Abstract
The objective of this study was to validate the use of stem measurement data from cut-to-length (CTL) harvesters and tree height and density data from airborne laser scanner (ALS) to predict product recovery. Comparisons were made between three ALS-based inventory methods: (1) LaserSTM (ALS combined with CTL-harvester measurements); (2) LaserDBH (ALS combined with field-plot measurements); and (3) LaserAVG (ALS stand-average estimates). The prognosis of product recovery was based on bucking simulations. Forest data from a Swedish test site were used in the study and included 17 forest stands. All stands were evaluated at the sub-stand level, with seven of the stands also being evaluated at the stand level. The focus of the validation was on the geometry of the trees. Tree species and stem defects were not considered in the prediction of product recovery. The diameter distributions of logs were evaluated using the Reynolds error index. Using 2 cm classes, the error indices were 0.16, 0.22, and 0.21 for LaserSTM, LaserDBH, and LaserAVG respectively. With 1 cm classes, these figures were 0.21, 0.28, and 0.39. In the Swedish case, where stem taper functions are available, the stem taper information from the harvester seems to have a limited contribution to the accuracy of volume estimates for different product groups. However, the estimates of log diameter distributions are being improved based on more detailed description of the trees. The results shows that stem measurements from CTL harvesters can replace labor-intensive field-plot measurements as training data for pre-harvest inventory models based on ALS data.
Acknowledgement
This research was financed by the Swedish Research Council Formas, as part of the WW-IRIS project within the ERA-Net Wood Wisdom project, and by the Heureka research program. We thank those people who have been involved in the harvest operation at Strömsjöliden: Alexandra Frank and her colleges at Sveaskog in Västerbotten for planning and taking care of the harvest operations; entrepreneur and machine operator Lars Anders Gustafsson at LA Skog in Örträsk for cutting down the trees and manually registering our sample trees in his onboard computer; Lena Jonsson at the Unit for Field-Based Forest Research at SLU for hosting and supporting our project; and Torbjörn Cruse at Cruse Datakonsult for developing a GIS tool for registering trees in the harvester. The authors also thank Kenneth Olofsson at SLU, John Arlinger, Nazmul Bhuiyan, Björn Hannrup, Johan J Möller, and Maria Nordström and Lars Wilhelmsson of Skogforsk for valuable guidance, comments, and discussions during the study.