16 December 2013 Science Briefs

SNAMP Pub #18: Tradeoffs between lidar pulse density and forest measurement accuracy

Article Title: Tradeoffs between lidar pulse density and forest measurement accuracy.

Authors: Marek K Jakubowski, Qinghua Guo, Maggi Kelly.

Research Highlights:


  • Collection of lidar (light detection and ranging) data can be costly, and costs depend on the density of the resulting data (pulses or “hits” per m2).

  • Most lidar acquisitions capture the highest possible density of data (up to 12 pulses/m2); but do we need that density of data? The benefit of collecting less dense data might be that data would be able to be captured over a larger area for the same cost.

  • We investigated the ability of different densities of lidar data to predict forest metrics at the plot scale (e.g. 1/5-hectare or ½-acre).

  • We examined ten canopy metrics (maximum and mean tree height, total basal area, tree density, mean height to live crown base (HTLCB), canopy cover, maximum and mean diameter at breast height (DBH), and shrub cover and height) based on varying pulse density of lidar data – from low density (0.01pulses/m2) to high density (10 pulses/m2).

  • High-density lidar data yielded higher accuracy predictions than low-density lidar data; but low-density data was also valuable at predicting canopy metrics at the plot-scale.

  • Many plot-level forest measures could be accurately predicted with lidar at moderate densities: 1 pulse/m2.

Background:
Airborne lidar (light detection and ranging) data is increasingly collected over large areas (such as counties, states, and even countries) to build topographic models and to provide information about ecosystems. For many reasons, high-density lidar, where many “returns per square-meter” are collected, is more expensive to acquire and process than low-density data, given the same project area. However, by lowering the density of lidar data we run the risk of not capturing enough information to accurately characterize our target. Project managers frequently face this tradeoff between cost, data density, and area surveyed. We wanted to explore this tradeoff, and determine which densities might be best for which types of targets, and to determine if lower density, and less costly, lidar data could accurately measure plot-scale forest measurements.

In this paper, we derive ten forest canopy metrics (maximum and average tree heights, total basal area, average height to live crown base, maximum and average diameter and breast height (DBH), tree density, canopy cover, and shrub cover and height) at the scale of a typical forest plot (e.g a plot with a radius of 12.5 m or 41 ft) based on continually decreasing lidar data density. We simulated lidar densities between low (0.01 pulses/m2) and high (9 pulses/m2). We compared the accuracy of each forest metric derived through lidar against actual tree measurements taken by field crews.

Results
The accuracy of the lidar predictions for all ten metrics increased as the lidar density increased from 0.01 pulses/m2 to 1 pulse/m2. However, the accuracy of many of the metrics showed very little improvement after that. Metrics that described forest cover (e.g. forest canopy and shrub cover) required higher densities of lidar data to be mapped accurately. In general, the results confirm findings from previous studies: the overall accuracy of a predicted forest structure metric decreased roughly with its vertical position within the canopy: metrics that estimate the tops of forests are more accurately mapped with lidar than those in the middle of the canopy or on the forest floor and so require less dense data for most applications.

Conclusions:


  1. Many plot-scale forest canopy measures (e.g. maximum and mean tree height, total basal area, maximum and mean diameter at breast height (DBH)) are well predicted with moderate density lidar data: 1 pulse/m2. More detailed features, such as individual trees, would likely require high-density lidar data.

  2. Coverage metrics (canopy cover, tree density, and shrub cover) were more sensitive to pulse density.

  3. This study should help managers evaluate tradeoffs between lidar density and coverage: if a manager needs plot-scale forest measurements (i.e. measurements summarized at scale around ½-acre or 1/5-hectare), they might be able to cover a larger area with lower density lidar data for the same cost as high density lidar data over a smaller area.

Full Reference:
Jakubowski, Marek K., Q. Guo, M. Kelly. 2012. “Tradeoffs between lidar pulse density and forest measurement accuracy,” in Remote Sensing of Environment, 130: 245-253.

The full paper is available here.

For more information about the SNAMP project and the Spatial team, please see the: Spatial Team Website.

To learn more about lidar data, check out our lidar FAQs sheet. For more information on lidar data in SNAMP, see the spatial team website: http://snamp.cnr.berkeley.edu/teams/spatial, and our spatial team newsletters that focus on lidar: Vol. 2, No. 3, and Vol. 5, No. 1.

No Comments Yet

Post a New Comment