3D Fuels Project

We are sampling 3D fuelbeds for regional fuel types in the southeastern (SE) and western US Explore interactive map of study sites with plot pictures (ArcGIS website).
that are most commonly burned within prescribed burning programs on U.S. Department of Defense and Department of Energy installations.

Sites by category are as follows:

1. SE longleaf pine (mesic flatwood) understories (4 sites)
   
2. SE loblolly pine-sweetgum forest understories (4 sites)
   
3. Western grasslands and grass-dominated pine savannas (4 sites)
   
4. Western ponderosa pine forest understories (4 sites)
   
5. Additional 3D Fuels sites
   

In this project, we are focused on some of the most commonly burned fuel types on DoD lands. We intentionally selected geographically distinct vegetation that has similar structures (e.g., southern pine and ponderosa pine forest understories). Through sampling of parallel fuel structures, we will evaluate how the process of 3D fuel characterization at the voxel level can be applied to structurally similar vegetation and fuels and customized to the fuel properties that may vary by species (e.g., S:V, bulk density) and fuel conditions that vary by geographic region and day-of-burn conditions (e.g., fuel moisture).

Our hierarchical sampling design incorporates lidar, photogrammetry and field-based measurements to characterize canopy and surface fuels at each study site. Recent ALS data are a selection requirement for each site. Paired TLS scans and SfM imagery (where UAS is permitted) are used to cover a 200 x 200 m area. Ground-based destructive sampling is used to calibrate the 3D point clouds with sampled occupied volume and bulk density of understory fuels.

Coarse wood and stumps are not adequately sampled by either 5x5 m plot scans or destructive plots. We plan to use synoptic TLS and SfM scans to survey coarse wood and will use a combination of measured and published bulk density values to estimate the biomass of these fuels.

Samples from all consumable fractions of live and dead fuels are collected in 3D sample plots and analyzed in the laboratory to build a library of surface fuel properties including bulk density by vegetation and fuel type. For shrub and tree species with complex architecture, foliage and branch samples are collected and individually scanned with TLS, facilitating 3D modelling of shrubs, trees and other plants as coherent geometric structures. The 3D sample plot dimensions are 0.5 m on each side in x and y, and 1-2 m in z., segmented into 10-cm vertical strata. For more details, please see: Hawley et al. (2018).

Voxel sampling of a forest understory plot at the Sycan Forest site (south central Oregon):

Voxel Survey using 3D frame

Clipping vegetation for biomass measurement

Voxel Sampling Frame

Eric Rowell and Michelle Bester conducting terrestrial lidar scanning at Lubrecht Experimental Forest, MT

Susan Prichard, Lyndsay Lascheck, Jesse Thoreson and Nick Tripodi field sampling at Tates Hell State Forest, FL

Jonathan Batchelor collecting drone-based photogrammetry scans at Tenalquot Prairie, Center for Natural Lands Management, WA

A fuel properties database and online library is being developed to house published and measured values for fuels at the object (e.g., shrub or litter layers) and element (e.g., grass blades or pine needles) scales. Values will be used to parameterize fields in 3D fuel models with computational inputs to computational fluid dynamics (CFD) models that cannot be measured remotely. We also will develop quantitative models that relate remotely sensed attributes with fuel properties.

Fuel can be distilled down to individual objects that comprise elements within a fuelbed. Individually, these objects each occupy a specific volume and often have different fuel properties, such as bulk density, varying with fuel type. In physics-based computational fluid dynamics (CFD) models such as FIRETEC (https://www.frames.gov/firetec/home) and WFDS (https://www.fs.usda.gov/pnw/projects/wildland-urban-interface-fire-dynamics-simulator-wfds), objects are often aggregated to represent total fuels and bulk density for a discrete volume (e.g. voxel grid cell).

We are using multiple approaches for object-based fuels characterization:

1. High-resolution simulations, in which individual fuel vegetation types (e.g. grasses, leaf litter, coarse woody debris) or objects will be described as three dimensional meshes that have a specific surface area and fuel mass that can be predicted as a function of unit mass per unit of surface area (g/cm²). These simulations can be assembled into mixed representative fuelbeds and distilled to estimate bulk density and mass per unit volume (g/cm³).
   
2. Quantitative structural modeling (QSM) is an analytical technique that distills point clouds with mathematical models that systematically and iteratively filter and fit various shapes to the objects present in a point cloud, thereby accounting for the incomplete nature of LiDAR scans due to occlusion and point densities.
   

3. The third method is segmentation and classification of surface and canopy fuel strata using combinations of point cloud metrics and UAS digital imagery. Point cloud objects and fuels characterization uses a new prototype algorithm to separate points into four fundamental categories and assignment of mass and volume.