to develop models for estimation of forest stand attributes, the data from field surveys conducted within regular forest inventory in 2011 were used as dependent variables, whereas metrics extracted from raster CHM were used as independent variables with addition of several variables from forest management plan (stand age, soil type, site class). A large number of potential independent variables were extracted from raster CHM, and only variables with the strongest correlation with field stand data were used in statistical modelling (multivariate linear regression). In total, 60 stands were used for models development, while 30 stands were used for models validation. The validation of stand-level models resulted with the following R2 values between observed and predicted estimates: 0.86, 0.91, 0.76, 0.71 and 0.80 for mean diameter, mean height, stand density, basal area and volume, respectively. The obtained results indicate that CHM derived from digital aerial images and national digital terrain data has potential for operational use in forest inventory of lowland pedunculate oak forests.

aerial images ; canopy height model ; digital photogrammetry ; forest inventory ; image matching

nije evidentirano