Investigations on forestry applications in Sweden using ALOS PALSAR
Maurizio Santoro(1), Johan Fransson(2), Leif Eriksson(3), Mattias Magnusson(2), Klas Folkesson(3), Lars Ulander(3) and Håkan Olsson(2)
(1) Gamma Remote Sensing, Worbstrasse 225, 3073 Gümligen, Switzerland
(2) Swedish University of Agricultural Sciences, Skogsmarksgränd, 901 83 Umea, Sweden
(3) Chalmers University of Technology, -, 412 96 Göteborg, Sweden
Intended for the K&C Special Session
Objective of this paper is to present recent advances on forest monitoring using ALOS PALSAR data in Swedish boreal forest. Because of the rather strong sensitivity of the L-band signal to forest structural properties, a number of applications can be assumed to be feasible. Here we present results on ongoing analysis concerning clear-cut detection and forest stem volume retrieval. Initial results have been presented in (Eriksson et al., Proc. 5th BioGeo, 2007, Fransson et al, Proc. IGARSS’07, 2007).
With respect to the previous investigations, this analysis is based on a much larger dataset of images, which have been acquired since the start of the ALOS mission back in 2006. SAR backscatter data are available both at the local scale and for larger areas. At the local scale the investigations focus on two test sites, Remningstorp and Brattåker, for which accurate and updated forest inventory measurements are available. At the regional scale we focus on the counties of Västra Götaland and Västerbotten, of which the two test sites are part of respectively. The test sites serve to gain understanding on the PALSAR signatures and for the development of mapping methods. Upon successful establishment of the methods, these are then applied to obtain county-wide estimates of clear-cut areas and stem volume.
The SAR dataset for the forest test sites consists of stacks of geocoded and co-localized SAR backscatter images in Fine Beam Single, Fine Beam Dual and Polarimetric modes with 20 m pixel size. The SAR dataset for the regional studies consists of SAR backscatter data strips acquired in Fine Beam modes and obtained through JAXA’s Kyoto & Carbon Initiative. All strips have been geocoded to 50 m pixel size and co-registered to each other to ensure accurate geo-location.
An analysis of recently felled areas revealed that they can present a drop in SAR backscatter up to about 2 to 3 dB, thus showing that clear cuts seem to be detectable. Nonetheless, the SAR backscatter signal presents significant variations depending on environmental conditions, which needs to be taken into account when developing an automatic method for clear cut detection. Stem volume retrieval also depends upon environmental conditions. An initial assessment based on a small dataset showed that for winter-dry conditions a retrieval error of about 30% can be obtained. With the availability of the larger dataset, it is expected that a multi-temporal combination will decrease the error, as already demonstrated in (Askne et al., TGRS, 2003; Santoro et al., IJRS, 2006).