Q24 – B. Vandendaele

Methods development to estimate tree structural attributes using LiDAR on an Unmanned Aerial Vehicle

The recent developments of UAV and LiDAR systems allow new procedures to improve current forest inventories. The operational flexibility and high spatial and temporal resolution of UAVs combined with the capacity of LiDAR signal to break through forest canopy makes this technology a promising tool for the characterization of forest structure at the tree-scale. Due to the low altitude, the large scanner field-of-view, the small footprint size and the high effective scan rate, ULS data can be used to derive 3D models of individual trees and characterize understory vegetation. This project aims to develop methods to estimate tree-level (DBH, stem taper, crown dimensions and height) and plot- to stand-level structural attributes (diameter distribution, basal area, merchantable volume) from high density ULS data applicable to managed and unmanaged forest stands.

Methods developments are supported by a wide range of datasets collected at two test sites: one in the boreal deciduous forest (northwestern New-Brunswick) and the second one in the boreal coniferous forest (eastern Ontario). For both study sites, in situ field measurements were collected to serve as reference dataset to compare the results obtained with ULS data. Results from Terrestrial (TLS) and Airborne (ALS) systems are also used to assess the gain obtained with the ULS data. All data sources have now been aligned, enabling the cross comparison of results. Bastien is currently working on adapting TLS and ALS methods to ULS data namely to segment individual tree and estimate their DBH. He developed a first automatic point cloud processing chain in the open source software Computree (from Office National des Forêts (ONF) of France). In this process, trees are automatically segmented from the point cloud, stems are isolated and a shape fitting algorithm is employed to estimate DBH. The first results generated on the deciduous forest site (acquired with ULS in leaf-off condition) are promising as they match closely what can be estimated from in situ or TLS data and exceed what can be estimated with ALS (see Fig. 1). For 532 trees, the tree detection rate and the correctly segmented trees are 104 % and 78 % respectively. The next research steps involve the development of stem taper estimation and the extraction of additional tree structural attributes related to the crown and the tree form. Finally, it involves the understory characterization and the adaptation of methods for unmanaged forest types.

Figure 1: (A) Alignment of three LiDAR data sources: TLS (black), ULS acquired in leaf-off conditions (green) and ULS acquired in leaf-on conditions (red) with a zoom at the tree-level on the upper right corner; (B) Automatic tree segmentation; (C)  Sample of fitted cylinders (grey) on ULS point cloud (leaf-off) for the automatic and direct estimation of Diameter at Breast Height generated in Computree’s open source software; (D) Results for tree detection, tree segmentation and tree diameter distribution in a 1 ha deciduous stand (n= 407 trees) of McCoy Brook Forest (northwestern New-Brunswick); dark grey: in situ field inventory; light grey: ULS.

 

Presentations and Workshops

Vandendaele B., Fournier R.A., Coté J.-F., (2016) – Use of terrestrial LiDAR in forestry – Workshop – AWARE AGM Newfoundland;

Vandendaele B., (2017) – Sustaining the Hardwood Resource Value Chain – Workshop – NHRI Edmundston;

Vandendaele B., (2017) – Development of methods to estimate tree structural attributes using LiDAR on an Unmanned Aerial Vehicle – AWARE AGM Edmundston;

Vandendaele B., Goodbody T., Vepakomma U., (2017) – Use of UAVs in forestry – AWARE AGM Edmundston;

Vandendaele B., Bolton D., Rakofsky J., (2017) – Introduction to LiDAR Remote Sensing and the use of FUSION (USDA) – AWARE Cross Country Workshop –Edmundston and Corner Brook;

Vandendaele B., (2018) – Inventory and complexity of forest stands, comparing LiDAR from various platform – FPInnovations, Pointe Claire;

Vandendaele B., Abdelmounaime, (2018) – Batch processing for LiDAR data – Workshop– ULaval;

Vandendaele B., Besserer – Lemay J., (2018) – Use Computree – an open source platform for terrestrial LiDAR data processing – Workshop – UQAM

Posters

Vandendaele B., Fournier R.A., Vepakomma U., Pelletier G., Lejeune P., (2017) –Assessing the capacity of UAV-based LiDAR to support Operationnal-level Forest Inventory – CEF Montreal

Vandendaele B., Fournier R.A., Vepakomma U., Pelletier G., Lejeune P., (2017) – Assessing UAV-based LiDAR capabilities to support operationnal-level Forest Inventory – EO Summit Montreal

Vandendaele B., Fournier R.A., Vepakomma U., Pelletier G., Lejeune P., (2018) – Development of methods to estimate tree structural attributes using LiDAR on an Unmanned Aerial Vehicle – CEF Québec