Mapping and verifying potential continuous cover forestry areas from LiDAR data

Abstract

Continuous cover forestry (CCF) is a forest management approach where clear-cuts are avoided. Instead, the largest trees are removed, which leaves room for the smaller trees to mature, and enables natural regeneration in the canopy gaps. In this study, potential CCF stands were mapped from airborne laser scanning (ALS) data. This was achieved by estimating the presence and spruce-dominance (spruce vs non-spruce) of understorey trees to be grown in CCF. As a ground reference, we had tree-mapped field plots where all trees had been measured. The field-measured tree diameter and laser-scanned echo height distributions were kernel smoothed, and the derivatives of the smoothed curve were used to interpret the presence of understorey trees. Also, the proportion of laser points or trees below a defined minimum diameter limit was used as an indicator of understorey presence. This resulted in estimates of understorey from both field and ALS data, and thus an estimate on the accuracy of ALS data in the detection of understorey presence. Furthermore, area-based metrics were calculated from the ALS data separately for each recognized tree layer. These data were then used to classify the plots according to the presence and species (spruce vs non-spruce) of understorey trees. The results showed that an overall accuracy of about 0.7 was obtained both in the comparisons of smoothed distributions and in the cross-validated empirical classification to three classes. The F1-score was higher with direct comparison (0.66) compared to the empirical 3-class classification (0.56).