Detection of Root, Butt, and Stem Rot in Picea Abies with Remotely sensed Data


Benjamin Allen successfully defended his doctoral thesis, “Detection of Root, Butt, and Stem Rot in Picea Abies with Remotely sensed Data”, on 14 March 2023.

The topic for the trial lecture was “Challenges in area-wide biodiversity inventories in Scandinavian boreal forests”. We congratulate!


Thesis abstract

Root, butt, and stem rot fungi (RBSR) are among the most significant forest pathogens in Norway. RBSR leads to decay of the tree stem and disruption of vascular tissue. Impaired vascular function reduces tree growth and can be fatal to certain tree species. Decay in the tree stem leads to reductions in wood quality and associated economic losses. Several methods exist for management of RBSR in Norway spruce (Picea abies L. Karst), including early harvest of heavily infected stands and alteration of tree species mixes during stand regeneration. Acquiring information about the prevalence of RBSR with field methods is time-consuming and cost-prohibitive due to the lack of obvious visual symptoms. Remotely sensed data has been proposed as an alternative method of detecting RBSR. The primary objective of this thesis was to test the utility of several remotely sensed data sources for detecting RBSR in Norway spruce.

The first study involved the use of airborne hyperspectral imagery to detect RBSR. Two classification algorithms were used: support vector machines and random forest. Overall classification accuracies of 64.8% (κ=0.27) were obtained with airborne hyperspectral imagery. The second study explored the potential for detecting RBSR with UAV-based hyperspectral imagery and compared the use of two different sensors: a 488-band hyperspectral sensor and a 29-band hyperspectral sensor. Higher accuracies were obtained with the 488-band sensor (75.8%, κ=0.24) than with the 29-band sensor (60.1%, κ=0.13). Comparisons of spectral indices were also performed; results suggest that RBSR may affect the values of several spectral indices. The third study compared the classification accuracies obtained with airborne hyperspectral imagery, airborne laser scanning, and historical multispectral imagery used alone and in combination. Fusion of airborne hyperspectral imagery and airborne laser scanning was found to deliver the highest classification accuracy (66.1%, κ=0.32). Hyperspectral imagery delivered the highest accuracy of any data source when used alone (64.3%, κ=0.28), followed by airborne laser scanning (59.3%, κ=0.19).

This thesis shows that detection of RBSR infection in Norway spruce with remotely sensed data is possible, although classification accuracies obtained were modest. Hyperspectral imagery and airborne laser scanning appear to be the most promising technologies for detecting RBSR with remotely sensed data. Further research should 4 aim to apply these methods to detecting RBSR in other tree species and to assess the utility of applying these methods in operational forestry.


Main supervisor: Professor Terje Gobakken, MINA-NMBU
Dr. Michele Dalponte, Fondazione E. Mach
Dr. Hans Ole Ørka, MINA-NMBU
Professor Erik Næsset, MINA-NMBU

Evaluation committee:

Professor Fabian Faßnacht, Freie Universität Berlin
Professor Markus Holopainen, University of Helsinki
Associate Professor Kyle Eyvindson, MINA-NMBU