Detection of small single trees in the forest-tundra ecotone using airborne laser scanning

Nadja Stumberg presented a lecture on a predefined topic and defended her doctoral thesis on Friday 16, November 2012. The topic for the trial lecture was “Remote sensing aided monitoring of the forest-tundra ecotone”  and the title of her thesis was “Detection of small single trees in the forest-tundra ecotone using airborne laser scanning”. We congratulate with the degree!

Thesis abstract

Alpine and arctic tree lines are expected to advance to higher altitudes and further north due to global warming. The forest-tundra ecotone in particular, is highly sensitive to climatic changes since many of the species found there are at their tolerance limits. Thus, the development of suitable methods for monitoring these changes is of great importance and interest. For the monitoring of such vast areas as the forest-tundra ecotone, airborne laser scanning (ALS) may provide a well-suited tool because of its capability to estimate biophysical parameters on single tree level at different geographical scales.

Small birch tree in the forest-tundra ecotone

The main objective of this thesis was to investigate the potential of using high-density ALS data for detection of small individual trees located in the forest-tundra ecotone. The specific parts of the thesis focus on (1) single tree detection using ALS height values in combination with variables describing tree characteristics and the site, (2) laser echo classification using laser height and intensity, geospatial and terrain variables, as well as geostatistics and statistical measures. Furthermore, (3) the potential of an unsupervised classification of raster cells for automated monitoring programs of small single trees was assessed. Along a 1,500 km long transect stretching from northern Norway (66°19’ N) to the southern part of the country (58°3’ N) field measurements of 744 small individual trees as well as ALS data were collected. Generalised linear models (GLM), a generalised linear mixed model (GLMM), support vector machines (SVM), and a raster-based algorithm concept were employed for the detection and classification of both trees as well as tree and non-tree laser echoes using different variables. Successful single tree detection using laser height values in combination with tree characteristics and spatial influences as latitude and region was verified for trees exceeding a height of 1 m using GLM and GLMM models. The results form a solid basis for generalisation and inference that goes far beyond previous research because of the huge geographical extension of the dataset. Secondly, the capability of the ALS data for classification into tree and non-tree echoes using laser measurements, geospatial and terrain variables was confirmed using the two different modelling techniques GLM and SVM. Furthermore, an extension of the classification models with geostatistical and statistical measures employing GLM and SVM revealed a significant improvement. Finally, the suitability of an unsupervised classification approach for the automatic detection of small single trees was verified for parameter values ensuring a justifiable trade-off between rates of successful detection and commission errors.


Professor Erik Næsset (main supervisor) INA, UMB
Professor Terje Gobakken INA, UMB
Researcher Ole Martin Bollandsås INA,UMB

Evaluation committee

Professor Benoit St-Onge, Université du Québec à Montréal (UQAM), Canada
Professor Håkan Olsson, Swedish University of Agricultural Sciences, Sweden
UMB coordinator: Associate Professor Kari Klanderud, INA, UMB


  • N. Stumberg, “Detection of small single trees in the forest-tundra ecotone using airborne laser scanning,” PhD Thesis, 2012.
    [Bibtex] [Download PDF]
    Title = {Detection of small single trees in the forest-tundra ecotone using airborne laser scanning},
    Author = {Nadja Stumberg},
    School = {Norwegian University of Life Sciences},
    Year = {2012},
    Owner = {hanso},
    Timestamp = {2012.11.21},
    Url = {}

Leave a Reply

Your email address will not be published. Required fields are marked *