Sagebrush Measurements Using Small-Footprint Discrete-Return LiDAR

Sagebrush Canopy Height and Shape Measurements Using Small-Footprint Discrete-Return LiDAR


Canopy height and shape are critical measurements in sagebrush steppe ecosystems for vegetation structure and cover characterization as well as for estimations related to biomass, site productivity, soil erosion, fuel loading and wildlife habitat assessment (e.g., sage-grouse). While field measurements of sagebrush height and shape are reliable, the methods are difficult to apply across large areas.  LiDAR (airborne laser scanning) has the potential for estimating canopy height and shape at a range of scales appropriate for landscape assessments.  Separating discrete LiDAR returns in low-height rangeland vegetation is difficult because the vegetation canopy return is often close to the ground return in both space and time.  Furthermore, there are fewer vegetation returns in sparsely vegetated semiarid ecosystems than in more foliated ecosystems.  A large number of studies have found that airborne LiDAR systems tend to underestimate canopy.  Either the top of the canopy is missed, or laser pulses penetrate the canopy and the return signal is from material within the canopy.  Factors that contribute to LiDAR prediction errors for shrub height and shape include target characteristics (e.g., slope, surface roughness, canopy density), sensor characteristics (e.g., scan angle, footprint diameter, and LiDAR return density) and associated with integrating ground reference and LiDAR return data (e.g., field measurement techniques, bare earth modeling methods).



  • Quantify shrub height and shape prediction errors by comparing LiDAR point-cloud data to sagebrush canopy characteristics measured in the field

  • Evaluate differences in LiDAR height and shape estimation results when using raw point-cloud data versus using maximum vegetation height models derived from the data (at 0.5 m and 1.0 m grid resolutions)

  • Evaluate sources of LiDAR underestimation error associated with target characteristics.


Accomplishments through 2009

From May to October 2009, height and shrub measurements were collected for 107 individual sagebrush and bare earth elevations were collected throughout 11 circular plots 20 m in diameter (figure below).  The sagebrush ground reference data were used to evaluate high density (average density of 9.46 points m-2), small-footprint discrete return LiDAR data collected over portions of the INL on December 13, 2006.  We transformed raw LiDAR point-cloud into raster data products (i.e., 0.5 m and 1 m grids of maximum vegetation height, bare earth elevation, total point density, vegetation roughness, and local roughness) using a LiDAR tools extension developed for rangeland vegetation (  LiDAR height and shape underestimation error was quantified by regressing LiDAR predictions against ground measurements.  Preliminary results were presented at Silvilaser 2009: The 9th International Conference on LiDAR Applications for Assessing Forest Ecosystems in Houston, Texas (October 14-16th) and at the American Geophysical Union Fall Meeting in San Francisco, California (December 14-18th). 

Demography Plot Schematic.wmf

Plot sampling schematic (not to scale) for the Sagebrush Demography Project.  The drawing shows the types and locations of vegetation data collected at each plot. 



Our results demonstrated that the LiDAR-derived sagebrush height estimates were significantly and strongly correlated with corresponding field-based height estimates, with observed coefficients of determination of 0.84-0.86.  Similarly, LiDAR predictions of shrub shape and area were significantly and strongly correlated with field-based measurements resulting in coefficients of determination of 0.65-0.78.  Regression analysis for both shrub height and shape indicated a tendency toward LiDAR underestimation.  Although shrub area was underestimated, our shrub shape and area measurements provide the first LiDAR-based low-height vegetation shape descriptors.  It is likely that the primary source of shape estimation error is related to the task of properly delineating individual shrubs in the LiDAR data.  While prediction error could account for as much as 35.6 percent of the average height and 32.8 percent of the average canopy area of shrubs sampled, alternative estimation methods (e.g., Interferometric Synthetic Aperture Radar (InSAR), stereoscopy) are likely to yield much less accurate results. 


Plans for Continuation

Plans for continuation of this study include evaluating LiDAR-derived vegetation roughness estimates as an indicator of shrub height and investigating the potential for future research on shrub biomass estimation through the fusion of LiDAR data with hyperspectral imagery acquired from an unmanned aerial vehicle platform.  The results of this study are being submitted for publication in the journal Photogrammetric Engineering and Remote Sensing (April 2010) and will be included as a chapter in the dissertation of the student investigator (2010).  Our collaboration with INL through the LDRD will continue in fiscal year 2010.


Publications, Theses and Reports

Mitchell, J. J., Glenn, N. F., Sankey, T. and Derryberry, D. (2009a), LiDAR canopy height measurements in a sagebrush steppe ecosystem, Silvilaser 2009, the 9th International Conference on LiDAR Applications for Assessing Forest Ecosystems, October 2009, College Station, Texas.

Mitchell, J. J., Glenn, N. F., Sankey, T. and Derryberry, D. (2009b), LiDAR Canopy Height and Shape measurements in a sagebrush steppe ecosystem, Eos Trans. AGU, 90(52), Fall Meet. Suppl., Abstract B31A-0320.

LDRD FY09 Final Report