Assessing condition and changes in forest ecosystems using remotely sensed data

Marie-Claude Jutras-Perreault defended her PhD-thesis entitled “Assessing condition and changes in forest ecosystems using remotely sensed data” on August 25, 2023.

Photo: Ole Martin Bollandsås

The trial lecture was entitled “Assessing biodiversity in boreal forests by remote sensing”

Forest ecosystems yield a range of ecosystem services that directly or indirectly contribute to human wellbeing. The potential of forest ecosystems to deliver services is, however, directly related to their condition, or state, and the pressure they are subjected to. To assess the condition of forest ecosystems in terms of biodiversity, the presence of dead trees and natural forests are two indicators often used. In Europe, the changes in forests are mainly driven by forest activities. Forest cover change is therefore an important indicator of the pressure on forests. Remotely sensed data can provide an objective and systematic approach to perform a wall-to-wall assessment of forest ecosystem condition and to map forest cover change over large areas. The main objective of this research was to explore the possibilities of using remotely sensed data to assess condition and changes in Norwegian forest ecosystems.

The first study assessed the potential of airborne laser scanning (ALS) data in combination with optical remotely sensed data to predict the presence of standing dead trees in a managed forest in Norway. Two approaches were compared: areabased and tree-based. The results revealed that, in the context of small occurrences of standing dead trees, a tree-based approach was more performant than area-based regression models to predict the presence of standing dead trees. Building upon the findings of the first study, the second study compared the performance of remotely sensed data with different spatial and spectral resolutions to detect standing dead trees. The study indicated that the presence of standing dead trees could be predicted over large areas by combining ALS-derived canopy height models for tree identification and optical images with a spatial resolution of <3m, including a nearinfrared band, for determining the tree status. In the third study, the presence of natural forests in three Norwegian counties was predicted using ALS-derived variables in combination with National Forest Inventory data. The study underscored the value of ALS data as an effective tool for predicting the presence of natural forests. It was found that stratifying the field plots by dominant tree species improved the accuracy of predictions, while acquiring additional field data targeting natural forests did not yield significant improvements. The last study compared two algorithms, LandTrendr and Global Forest Watch, for estimating stand- and landscape-level forest cover changes. It was concluded that LandTrendr provides
efficient spatial and temporal indicators of forest change dynamics and outperformed Global Forest Watch in identifying and monitoring clear-cuts in a Norwegian boreal forest.

This research has demonstrated the potential of using remotely sensed data to assess forest ecosystem condition and forest cover changes over large areas. Our findings provide valuable insights that can support operational forest management planning efforts aimed at preserving the condition of forest ecosystems to a level that ensures the provision of essential services.

Veilederne var
Hans Ole Ørka (NMBU)
Terje Gobakken (NMBU)
Erik Næsset (NMBU)

Evalueringskomiteen bestod av
Eva Lindberg (SLU)
Petteri Packalén (LUKE)
Maarit Kallio (NMBU)