Dataset
ForestScan Project : Terrestrial Laser Scanning (TLS) of FBRMS-01: Paracou, French Guiana 1ha plot FG5c1, September to October 2022
Abstract
Terrestrial laser scanning (TLS) was conducted at three ForestScan 1ha (100m x 100m) Forest Biomass Reference Measurement Site (FBRMS) plots in French Guiana from September to October 2022 by Cecilia Chavana-Bryant using a Riegl VZ-400i scanner. Data collection assistance was provided by UCL PhD student Wanxin Yang and a local team of field assistants, data processing assistance was provided by Mr Peter Vines. This data collection was part of the European Space Agency (ESA) funded ForestScan project designed to improve the use of new Earth Observation (EO) estimates of above ground biomass (AGB) by providing terrestial (TLS), unpiloted airborne vehicles (UAV-LS)- and airborne (ALS) LiDAR scanning-derived AGB and tree census data to compare to allometric and EO-derived estimates.
Scans were acquired using chain sampling at 121 locations along a 10m Cartesian grid to ensure sufficient data overlap to produce high-quality point clouds for all ForestScan 1ha FBRMS plots. Due to the scanner's 100° field of view, capturing a complete sample of the scene at each scan location required two scans -an upright scan and a tilt scan. Upright scans are odd-numbered while tilt scans are even-numbered. The first scan at each plot is collected at the southwest corner, i.e. scan position 0,0 (unless something impedes it, e.g. stream, large tree fall, etc. or if the plot is oriented differently). To facilitate scan registration, five retro-reflective targets were located between scan positions with all tilt scans along the first sampling line were oriented towards the same sampling position along the next sampling line and tilt scans at the ends of sampling lines (i.e. tilt scans along plot edges) were oriented towards the inside of the plot. This aids scan registration as it allows tilt scans to capture the previous scan location within its field of view. A total of 242 scans were collected at each plot.
The Riegl operating and processing software RiSCAN PRO version 2.14.1 was used to generate a plot-level point cloud, scans were coarse registered using the shared retro-reflective targets located between consecutive scan positions. Coarse registration was then fine-tuned using Multi Station Adjustment 2 (MSA2).
Data for each of the three FBRMS plots is found within plot directories: FG5c1, FG6c2 and FG8c4. Plot directories contain a main project directory (named using the starting date of data collection, e.g. 2022-10-10_FG5c1.PROJ) with nine data subdirectories and a tile_index.dat file as shown in the archived document /neodc/forestscan/data/french_guiana/paracou/TLS_Plot_FG5c1/ForestScan_example_data_directory_structure.pdf which details the data structure shared by all FBRSM plot TLS datasets.
Details
| Previous Info: |
No news update for this record
|
|---|---|
| Previously used record identifiers: |
No related previous identifiers.
|
| Access rules: |
Public data: access to these data is available to both registered and non-registered users.
Use of these data is covered by the following licence(s): http://creativecommons.org/licenses/by/4.0/ When using these data you must cite them correctly using the citation given on the CEDA Data Catalogue record. |
| Data lineage: |
Data were produced by the project team and supplied for archiving at the Centre for Environmental Data Analysis (CEDA). |
| Data Quality: |
Data quality control conducted by UCL ForestScan project team
|
| File Format: |
Terrestrial LiDAR scanner data in csv, point cloud and Riegl Proprietary raw data format. Details of formats and data structure can be found in the archive /neodc/forestscan/data/french_guiana/paracou/TLS_Plot_FG5c1/ForestScan_example_data_directory_structure.pdf which details the data structure shared by all FBRSM plot TLS datasets. Further explanation of the files and their origin can also be found in the process computation section.
