This dissertation evaluates the applicability of airborne laser scanning (ALS)-based forest attribute interpretation, mobile-based machine vision methods and operational harvester data for improving forest inventories and management. The research focuses on assessing the accuracy of remotely-sensed forest attributes and mobile machine vision-derived volume attributes against operational harvester data, and improving remote sensing-based volume attribute estimates by applying harvester measurements and other Big Geodata.
My first study examined the accuracy of Metsään.fi forest inventory data, derived from ALS, by comparing it to operational harvester data. The findings revealed a tendency to overestimate sawlog removals, particularly Norway spruce (Picea abies (L.) Karst.) in clear-cut areas, although dominant tree species were accurately determined.
My second study assessed the Trestima smartphone app for pre-harvest measurements and my results showed that an insufficient number of photographs per forest stand led to poor accuracy levels, although when the recommended data collection protocol was closely followed there was an improvement in performance. The app provided accurate estimates of Norway spruce volume but slightly underestimated Scots pine (Pinus sylvestris L.) volume.
My third study explored the use of operational harvester data for the prediction of sawlog volumes using Metsään.fi attributes and other Big Geodata sources. A Random Forest model provided the best results with regard to factual sawlog volumes. The model-based approach notably improved sawlog predictions for Scots pine compared to the original Metsään.fi estimates.
Findings of this thesis indicate that remote sensing and machine vision-based methods are satisfactory when timber assortments and sawlog proportions are predicted but could be improved by additions. While certain limitations remain, improved data collection practices and advanced modelling techniques can further enhance the accuracy and usability of forest inventory systems. The results of this dissertation will contribute to the development of more efficient and data-driven forest inventory practices that may facilitate better resource allocation and sustainability in Nordic forestry.