Large-area forest productivity estimation using bitemporal data from airborne laser scanning and digital aerial photogrammetry

Site index (SI) indicates the magnitude of timber production that can be realized at a given site and is a crucial variable in forest planning. In Norwegian forest management inventories, SI is commonly quanti ed with large uncertainty by means of aerial image interpretation, eld assessment and information from previous inventories. Airborne laser scanning (ALS) and digital aerial photogrammetry (DAP) have revolutionized the eld of forest inventory in recent decades, however operational practices for SI estimation have remained unchanged since the 1970s. The main objective of this thesis was to develop practical methods of SI estimation using bitemporal tree height data derived from ALS and DAP. The first study presented two practical methods of SI estimation; (i) the direct method, in which models are applied for direct prediction of SI from bitemporal ALS metrics, and (ii) the indirect method in which the SI is derived indirectly from estimates of canopy height development over time. Both methods provided reliable SI estimates, however the direct method was most accurate. Operational application of the methods requires undisturbed forest growth. Hence, the second study assessed the use of bitemporal ALS data for classification of various types of changes in forest structure, and showed that such changes can be classi ed with high accuracy at plot level. In the third study, a practical method for predicting and mapping SI in repeated ALS-based forest inventories was demonstrated. The method included a forest disturbance classi cation, and the direct method was then applied to forest areas classi ed as undisturbed. The last study compared the economic utility of six methods of ALS- and DAP-based SI estimation and conventional practices in a cost-plus-loss analysis, by which the economic losses due to sub-optimal treatment decisions were added to the inventory costs. The study showed that SI can be estimated from bitemporal combinations of ALS and DAP data with unprecedented accuracy and at a lower cost than conventional methods. This thesis shows that bitemporal ALS and DAP data are highly suitable for the estimation of SI. The methods presented here can be used to predict, estimate and map SI at sub-stand level automatically over large areas of forest. They are practically applicable and cost-effcient, and can be adopted to replace conventional practices of SI estimation in repeated forest management inventories