|
Process overview
Independent Instruments
| UCL RIEGL VZ-400 Terrestrial LiDAR scanner |
Computation Element: 1
| Title | TLS2trees processing pipeline for ForestScan 1ha plots FG5c1, FG6c2 and FG8c4 in FBRMS-01: Paracou, French Guiana |
| Abstract | Data for each of the four Gabon FBRMS plots is found within plot directories: OKO-01; OKO-02; OKO-03 and LPG-01. Plot directories contain a main project directory (named using the starting date of data collection and the plot ID, e.g. 2022-06-24_LPG-01_PROJ) with nine data sub-directories, a tile_index.dat file and a 2022-06-24_LPG-01.kmz file as shown in the archived ForestScan_example_data_directory_structure.pdf document. The raw project sub-directory contains all registered scans for each FBRMS 1ha plot. The matrix project sub-directory contains each scan’s Sensor's Orientation and Position (SOP) matrix with the GNSS coordinates (geographical coordinate system: WGS84 Cartesian) for all scans saved separately and made available via a .kmz file under the project main directory, e.g. 2022-06-24_LPG-01.kmz. In order to estimate woody volume and above ground biomass (AGB) for each plot, the TLS2trees processing pipeline was used. TLS2trees is an automated processing pipeline and set of Python command line tools that segments individual trees from plot level point clouds. It consists of existing and new methods and is specifically designed to be horizontally scalable. The TLS2trees pipeline includes three preparatory data steps followed by two segmentation steps: semantic & instance segmentation. Quantitative Structure Modelling (QSM) is then used to estimate morphological and topological tree traits via a four-step process: generate TreeQSM inputs, run TreeQSM, generate optQSM commands and run optQSM. Two final processing steps generated 1) a tree attributes .csv file and 2) tree figures of individually segmented trees arranged by tree DBH size. The complete set of TLS2trees processing files is available for each of the four ForestScan FBRMS plots in Gabon, the step-by-step processing summary below provides details for these files. The first of three preparatory data steps segmented the 100m x 100m plot point clouds into 10m x 10m data tiles and converted each tile from the RIEGL proprietary file format .rxp to .ply format. The resulting <0-NNN>.ply files (NNN is the assigned tile ID number) + a subdirectory named bounding_box containing bounding geometry files + a tile_index.dat file were saved into the rxp2ply project subdirectory. The second preparatory data step down-sampled the data tiles with results saved as tileID.downsample.ply files in the downsample project subdirectory, e.g. 000.downsample.ply. The third preparatory data step generated a tile_index.dat file saved under the project directory. Next, a semantic segmentation step classified the tiled data into leaf, wood, ground or coarse woody debris. For each data tile, three different files tileID.downsample.dem.csv, tileID.downsample.params.pickle, tileID.downsample.segmented.ply + a temporary subdirectory tileID.downsample.tmp were generated and saved in the fsct project subdirectory. Instance segmentation was then used to automatically segment the semantically classified tiled data into individual tree files. Two automatically segmented versions of each tree (with and without canopy leaves) were generated and saved in subdirectories arranged by increasing DBH size (i.e. subdirectory 0.0 contains the smallest trees in the plot) under the clouds project subdirectory, e.g. clouds/N.N/tileID_TreeID.leafon.ply and clouds/N.N/tileID_TreeID.leafoff.ply. Quantitative Structure Modelling (QSM) was then used to enclose the wood-only file version (i.e. tileID_TreeID.leafoff.ply) of each individually segmented tree in a set of geometric primitives i.e. cylinders. This allowed for the estimation of morphological and topological traits such as volume, length and surface area metrics for each successfully modelled tree. The first QSM processing step generated 125 modelling input files representing 125 different parameter combinations for each individually segmented tree. These files were saved as tileID_TreeID_NNN.m (NNN ranges from 0 to 124) in the models/intermediate/inputs project subdirectory, e.g. models/intermediate/inputs/tileID_TreeID/tileID_TreeID_<0-124>.m. Next, up to 625 different model candidates for each segmented tree were generated from the modelling input files and saved as tileID_TreeID-NNN.mat files (NNN ranges from 0 to 624) in the models/intermediate/results project subdirectory, e.g. models/intermediate/results/tileID_TreeID/tileID_TreeID-NNN.mat. QSM command files to find the optimal QSM for each segmented tree were then generated and saved as tileID_TreeID_opt.m files in the models/optqsm/commands project subdirectory, e.g. models/optqsm/commands/tileID_TreeID_opt.m. During the final QSM step, an optimal model was found for each successfully modelled segmented tree and saved as a tileID_TreeID.mat file in the models/optqsm/results project subdirectory, e.g. models/optqsm/results/tileID_TreeID.mat. After QSM modelling, a report file named projectID.tree-attributes.csv was generated for each plot and saved in the attributes project sub-directory, e.g. attributes/projectID.tree-attributes.csv. This report contains estimates of morphological and topological traits for all modelled trees. Due to the >300m scanning range of the Riegl VZ-400i scanner, reports contain trees located both inside and outside the plots which can be filtered using the in_plot variable. Each row in these reports represents a tree with both successfully and unsuccessfully (empty attribute variables) modelled trees included in the reports. The last processing step generated tree figures arranged by descending tree DBH size and saved as projectID.nn.png files (nn refers to the order in which the figures were generated with figure projectID.0.png containing the largest trees) in the figures project sub-directory, e.g. figures/lpg_01.0.png. |
| Input Description | None |
| Output Description | None |
| Software Reference | None |
| Output Description | None |
No variables found.
Temporal Range
2022-10-10T00:00:00
2022-10-17T00:00:00
Geographic Extent
5.2726° |
||
-52.9297° |
-52.9291° |
|
5.2716° |