Polarimetric observations of Tundra landscapes with RADARSAT-2

Jennifer.Sobiech [ at ] awi.de


Radar remote sensing offers the possibility to monitor the vast uninhabited polar landscapes remotely and independent of the presence of sunlight and cloud cover. Fully polarimetric radar observations offer the potential to improve the separation of different surface classes and achieve information about the contributions of particular scattering mechanisms, which are related to the structure of the scattering objects. Single channel radar intensity images from Tundra landscapes are in certain cases difficult to interpret. For example, often it is not clear if the main signal is received from a moss cover or the soil underneath. This study analyses the potential of polarimetric decomposition methods to determine the origin of the backscattered radar signal. Our dataset consists of 20 fine quad-pol RADARSAT-2 images from the southern central part of the Lena River Delta, Siberia, Russia, at 72°N, 126°E, with incidence angles in the range of 32.40° to 38.00° (FQ 13, FQ 15, FQ 17). The images span the time period from late April to November 2011 with acquisitions every 10 to 14 days, thus covering all seasons from late winter over spring, summer, and fall to the next winter. All images were geocoded to a pixel size of 10 m x 10 m. Our test site is a typical high arctic tundra landscape, underlain by continuous permafrost. The images cover three different geomorphological units with different soil properties and vegetation covers, several river channels and lakes up to 1.3 km² in size. Two automatic weather stations (AWS) are located in the investigation area on two of the different geomorphologic units, recording air temperatures in 0.5 m and 2 m heights, snow depth, wind speed and direction, incoming and outgoing shortwave and longwave radiation, soil temperature and soil moisture in various depths, and precipitation. An automatic camera, taking one image per day, is located at one of the AWS in addition. During a field campaign in 2010, the vegetation cover and spatial distribution of the soil moisture patterns were mapped extensively. A high resolution aerial image (0.3 m resolution) is available for Samoylov Island, located within the investigation area. Change detection methods were applied separately on the HH, HV, VH, and VV polarized images during the whole time period. Strong backscatter variations of around 3 dB appear due to snowmelt, the onset of active layer thaw in spring and refreeze during fall. During summer, no significant backscatter variations could be observed. The decomposition parameters entropy, alpha and anisotropy were calculated from the RADARSAT-2 images. First results display the pattern of the snowmelt during spring 2011 in agreement with the in-situ data from the AWS. The differences in the vegetation cover between the geomorphologic units are clearly distinguishable when using the decomposition scheme. The segmentation into the H / alpha space show that the signal received by the radar is dominated by surface scatter in case of the polygonal tundra covered by a thick moss layer, and by volume and dihedral scattering in case of the floodplains, which are vegetated by small Salix bushes, Equisetum and grasses. The difference between the different units becomes more pronounced during summer in comparison to spring, which is in agreement with the greening of the vegetation. Further analyses addressing the potential of quad-pol C-band SAR for the monitoring of soil moisture variations will be carried out in the near future, using additional polarimetric parameters.

Item Type
Conference (Poster)
Primary Division
Primary Topic
Peer revision
Not peer-reviewed
Publication Status
Event Details
Earth Observation and Cryosphere Science, 13 Nov 2012 - 16 Nov 2012, ESA, Frascati, Italy.
Eprint ID
Cite as
Sobiech, J. , Boike, J. and Dierking, W. (2012): Polarimetric observations of Tundra landscapes with RADARSAT-2 , Earth Observation and Cryosphere Science, ESA, Frascati, Italy, 13 November 2012 - 16 November 2012 .


Research Platforms


Edit Item Edit Item