GI013
Time-series of lidar data for the H. J. Andrews Experimental Forest and Willamette National Forest, 2008-2021

CREATOR(S): David Bell

PRINCIPAL INVESTIGATOR(S): David Bell

ORIGINATOR(S): David Bell

DATA SET CONTACT PERSON: David Bell

METADATA CONTACT: David Bell

DATA SET CREDIT:

Original data were flown by a series of different vendors and are available from Oregon Department of Geology and Mineral Industries or USDA Forest Service Region 6

MOST RECENT METADATA REVIEW DATE:

3 Apr 2023

KEYWORDS:

stand structure, geology, geomorphology, spatial properties, geographic information systems, disturbance, landscape change, organic matter

PURPOSE:

The purpose of this work was to generate standard geospatial products from differing lidar acquisitions for researchers to use to represent state and change of the landscape topography and vegetation. The data for individual acquisitions was collected to provide a highly accurate and comprehensive base layer of elevation data and vegetation cover for differing areas of interest that overlap portions of the H. J. Andrews Experimental Forest. The purpose of the data is to provide users with a very accurate view of the topography and vegetation of the study area. The data are suitable for creating visualizations, deriving watershed boundaries, creating stream networks, identifying structures such as roads and water features, vegetation modeling and mapping, and for identifying landslides and geological features.

METHODS:

Experimental Design - GI013:

Description: For the purposes of generating a lidar time-series, we focused only on growing season (June – October) lidar acquisitions during or after 2008 that overlap portions of the H. J. Andrews Experimental Forest and LTER, excluding acquisitions that were spatially discontinuous, such as lidar strips. The extent of the area was chosen to fully encompass the Blue River Watershed and surrounding portions of the Willamette National Forest.

Field Methods - GI013:

Description: Field methods are described in lidar acquisition reports. For users, the most relevant information is likely associated with lidar scan overlap, point densities, and time of year for acquisition. All acquisitions were designed to have at least 50% sidelap in lidar strips, an average of 8.0 points per meter square, and were collected from June to October.

Processing Procedures - GI013:

Description:

Lidar data processing was accomplished in the office using the delivered point clouds for each acquisition. Point cloud data (.las or .laz files) were downloaded from Oregon Department of Geology and Mineral Industries or USDA Forest Service Region 6 servers and stored locally. For each acquisition, we applied a single processing workflow. All processing was performed in R using the raster, terra, sf, and lidR packages (Roussel et al. 2020 , Roussel and Auty 2023 ).

Point cloud data were reprojected from to NAD83 UTM zone 10N (EPSG:26910). Where needed (2014 and 2016), point heights rescaled from feet to meters. Point clouds were then saved as .laz files for further analysis.

Digital terrain models (DTMs) were generated based on vendor point classification. DTMs were generated at 1-m resolution using the triangular irregular network (TIN) algorithm. The resulting DTM was saved as a GeoTiff.

After DTM generation, we performed an automated cleaning of the point cloud. We removed all points classified as noise by the vendor. We also applied the Statistical Outlier Removal (SOR) method (Rusu 2009 ). We used an area-based approach to detect and remove outliers, where outliers were defined as points exceeding 120% of the 99th percentile of height in 10 x 10-m moving windows (described in a vignette for the lidR package). Finally, all points more than 1.0 m below or 100.0 m above the ground surface were removed.

Following point cloud cleaning, digital surface models (DSMs) and canopy height models (CHMs) were generated at 1-m resolution. For DSMs, we used the p2r algorithm using a sub-circle of 0.15 to approximate the diameter of the laser, with TIN for filling pixels with no data. CHMs were created by differencing DSMs and DTMs. Points used for DTM generation included never-classified or unclassified points (0, 1), ground (2), vegetation (3-5), and water (9) (ASPRS 2019 ).

To generate topographic variables, we used the DTM and the terra package in R. Aspect and slope were calculated at 1-m resolution using the terrain function, with units equal to degrees. To provide an easily visualized representation of ground and canopy surface, we generated hillshade and corwnshade at 1-m resolution using the shade function with slope and aspect in radians. Slope, aspect, hillshade, and crownshade were saved as GeoTiffs. We also aggregated the DTM to 10-m using a mean function to create a 10-m digital elevation model as the basis for 10-m, 25-m, and 50-m contour lines based on the 10-m, saved as Esri Shapefiles.

To generate vegetation variables, we generated rasters at 5-m and 25-m resolutions summarizing vertical and horizontal variation in height above ground. Two general types of vegetation variables, or metrics, were generated: point-based and pixel-based.

Point-based metrics are those that were based directly on the lidar point cloud data. When metrics were based on point clouds, we retained never-classified or unclassified points (0, 1), ground (2), vegetation (3-5), and water (9) (ASPRS 2019 ). Points were normalized based on the DTM to generated height relative to ground surface. We then generate standard lidar metrics based on height, including number, max, mean, standard deviation, kurtosis, skew, entropy, percentage of returns above the mean, percentage of returns above a given height, and the height of the xth percentile for points. These were generated for all returns, first returns, and first returns at least 2 m above the ground surface.

Citation:

Roussel J, Auty D, Coops NC, Tompalski P, Goodbody TR, Meador AS, Bourdon J, de Boissieu F, Achim A (2020). “lidR: An R package for analysis of Airborne Laser Scanning (ALS) data.” Remote Sensing of Environment, 251, 112061. ISSN 0034-4257, doi:10.1016/j.rse.2020.112061, https://www.sciencedirect.com/science/article/pii/S0034425720304314.

Roussel J, Auty D, Coops NC, Tompalski P, Goodbody TR, Meador AS, Bourdon J, de Boissieu F, Achim A (2020). “lidR: An R package for analysis of Airborne Laser Scanning (ALS) data.” Remote Sensing of Environment, 251, 112061. ISSN 0034-4257, doi:10.1016/j.rse.2020.112061, https://www.sciencedirect.com/science/article/pii/S0034425720304314.

Rusu, R.B. (2009). Semantic 3D object maps for every-day manipulation in human living environments. PhD thesis, University of Munich

LAS Specification 1.4-R14. The American Society for Photogrammetry & Remote Sensing. Available online: http://www.asprs.org/wp-content/uploads/2019/03/LAS_1_4_r14.pdf

LAS Specification 1.4-R14. The American Society for Photogrammetry & Remote Sensing. Available online: http://www.asprs.org/wp-content/uploads/2019/03/LAS_1_4_r14.pdf

Quality Assurance - GI013:

Description: Raster products were visually assessed. Detailed alterations of rasters, as is often performed by vendors, was not performed here. Users should inspect layers and implement methods to minimize the effects of, for example, areas with low return densities.

SITE DESCRIPTION:

Lane and Linn Counties, Oregon, USA

TAXONOMIC SYSTEM:

None

GEOGRAPHIC EXTENT:

H. J. Andrews Experimental Forest and surrounding landscape, Western Cascades, Oregon, USA

MEASUREMENT FREQUENCY:

Irregular (1-6 years)

PROGRESS DESCRIPTION:

Active

UPDATE FREQUENCY DESCRIPTION:

asNeeded

CURRENTNESS REFERENCE:

Observed

RELATED MATERIAL:

Acquisition reports, various additional geospatial layers, and original data are available from Oregon Department of Geology and Mineral Industries or USDA Forest Service Region 6. We also provide links to these reports at {TBD}