Modelling presence of swamp forest and forest dwelling birds in a boreal forest reserve using airborne laser scanning

John Wirkola Dirksen presented a lecture on a predefined topic and defended his doctoral thesis on November 1,  2013. The topic for the trial lecture was “Reserve selection approaches with particular emphasis on forest biodiversity”  and the title of his thesis was “Modelling presence of swamp forest and forest dwelling birds in a boreal forest reserve using airborne laser scanning”. We congratulate!

Thesis abstract

Since the millennium there has been a huge increase in ecological studies which deals with distribution modelling, and it is a fact that distribution modelling has branched of from mainstream ecology and is now considered an independent branch within ecological science, as a key subfield of Conservation Biology under the name: Conservation Biogeography. When developing distribution models regardless of whether it is for the distribution of species or nature types it is essential to have environmental data of high quality covering large geographical areas. Such data is available through remote sensing. Airborne laser scanning (ALS) is a relatively new remote sensing technique providing modellers with 3D data which can be used as proxies for environmental variables e.g. terrain slope, vegetation height and density of forest canopies. Within the last decade ALS has been implemented in many studies which deals with distribution modelling.

In this thesis the main objective was to implement ALS data in ecological studies to gain more knowledge about the object in focus and to create guidelines for future use of ALS in similar studies. This thesis consists of three studies which all were conducted within an old-growth forests reserve in southeast Norway. Specific objects were addressed in the individual papers: For Paper 1 and 2 the main objective was to develop distribution models for boreal swamp forests. The two studies differed in the sense that Paper 1 used variables which described the topography of the terrain as environmental variables. Models were developed using the distribution modelling software Maxent. Two distinct MaxEnt modelling methods were compared, the standard MaxEnt procedure (SMP) where models were created using default settings in the software and the alternative MaxEnt procedure (AMP) where default settings were deactivated and manual selection of variables were conducted when models were built. In Paper 2 proxies for forest structures were used as environmental variables when GLM models were built. In Paper 3 the object in focus was forest dwelling birds. Models were developed for both total species richness and eight single bird species.Distribution models were developed using both GLM and Random Forest. We used ALS variables to describe forest structure, and variables derived from spectral images to describe tree composition in each of 157 counting stations.

Highly informative distribution models were developed in Paper 1. SMP models tend to be overfitted compared to AMP models. Slope, topographic position, ruggedness and curvature play an important role for the location of swamp forests. In Paper 2 we found that forest structures differ between the two nature types. The mean vegetation height is approximately 3 m lower within swamp forests compared to non-paludified forests. The best model classified 62% of the cells correctly. For Paper 3 variables which described the density of the canopy within certain one meter sections turned out to be the most important sort of variables for six out of the eight most abundant bird species and species richness. The best distribution models were created for Goldcrest (Regulus regulus) and Wren (Troglodytes troglodytes), for the other species and species richness the models predicted poorer but still in accordance with previous field measures. We conclude that ALS data has an important role to play in the future development of distribution models and thereby Conservation Biogeography.

We recommend future modellers to always evaluate distribution models on an independent evaluation data set before implementing them into management plans.

Supervisors

Professor Mikael Ohlson (main supervisor) (INA, UMB)
Professor Rune Halvorsen (University of Oslo, Natural History Museum)
Dr. Katrine Eldegard (INA, UMB)
Professor Erik Næsset (INA, UMB)
Professor Terje Gobakken (INA, UMB)

Evaluation committee

Professor Lena Gustafsson (Department of Ecology, Swedish University of Agricultural Sciences)
Dr Ross Hill (Bournemouth University, School of Applied Sciences)
Professor Jon Swenson (INA, UMB)

Reference

  • Dirksen, J. W.. ((2013). Modelling presence of swamp forest and forest dwelling birds in a boreal forest reserve using airborne laser scanning.). PhD Thesis.
    [Bibtex]
    @PhdThesis{Dirksen2013,
    Title = {Modelling presence of swamp forest and forest dwelling birds in a boreal forest reserve using airborne laser scanning},
    Author = {John Wirkola Dirksen},
    School = {Norwegian University of Life Sciences},
    Year = {2013},
    Owner = {hanso},
    Timestamp = {2013.11.05},
    Url = {https://static02.nmbu.no/mina/forskning/drgrader/2013-Dirksen.pdf}
    }