Forest Attribute Prediction and Mapping using 3D Remote Sensing Data

Abstract

Forest inventory enables collection of essential data on forest attributes such as volume (VOL), aboveground biomass (AGB), species composition, age, and forest health. Knowledge about these attributes are vital for strategic and tactical forest management purposes, including planning timber harvests, conserving biodiversity, estimating carbon sequestration, and forecasting future yields. Forest inventory practices have evolved significantly over the past century along with the development of remote sensing (RS) assisted inventory approaches. This thesis focuses on using 3D RS data acquired from different platforms and with different remote sensors, for example - airborne laser scanning (ALS), digital aerial photogrammetry, and synthetic aperture radar. The individual papers focused on different forest regions and different spatial extents of acquired RS data for the prediction, estimation, and mapping of forest attributes such as, VOL and AGB, for various cases of model-based inference. The included papers have shown that, 3D RS data can be successfully integrated as auxiliary data and reference data within model-based inference frameworks. A combination of dense and sparse ALS data can be used for forecasting forest VOL growth through VOL models. Several methods are also employed in the individual papers to quantify uncertainty, including root mean square error, confidence intervals, and prediction intervals. Overall, this thesis concludes that 3D RS data is efficient for accurate forest attribute prediction, supporting cost-effective forest monitoring and management solutions. The integration of RS data into forest inventory practices continues to evolve, offering new opportunities for large-scale forest monitoring, resource management, and biodiversity conservation.