This Is AuburnElectronic Theses and Dissertations

Using Airborne and Spaceborne lidar for Mapping Forest Attributes in Southern Alabama.




Brown, Schyler

Type of Degree

Master's Thesis


Forestry and Wildlife Science

Restriction Status


Restriction Type

Auburn University Users

Date Available



Remote sensing has become a major resource for assessing ecosystems and global carbon cycling. In forest applications, remote sensing data have been used to estimate the amount of carbon stored or lost in forests due to fire, deforestation, and reforestation. In particular, airborne light detection and ranging (lidar) has shown practical effectiveness in estimating forest variables over greater spatial extents. Using models developed from an association of field inventory samples with lidar data, attributes such as basal area, volume, and aboveground biomass can be predicted over larger tracts of forests, allowing for wall-to-wall mapping. Spaceborne lidar can also collect elevation data at broader scales and may be used to efficiently characterize forest structure over an even larger area of land. This work consisted of two main objectives: (1) to develop wall-to-wall maps of basal area, volume and aboveground biomass using airborne lidar for the Solon Dixon Forestry Education Center in southern Alabama, and (2) to investigate the Global Ecosystem Dynamics Investigation (GEDI) lidar for predicting forest structure over a larger spatial extent. The reference maps from chapter 2 had modest accuracy (R2 = 0.36-0.53), with much of the unexplained model variation due to field data lacking complete vegetation. In the second study, models were developed from GEDI lidar with lower accuracy (R2 = 0.30-0.36) and the maps produced from GEDI footprints and auxiliary imagery had low accuracy (R2 = 0.10-0.11). The low accuracy was likely caused by a propagation of errors in the reference maps due to incomplete field inventory and could be further investigated. While lidar data are sufficient for forest mapping, texture variables derived from multispectral imagery are valuable, especially where lidar data is not spatially continuous.