Opportunities for LIDAR to improve and validate tree data sets in the Netherlands

Organisator Laboratory of Geo-information Science and Remote Sensing

do 10 april 2014 09:30 tot 10:00

Locatie Gaia, building number 101
Droevendaalsesteeg 3
6708 PB Wageningen
+31 317 48 16 00
Zaal/kamer 1

By Uliana Volkova (Russia)


Light detection and ranging (LiDAR), an active remote sensing (RS) technique, is able to describe tree structural attributes by measuring responses of emitted laser pulses from a tree. Alterra in cooperation with Geodan and NEO created a tree database which was based on raster airborne laser scanning (ALS) LiDAR data and included information about tree locations, tree crown projection perimeters, and several tree shape parameters according to SILVI-STAR method. SILVI-STAR method was was designed to save tree attributes as a parameterized, 3D model.

The aim of the study was to find a new way of delineating trees and deriving their parameters using terrestrial laser scanning (TLS) and ALS LiDAR point data, which can improve the existing method of tree parameter extraction and to assess the quality of the existing method.

The delineation essence was in extraction of aggregated and classified as high vegetation TLS points. Individual TLS and ALS tree point clouds were processed in R to derive tree parameters. Visual assessment showed that tree height, peripheral height and peripheral points extracted from TLS point data were accurate in more than 90% of cases, while the same parameters extracted from ALS point data were correct only in 46-77% of the cases. Meanwhile, only in 60.41% of cases the height of the first living fork and in 63.07% of cases the DBH and tree location have been correctly computed, thus the method of tree parameter extraction from point data needs improvements. The suggested enhancements are the separation of solitary trees from the aggregation of trees and creation of a better technique for noise removal.

The quality of tree parameters derived from ALS raster and point data was assessed using parameters derived from TLS point data. Based on the results of validation, the tree location calculation from ALS raster data was not as precise as urban tree managers may be requiring, producer’s accuracy was 0.23 and user’s accuracy was 0.15. However, this result is not very much reliable, as only in 63.07% of cases tree locations derived from TLS point data were correct. In addition, height parameters retrieval from ALS data were reliable only for tree heights extracted from ALS point data (R2 = 0.71 and RMSE = 3.89). This could be explained by the fact that the accuracy of ALS raster data is much lower, which is caused by the low density of points and averaging these points during the transformation of point data into 0.5m×0.5m raster cells. The extraction of periphery points from ALS point, as well as raster, data was perfect (R2 was no smaller than 0.9). In general validation showed that extraction of parameters using ALS point data rather than ALS raster data gives more precise result.

Undoubtedly, extraction of tree parameters from TLS point data is more precise compared to ALS data. However, taking into account the fact that acquiring TLS data in comparison with ALS data for the same area takes more time and requires more labour, the perspective of using ALS point data to derive tree parameters seems more realistic.

Keywords: Terrestrial Laser Scanning; Airborne Laser Scanning; LiDAR; SILVI-STAR tree parameters