Result TO01000315-V3
Within the following text, you’ll find an in-depth exposition of the findings from TO01000345-V3 – specialised maps of forest variables related to production and regulation functions for selected areas in the Czech Republic and Norway derived from airborne remote sensing data. This result includes a description of the estimation of forest aboveground biomass (AGB) from airborne laser scanning (ALS) data at DendroNetwork sites.
Forest AGB maps are based on ALS data acquired in September and October 2021 and September 2023 over 68 DendroNetwork sites. Each ALS data acquisition was organized as a pair of perpendicular flight lines over the centre of a DendroNetwork plot. Flight altitude and other acquisition parameters resulted in a point cloud density of around 30 points per m2 in the region of interest. The ALS data underwent standard pre-processing steps including wave decomposition, geometric correction and stripe adjustment. In the second step, the noise points (if present) were removed from the cloud, cloud points were classified as terrain, buildings, higher vegetation and other, and the Z coordinate was recalculated from altitude above sea level to height above terrain.
Field estimates of AGB are computed from forest inventory data measured for all trees within a 30 m × 30 m plot and allometric equations. Field level AGB obtained from forest inventory measurements were used to validate ALS-based methods. AGB estimation from ALS data uses the area-based approach described in Brovkina et al. (2022). The methodology combines predictors derived from point clouds of ALS data and a machine learning model. For each DendroNetwork site a regular grid with a step of 2 m was constructed, spanning 65 m to all four directions from the plot centre. At each grid node, a cylindrical sub-cloud was cut out and a set of statistical variables describing the distribution of the point sub-cloud was calculated. These predictors were fed into a machine learning model to obtain an AGB estimate for the particular grid node. We used a machine learning meta-model trained and verified on the field data from several forest sites in the Czech Republic independent of the DendroNetwork data.
References
Brovkina O, Novotný J, Navrátilová B, Hanuš J, Ciencala E, Albert J (2022) Forest aboveground biomass assessment using an area-based approach (verified technology). Global Change Research Institute CAS, Brno.