Public defence in geoinformatics, M.Sc. (Tech.) Eelis Halme

Title of the thesis: Retrieving boreal forest structure from remote sensing data using reflectance modelling and machine learning

Thesis defender: Eelis Halme
Opponent: Professor Ruben Valbuena, Swedish University of Life Sciences, Sweden
Custos: Professor Miina Rautiainen,Aalto University School of Engineering, Department of Built Environment

The boreal forest is the largest terrestrial biome on Earth and plays an integral role in the global climate system. Boreal forests, which are increasingly affected by climate change, hold significant ecological value and are central to the global carbon cycle. Therefore, effective monitoring of these forests is of great importance.

This doctoral thesis set out to advance methodologies for monitoring the ecologically important northern European boreal forests, with a primary focus on Finnish forests. Specifically, the thesis aimed to develop a new hybrid method that combined a forest reflectance model with a machine learning algorithm for retrieving forest structural variables from optical remote sensing data.

The results revealed that the utility of hyperspectral imaging stands out for variables linked to species-specific information, whereas traditional spaceborne multispectral remote sensing data is sufficient for accurately retrieving common forest structural variables. Additionally, the results suggested that the spatial pattern of trees and the ratio of branch area to leaf area have the greatest influence on the accuracy of forest reflectance modelling. Furthermore, the results demonstrated that hybrid methods hold great promise for retrieving forest structural variables in northern European boreal forests.

While the thesis presents valuable findings for advancing boreal forest monitoring methodologies, it also outlines several challenges and limitations that must be addressed. Tackling these issues will require continued investigation, underlining the importance of further research. Overall, the thesis establishes a solid foundation for developing hybrid methods that use forest reflectance models.

Keywords: remote sensing, Sentinel-2, hyperspectral imaging, boreal forest, reflectance modelling, machine learning, hybrid inversion

Thesis available for public display 10 days prior to the defence at Aaltodoc.