ºÚÁÏÍø

Title of the thesis: Close-range remote sensing of trees using multiwavelength terrestrial laser scanning and drone-based photogrammetry with spectral imaging

Thesis defender: Olli Nevalainen
Opponent: Professor Andreas Wieser, ETHZ, Switzerland
Custos: Matti Vaaja, Aalto University School of Engineering, Department of Built Environment

Forests and their soils are the largest carbon stock in land ecosystems; thus, knowing the biomass stored in trees and understanding their current and future carbon uptake is needed for climate science and policy, commercial forestry, and building accurate digital twins. This doctoral thesis presents new laser- and drone-based technologies that enable detailed measurements of individual trees. 

In the first part, one of the world's first multiwavelength laser scanners was used to measure pine shoots and a single pine tree across a growing season. The data revealed changes in chlorophyll content, an indicator of a tree's health and photosynthetic activity, and produced three-dimensional data showing how chlorophyll varies within the crown of a tree. 

In the second part, drones equipped with RGB, multispectral, and hyperspectral cameras collected data from 15 forest plots in Finland. Automated methods for detecting and classifying individual trees and estimating standard forest inventory variables, such as height, diameter, and volume, at the individual tree and plot levels were developed. The developed tree detection methods for the drone-based camera data performed as well as established laser scanning-based approaches. Machine learning models classified four common Finnish tree species–pine, spruce, birch, and larch–with over 90% accuracy. Hyperspectral cameras provided the best species classification but required an RGB camera alongside for accurate tree detection and estimation of tree inventory variables. On the other hand, a multispectral camera alone came close to matching this performance using only one device. 

Forests are central to mitigating climate change, and automated measurement of individual trees is needed for climate research and policy, commercial forestry, and building detailed digital models, or "digital twins", of forests. As drones carrying laser scanners and spectral cameras become even more common, these findings, among the first of their kind, can guide the development of future remote sensing sensors and methods for monitoring forests tree by tree.

Key words: remote sensing, laser scanning, spectral imaging, drones, forestry

Thesis available for public display 7 days prior to the defence at . 

Contact information: olli.nevalainen@fmi.fi