Retrieving Suppressed Trees from Model-Based Height Distribution by Combining High-and Low-Density Airborne Laser Scanning Data

The individual tree detection (ITD) using airborne laser scanning (ALS) data suffers from the problem of underestimating forest total attributes, caused by deficient detectability of suppressed trees. A new method is proposed in this study to retrieve height information for suppressed trees in order to improve the estimation of forest total stem volume at plot level. Area-based approach (ABA) and ITD were, respectively, applied to predict plot-level height distributions. Automatic detection of the cut points that distinguish dominant and suppressed trees was developed. Suppressed trees in the ITD-derived height distribution were retrieved from the corresponding height classes of the ABA-derived height distribution. The 2 basic and 5 calibrated height distributions were assessed using the Reynolds error index and stem volume estimates. ITD, although exposed to systematic underestimation, obtained better accuracy for the stem volume than ABA did. After calibration, the relative root mean square error (RMSE) for the estimated stem volume decreased from 18.66% to 16.70%, and the bias decreased from 10.39% to -0.41%, at best.