Science Service System
You are here : Home : Proposals_Summary

Summary of Proposal MTH0430

TitleBackscatter modelling for TerraSAR-X using full-waveform airborne laser scanning data
Investigator Doubkova, Marcela - Vienna University of Technology, Institute of Photogrammetry and Remote Sensing
Team Member
Prof.Dr Wagner, Wolfgang - Vienna_University_of_Technology, Institute_of_Photogrammetry_and_Remote_Sensing
Mag Weichselbaum, Jürgen - GeoVille_Information_Systems_GmbH, Technical_Department
Dr Steinnocher, Klaus - Austrian_Research_Centers_GmbH, Earth_Observation_Application_Group
SummaryThe primary objective of the project is to assess the potential of TerraSAR-X for retrieving terrain roughness in an alpine environment. In order to allow a multi-dimensional study of the backscatter behaviour in dependence of vegetation, soil moisture and surface roughness airborne laser scanning techniques will be applied for mapping roughness as a means to collect spatially extensive reference data for studying radar roughness effects. Since the functional dependence of TerraSAR-X backscatter measurements to various airborne laser scanning (ALS) derived parameters is not known, such relationships have to be developed. The parameters include surface roughness at spatial scales in the range of the microwave wavelength, terrain roughness characterising the unevenness of the terrain (including rocks and vegetation) at scales in the order of meters to tens of meters, and volume scattering for vegetation modelling. Comparing TerraSAR-X measurements with the ALS derived parameters and available in-situ measurements will allow developing improved X-band backscatter models. Methods to be developed involve analysis of EO data mainly from TerraSAR-X. Nevertheless in order to generate reliable and robust environmental mapping products and for fully exploiting the information content of TerraSAR-X, additional EO data (ALOS L-Band, ALS data) and independent ground truth (state-of-the-art roughness measurements) will compliment the datasets. The results of this project will first comprise functional dependencies between surface parameters like surface/terrain roughness and volume scattering of different vegetation types and TerraSAR-X backscattering information. Secondly, a methodology for the derivation of terrain roughness map will be developed and evaluated within the specified test areas in Austria. The project is funded by the Austrian Space Application Programme (ASAP).
Final ReportThe proposal focuses on the investigation of the information content provided by TerraSAR-X data. A new scientific approach is pursued, which applies Airborne Laser Scanning (ALS, also referred to as LiDAR) data for deriving distributed, three-dimensional and area-wide reference data. The major aim is to increase the understanding of the backscattering processes in the TerraSAR-X band with parameters derived from laser scanning points. Particularly, the well-known but hardly standardized term and surface parameter roughness is explored. Secondly, a field work campaign was coordinated in gardens of Schoenbrunn in order to demonstrate the penetration depth of X band over deciduous forest. TerraSAR-X versus ALS roughness parameters High density three-dimensional data even in densely vegetated areas can best be provided by new ALS sensors. The following roughness parameters were already theoretically defined and should be derived from ALS data to increase understanding and modeling the SAR backscatter: Surface roughness: small scale (mm-dm) height variations in the magnitude of the microwave wavelength Terrain roughness: unevenness of the terrain surface (including rocks and low vegetation) at scales of meters to tens of meters Volume scattering layer: roughness defined by vegetation, which causes attenuation of the signal but also contributes to volume scattering; detailed geometrical description of the vegetation Following parameters were added to investigate the sensitivity of TerraSAR-X to upper most (25 cm) vegetation layer. Upper cross section: Mean backscatter cross section of points in highest 25 cm Upper mean echo: mean echo width of points located in the highest 25 cm The overall correlations of the TerraSAR-X intensity temporal statistics with the ALS roughness parameters representing lower (up to 3m) forest layer and the entire forest biomass were insignificant. Most probably the randomly oriented branches and twigs in the forest upper layers cause a random volume scattering and high saturation level preventing any relationship between forest density, biomass and surface roughness with TerraSAR-X signal. To exclude the effects of the forest saturation we also investigated linear relationships between lidar based roughness parameter representing only upper 25 cm and TerraSAR parameters (mean, range, stdev) derived from al 11 acquired scenes. Surprisingly, no simple linear relationship was found suggesting the complicated affair of forest structural components. The 69 ASAR WS scenes were obtained and processed. The parameters (mean, range and stdev) were computed open areas, young and old forests, respectively. The result provide a significant difference between open areas and forests, while not distinguishing between the two different forest