Quantification of uncertainty in aboveground biomass estimates derived from small-footprint LiDAR data

Abstract:

A promising approach to determining aboveground biomass (AGB) in forests comes through the use of individual tree crown delineation (ITCD) techniques applied to small-footprint LiDAR data. These techniques, when combined with allometric equations, can produce per-tree estimates of AGB. At this scale, AGB estimates can be quantified in a manner similar to how ground-based forest inventories are produced. However, these approaches have significant uncertainties that are rarely described in full. Allometric equations are often based on species-specific diameter-at-breast height (DBH) relationships, but neither DBH nor species can be reliably determined using remote sensing analysis. Furthermore, many approaches to ITCD only delineate trees appearing in the upper canopy so subcanopy trees are often missing from the inventories. In this research, we performed a propagation-of-error analysis to determine the spatially varying uncertainties in AGB estimates at the individual plant and stand level for a large collection of LiDAR acquisitions covering a large portion of California. Furthermore, we determined the relative contribution of various aspects of the analysis towards the uncertainty, including errors in the ITCD results, the allometric equations, the taxonomic designation, and the local biophysical environment. Watershed segmentation was used to obtain the preliminary crown segments. Lidar points within the preliminary segments were extracted to form profiling data of the segments, and then mode detection algorithms were applied to identify the tree number and tree heights within each segment. As part of this analysis, we derived novel “remote sensing aware” allometric equations and their uncertainties based on three-dimensional morphological metrics that can be accurately derived from LiDAR data.