A comparison of four individual tree height prediction methods for forest inventory

In an effort to same time and money, land managers typically subsample certain tree attributes during forest inventory. The calculation of unsampled attributes is then achieved through statistical models exploiting strong relationships between unsampled and sampled parameters. The development of new remote sensing technologies may augment such approaches or provide an alternative method for determining such unsampled tree attributes. This paper compares four separate methods of inferring unsampled tree heights in open canopy stands in North Idaho, USA. The results suggest that imputing missing tree heights via mixed effects modeling out-performs lidar-based estimates of missing tree heights in open forest stands. However, the difference between the mixed effects model estimates and estimates extracted from lidar data is negligible, suggesting that in open forested environments, lidar may provide accurate information which can employed to supplement forest inventory.