Using LIDAR as a Sampling Tool for Forest Inventory
on the Kenai Peninsula of Alaska

Airborne laser scanning (lidar) can provide very detailed, remotely-sensed measurements of forest structure, and therefore has the potential to significantly improve the efficiency of forest inventory and monitoring. The use of a multi-level inventory sampling design, consisting of information derived from 1) a relatively sparse grid of field plots, 2) continuous strips of high-resolution airborne lidar, 3) wall-to-wall satellite image-based classifications, could be particularly advantageous in extensive, remote regions such as interior Alaska, where it is logistically difficult and extremely costly to establish field inventory plots. In this study, we used detailed individual tree-level forest measurements (crown cover, overstory tree height, species type) obtained from ten continuous north-south (350-m wide) strips of high-density lidar data collected in 2004 over 120 FIA field plots on the western lowlands of the Kenai Peninsula of Alaska, in combination with a lower-resolution (30-m) comprehensive forest type classification (height, cover, vegetation class) provided through the LANDFIRE program, and allometric biomass models to develop and evaluate a model-based, multi-level inventory sampling design for estimation of total aboveground biomass (by forest type) over this entire 7800 sq. km. region. A simulation-based approach was used to evaluate the statistical properties (variance and bias) of the multi-level biomass estimator. Using a comprehensive, detailed vegetation classifi cation developed using aerial photo-interpretation and the FIA plot data as a basis of comparison, the effect of varying sampling intensities for lidar data strips and the effect of modeling error on biomass estimates, as well as potential for application in interior Alaska, was also investigated.