Druckansicht der Internetadresse:

Fakultät für Biologie, Chemie und Geowissenschaften

Lehrstuhl für Ökologische Modellbildung - Prof. Dr. Michael Hauhs

Seite drucken

Masterarbeit

Ragram: Radargrammetry R Package to Estimate Forest Canopy Height

Anja Kleebaum (02/2016-09/2016)

Betreuer: Holger Lange, Christina Bogner

Radargrammetry is a method to derive surface height information (i. e., a digital surface model, DSM) from two or more Synthetic Aperture Radar (SAR) images.

These images cover (partly) the same  area and are obtained from the same viewing direction, but they were recorded using different  incidence angles. Like in photogrammetry, pixels belonging to the same object (homologous  points) are identified on the corresponding images using an appropriate matching algorithm.  The height difference between these pixels  is calculated using the parallax, which is the angle between two straight lines that are crossing  each other on the homologous point.  Subsequently, a digital terrain model (DTM) is subtracted from the DSM. Thus, a canopy  height model (CHM) is obtained. The CHM is converted to abovegroun tree biomass using simple regression  relations obtained from ground truth data. The method has therefore  the potential to monitor biomass changes when being observed over time.
 
There is a range of satellite missions producing data which are suitable for radargrammetry.  For this thesis, Radarsat-2, TerraSAR-X and the new Sentinel mission will be used.  Sentinel-1 comprises two polar-orbiting satellites, i. e. 1A and 1B. Sentinel-1A was launched in April 2014, whereas Sentinel-1B  will be launched in 2016. 
 
In this thesis, an R-package is developed to derive a DSM from Sentinel-1 Radar images and  its accuracy to predict canopy heights is evaluated using e.g. Norwegian forest inventory data.  This work is a collaboration between the Norwegian Institute of Bioeconomy Research and the Department of Ecological Modelling at the University of Bayreuth.

FacebookTwitterYoutube-KanalBlogKontakt aufnehmen
Diese Webseite verwendet Cookies. weitere Informationen