Canadian Forest Service Publications

Abstract tree crowns in 3D radiative transfer models: impact on simulated open-canopy reflectances. 2014. Widlowski, J.-L.; Côté, J.-F.; Béland, M. Remote Sensing of the Environment 142: 155-175.

Year: 2014

Issued by: Atlantic Forestry Centre

Catalog ID: 35317

Language: English

Availability: PDF (request by e-mail)

Available from the Journal's Web site.
DOI: 10.1016/j.rse.2013.11.016

† This site may require a fee

Mark record

Abstract

Three-dimensional (3D) radiatives transfer models of vegetation canopies are increasingly used to study the reflective properties of specific land cover types and to interpret satellite-based remote sensing observations of such environments. In doing so, most 3D canopy reflectance models simplify the structural representation of individual tree crowns, for example, by using a single ellipsoidal envelope or a series of cubic volumes (known as voxels) to approximate the actual crown shape and the 3D distribution of scatterers therein. Often these tree abstractions ignore or simplify the woody architecture as well. Focusing on broad-leaved Savanna trees, this study investigates the impact that architectural simplifications may have on the fidelity of simulated satellite observations at the bottom of the atmosphere for a variety of spatial resolutions, spectral bands, as well as viewing and illumination geometries. The typical uncertainty associated with vicarious calibration efforts, i.e., 5%, is used as the quality objective for the simulated bidirectional reflectance factors (BRFs). Our results indicate that the size of the voxel as well as the spectral, viewing, and illumination conditions drive the BRF bias at a given spatial resolution. The simulation of remote sensing data at medium spatial resolution is not affected by canopy abstractions except in the near-infrared (NIR) for cases where woody structures are omitted. Here, the BRF simulations of the abstract tree crowns exceeded the 5% tolerance limit even at spatial resolutions coarser than 125 m. For high-resolution satellite imagery, i.e., for nominal pixel sizes of 1 × 1 m2 or finer, local BRF biases can be 10 times greater than the 5% tolerance criterion. Both positive and negative local biases are possible depending on the relative weights of the single-collided, single-uncollided, and multiple-collided BRF components.

Plain Language Summary

The use of satellite imagery for forest inventory purposes began in the 1970s. The techniques used have been refined over the years, and the results are becoming more and more precise.

To further improve the effectiveness of these techniques, researchers created models to simulate what the satellite sees in order to better understand the interactions between tree structure (trunk, branch, and foliage) and what is measured by satellite. How precisely these models can simulate what the satellite sees depends on how realistically the tree structure is defined. To this end, the researchers used highly detailed 3D models of trees found in savanna environments. They varied the definition and level of detail of tree structure attributes in order to note differences from satellite measurements. The researchers found that the more realistic this structure is, the greater the similarity between the results and the satellite data.

These results will help reduce the costs of using satellite images by reducing the number of field validation sites and the number of measurements to be performed therein.