classes. FIELDWORK (placing corner reflectors under deciduous forest in Schoenbrunn gardens, to judge about penetration depth of X band) Three study sites have been selected in the Schoenbrunn gardens that represented a) open field, b) deciduous forest (tree height 17-22 m) and c) deciduous forest (tree height 23 and more) (FIGURE 1, nDSM demonstrates height information of trees and other vegetation and have been derived from Liar points). The corner reflectors have been placed at these locations every day of the acquisition at the same orientation facing the sensor (Figure 3 and 4). This campaign has been limited on winter (non-leave) season due to the continuing cancelation of ordered scenes. Despite this limitation we could demonstrate some interesting results. The TerraSAR-X signal could only penetrate the more scarce deciduous forests (Number 3 In Figure 2) while demonstrating only limited increase of the backscatter from the corner reflector under the denser forest (Number 2 in Figure 2). By visual interpretation of field photos (Figure 3 and 4) and basic microwave concepts we assumed two hyypotheses: I) that the level of penetration is dependent on amount of twigs and branches at 2-3 cm width. These are positively dependent on the tree height and density II) that the uncovered large parts of the trees (branches, trunks) in the sparse forest may cause double bounce effects that allow signal to bounce towards ground, increase rapidly when reflecting from the corner reflector and bounce back. We expect strong seasonal dependency of signal penetration. This could not be demonstrated within scope of this project. The response of HH and VV polarizations differed due to the vertical orientation of the branches and demonstrated the general known fact of higher sensitivity of VV polarization to vegetation structures (Figure 5). CONCLUSION The experiments demonstrate the complexity of the relationship between forest biomass and SAR backscatter. The penetration of the X band under forest canopies in dormancy periods was demonstrated. To explain the penetration level two hypothesis were suggested: I) the X band penetration is inversely related to the amount of twigs and branches of a size of X band (3.1 cm) that cause signal extinction, II) the uncovered large parts of the trees (branches, trunks) in the sparse forest may cause double bounce effects that allow signal to bounce towards ground, increase rapidly when reflecting from the corner reflector and bounce back. Despite the demonstrated penetration we did not find any relationship between the vegetation depth/biomass as here presented by the Lidar parameters and the TerreSAR-X signal. Also no dependency was demonstrated between the TerraSAR-X signal and ALS parameters representing only upper 25 cm of the forests. We presume that only a small proportion of the signal can penetrate to the surface but is significantly increased by double bounce from the corner reflector during our field experiments. It’s also assumed that with absence of corner reflector (or other object with large cross-section) the affect of the volume scattering from randomly oriented small twigs and branches prevail. The volume scattering occurs already in the upper most tree layer and doesn't differ for forests with different biomass or vegetation height. In order to further test presented hypothesis and shed more light on the understanding of the TerraSAR-X signal an additional research over larger study areas is recommended. Especially analyses over larger and heterogeneous landscapes (combination of open fields, coniferous and deciduous forests) with both laser scanning and TerraSAR-X datasets are needed. PUBLICATIONS The ASAP-V reports: ---------------------------- Adams, M., Gangkofner, U., Stemberger, W., Doubkova, M., Höfle, B., Hollaus, M., Sabel, D., Aubrecht, C. & Steinnocher, K. (2009): SAR-X Environ: Environmental Mapping Applications using TerraSAR-X. Technical Report No. 817082, Final report submitted to the FFG in the frame of the project SAR-X Environ financed by the Austrian Space Applications Programme, pp. 1-235. W. Stemberger, B. Höfle, K. Steinnocher: "SAR-X Environ: Environmental Mapping Applications using TerraSAR-X"; Report for User Requirement Document submitted to the FFG in the frame of the project SAR-X Environ financed by the Austrian Space Applications Programme; Report No. 817082, 2008; 42 pages. W. Stemberger, B. Höfle, K. Steinnocher: "SAR-X Environ: Environmental Mapping Applications using TerraSAR-X"; Report for Midterm report submitted to the FFG in the frame of the project SAR-X Environ financed by the Austrian Space Applications Programme; Report No. 817082, 2008; 94 pages. Acknowledged in: ---------------- Höfle, B., Hollaus, M., Lehner, H., Pfeifer, N. & Wagner, W. (2008): Area-based parameterization of forest structure using full-waveform airborne laser scanning data. In: Proceedings of Silvilaser 2008. Edinburgh, Scotland, pp. 227-235 (on CD-ROM). Wagner, W., Hollaus, M. & Höfle, B. (2009): Terrain characterization and vegetation structural analysis with full-waveform airborne laser scanners. In: Proceedings of Remote Sensing and Photogrammetry Society Annual Conference 2009. Leicester, United Kingdom, pp. 208-213.

Back to list of proposals

© DLR 2004-2016