University of Twente Student Theses


Extraction of forest structural attributes as indicators of landslide-induced disturbance using LiDAR data

Gode, Ameya (2012) Extraction of forest structural attributes as indicators of landslide-induced disturbance using LiDAR data.

[img] PDF
Abstract:Europe’s forests were hit hard by storms in 1990 and 1999, which caused 120 and 180 million m3 of damage respectively .Since forests are important in the context of carbon sequestration and biodiversity conservation, this triggered off a series of studies of forest disturbances/damage. Natural Resource scientists are researching many of these disturbance factors in order to understand them better and to develop control or mitigation methods. One of these natural disturbance factors is landslide activity. This study focuses on the effect of such slow moving translational, rotational and earth flow processes on forest in the Bois Noir landslide in the Alpes de Haute province of France. Previous researches in this area to study landslides under forests have been done with different methods like dendrogeomorphology and climate data analysis. This study makes use of high resolution airborne LiDAR data (with a mean point density of 180 points per m2.). Since LiDAR has long proven to be the best known remote sensing application for studying forest structure, this study was also aimed at assessing the applicability and accuracy of using LiDAR data. The prime focus of this study was to extract various forest structural attributes from the LiDAR point cloud and compare these attributes between ‘stable’ and ‘unstable ‘zones. From previous research and ecological knowledge, it was decided that the following structural attributes are to be taken into consideration while doing analysis: Tree Height, Diameter at Breast Height (DBH), tree inclination and orientation and canopy gaps. The LiDAR point cloud was subjected to normalisation and gridding to form a canopy height model (CHM), which was used for most of the analysis. A region growing approach in ECognition was then used to segment tree crowns and canopy gaps. Also, any human induced features, edaphic features and ‘obvious’ areas where landslide had caused massive tree fall were excluded from analysis. This approach gave a ‘gap map’ of the area and also information about tree heights and tree density, which was then subjected to statistical analysis. To validate the LiDAR data and assess its accuracy, field data was collected in the month of September 2011. From the analysis, it was found that tree heights can be estimated with an R2 value of 0.72. The gap detection accuracy was 81%, in a canopy height model of 15cm grid size. Furthermore, it was proven that statistically there is a significant difference in the distributions of canopy gap area, gap shape and tree heights between stable and unstable zones. Thus, these can be considered as indicators of landslide activity, but it also needs some multi-temporal analysis to study the dynamics of these structural attributes.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page