The dissertation examines a range of issues related to the application of unmanned aerial vehicle (UAV) technology, optical imaging, and stereophotogrammetry methods for the inventory of pine stands in Ukraine. The dissertation was prepared to find the optimal parameters for remote sensing that will enable the most accurate mensuration of pine stand parameters. It also discusses the use of remote sensing methods for the
assessment of mensuration parameters of forest stands, stand volume, and the volume of biomass components.
In Ukraine, a considerable portion of forests comprises Scots pine stands, which hold important ecological and economic impacts within the country and globally. According to the State Forest Resources Agency, the total area of Scots pine forests in 2019 was 2.3 million hectares. Data collection information about forests and their comprehensive assessment of such an extensive territory is a very time-consuming process. The presence of areas contaminated with radioactive elements and hazardous remnants of objects from war poses a danger to personnel during pine stand mensuration on such territories. In such cases, remote photogrammetric data can be an important source of information about forest stands. Remote sensing data are widely used for forest mensuration, and optical images obtained from unmanned aerial vehicles and photogrammetry methods give a possibility for obtaining adequate, objective, and current data on forests.
The dissertation considered the historical background of implementing remote sensing methods in forestry. An overview of the systems used to obtain information about forest ecosystems, and their advantages and disadvantages have been discussed.
Considered were the specific features of data collection systems that apply to using unmanned aerial technologies. Analyzed the widely distributed methods of detecting individual trees in stands and the possibilities of measuring their characteristics. In addition, analyzed the main methods for modeling forest stands characteristics which can use individual tree detection data in a stand or area-based approach.
The author emphasizes the need for maximum digitization of the reference base and groundbased forest inventory data.
The study used 58 sample plots of natural and artificial strands of Scots pine in the Chornobyl exclusion zone.
Raster images of digital canopy height models were created from the remote stereophotogrammetry survey data, which represented the three-dimensional structure of the measured stands. Individual tree detection was performed using the local maximum search function in the «ForestTools» package implemented in the R programming language.
The study considered which parameters of spatial resolution and image overlap (which vary depending on the flight height and photo interval) would allow for accurate measurement of forest stand characteristics in pine forests. For this purpose, a study polygon was UAV data collected at three different flight heights and overlaps, resulting in nine combinations of image sets that were after processed using photogrammetric methods.
The results of modeling forest stand characteristics indicate the possibility of successful application of UAV remote sensing data for prediction of the number of dominant trees in plots, average height, diameter, basal area, relative stocking, and stand volume.
Mensuration of Scots pine forest stands parameters can be successfully performed using statistical distribution data of the CHM and individual tree detection data. A comparison of two approaches to modeling Scots pine forest stand characteristics indicates no preference for using a method that requires individual tree detection in the stand. For stands with relative stocking of more than 0.4, the difference between observed
and predicted stand characteristics was not significant. Therefore, it can be concluded that for the assessment of Scots pine forest stand characteristics, data from the statistical distribution of the canopy heights model and individual tree detection data can be successfully used. The estimated errors between predicted and observed stand characteristics with a relative stocking of 0.4 or lower indicate the necessity to allow this parameter, which can be taken into account using crown projection area coefficient of individual tree detection data or coefficient of variation of canopy height model.
