Adaptation of forest inventory information obtained from remote sensing technology data to support spatial and operational planning of LVM forestry Project leader Dagnis Dubrovskis Duration 2018 - 2018 Research focus 4. Development and adaptation of technologies for obtaining high-value agricultural and forest products, as well as in veterinary medicine Source of funding Collaboration research Project partners A/S Latvijas valsts mežiJānis KrūmiņšRaivis BaltmanisIngus Šmits Description of project The aim of research The aim of the project is to adapt the obtained forest inventory information to support LVM's forestry spatial and operational planning using the LiDAR combined forest inventory method (hereinafter CFI) from remote sensing technology data. The tasks of project: Within the project, so far acquired data and methods of processing were adapted to the formats used by LVMThe requirements for adaptation of taxation data processing algorithms were specified and algorithm solutions for using the results in the LVM system were prepared, as well as a methodology for collecting data necessary for calibration of the data and for calibration of data processing algorithms using LVM internal sample plots and harvester production data were prepared Results The algorithms have been tested in the vicinity of Jelgava by instrumental taxation methods, but in the Northern Kurzeme by comparing harvester production data with forest inventory data. Estimation of the accuracy of the calculation of taxation data in accordance with the existing legislation have been done. It was found that in the sample territory of the Jelgava area the best results for determination of volume for standing timber in sample plots were indicated for the pine (determination coefficient R 2 = 0.92), but for other tree species the coefficient of determination is lower, respectively for the spruce R 2 = 0.75, birch R 2 = 0.84, black alder R 2 = 0.78, aspen R 2 = 0.70, grey alder R 2 = 0.66. For all species together R 2 = 0.72. Automatic recognition of tree species was carried out on individual trees. The best results were obtained at a pixel size 0.15m on the ground. 88% of pine, 88% of spruce, 53% of birch, 17% of black alder, 84% of aspen and 88% of grey alder trees were recognized. Good results were obtained for determining the height and average height of a single tree, however, the results for determination of average height and diameter of the stand were not conclusive, as they depended on the number of elements of the stand and the differences in their height. Testing of algorithms in Northern Kurzeme showed significant differences in results between forest inventory data, CFI results and harvester production data. A weak correlation (R 2 = 0.34) between forest inventory (taxation) and production data was found, while CFI results and production data had close correlation with R 2 = 0.78. There is a weak correlation between the CFI and the taxation data with R 2 = 0.36, since the accuracy of the forest inventory (taxation data) is not high. Comparing CFI calculated standing stock with production data, the average relative error is 10.7%, which shows the possibility for integration of these results into the forest inventory system. Slightly close correlation was found also between the average tree height in taxation data and the CFI results. Unfortunately, the correlation between the other taxation data was not sufficient, but the results of the project are influenced by the fact that the accuracy of forest inventory data could not be verified. Testing of algorithms and determination of limit values for LiDAR data collected on an unmanned platform (drone) for estimation of ditch overgrowth and calculation of biomass yield have been done. The highest result for obtained correlation between the manually measured biomass and the scanned ditch tracks with R 2 = 0.643 showed algorithm configuration parameters used in the second iteration. Unfortunately, it was not possible to ascertain the accuracy of the results, as the piles of felling residues were not chipped during the project period. It was found that the use of drones for data acquisition is useful in individual felling areas or areas up to 50ha, while for scanning of larger areas is more efficient with use of a helicopter or airplane platform.
