A Segmentation Approach for Estimating Forest Structural Characteristics from Lidar and Radar: Analysis and Error Assessment

Paul Robert Siqueira, University of Massachusetts, siqueira@ecs.umass.edu (Presenter)
Razi Ahmed, University of Massachusetts, rahmed@ecs.umass.edu
Caitlin Dickinson, University of Massachusetts, dickinson@ecs.umass.edu
Bruce Chapman, JPL, bruce.d.chapman@nasa.gov
Scott Hensley, JPL, scott.hensley@jpl.nasa.gov
Kathleen Bergen, University of Michigan, kbergen@umich.edu

To help meet the needs of the planned DESDynI mission, our group has been investigating the use of segmentation algorithms to make best use of the mapping capabilities of synthetic aperture radar and the fundamental structural sensitivity of lidar, to make large-scale and regional estimates of forest biomass and vegetation structure. In order to achieve this goal, we have studied three sites in detail: 1.) The Harvard Forest, 2.) The Howland Forest , and 3.) The Injune region in Eastern Australia. The availability of ground validation resources for these sites is extensive, with airborne and spaceborne SAR and lidar available. The sites have been singled out for investigation because of the diversity of forest structure and because the terrain is relatively flat. Flat terrain is an important characteristic fo establishing a fundamental sensitivity between the remote sensing observables and the forest physical characteristics of interest.

Segmentation has been used for this work in order to aggregate regions of like-response to the SAR observations, and then to relate the regions to structural observations made by lidar. By using NASA’s UAVSAR and LVIS high resolution assets for prototyping, it has been possible to separate the effects associated with inherent instrumental error sources (e.g. radar speckle) from natural variations in the target characteristics. Using the extensive observations that have been collected by the Japanese Aerospace Exploration Agency’s ALOS/PALSAR, it has been possible to extrapolate how these results could be applied to a spaceborne instrument such as DESDynI. Further, by examining the different forest types at Injune, Howland and Harvard Forests, our team has been able to investigate in which structural and biomass regimes such a method could be successfully applied, and which not.

Finally, our group has also performed a formal (and extensive) error analysis in the estimation of forest biomass through the integration of ground validation measurements and remote sensing observations. Results from this work will be presented as well.

Presentation Type:  Poster

Session:  Global Change Impact & Vulnerability   (Tue 11:30 AM)

Associated Project(s): 

• Siqueira, Paul: A Segmentation Approach for Combining RaDAR Backscater, InSAR and LiDAR Measurements to Determine Vegetation 3D Structure and Biomass from Space ...details
• Siqueira, Paul: The impact of temporal decorrelation on InSAR vegetation 3-D structure retrieval algorithms ...details

Poster Location ID: 285