K&C Science Report – Phase 1 Detection of deforestation in Swedish forest

Abstract—An extensive dataset of ALOS PALSAR L-band Synthetic Aperture Radar (SAR) backscatter images is investigated for clear-cut detection in Swedish forest. SAR data were available for the counties of Västerbotten and Västra Götaland as well as for two local test sites (Remningstorp and Krycklan...

Full description

Bibliographic Details
Main Authors: Johan E. S. Fransson, Håkan Olsson, Leif E. B. Eriksson, Lars M. H. Ul, Maurizio Santoro
Other Authors: The Pennsylvania State University CiteSeerX Archives
Format: Text
Language:English
Subjects:
Online Access:http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.530.638
http://www.eorc.jaxa.jp/ALOS/en/kyoto/phase_1/KC-Phase1-report_SLU.pdf
Description
Summary:Abstract—An extensive dataset of ALOS PALSAR L-band Synthetic Aperture Radar (SAR) backscatter images is investigated for clear-cut detection in Swedish forest. SAR data were available for the counties of Västerbotten and Västra Götaland as well as for two local test sites (Remningstorp and Krycklan). A strong forest/non-forest contrast and temporal consistency were found for the Fine Beam Dual HV-polarized backscatter in summer/fall. Thus, a simple thresholding algorithm could be used for clear-cut detection. Using the ALOS PALSAR data and methods applied so far, most pixels in the clear-felled areas could be correctly classified as changed. For the county of Västerbotten in northern Sweden, up to about 50 % of the pixels were correctly classified as changed for about 90 % of the clear-felled areas using a 2 dB threshold. The results were less good for the county of Västra Götaland in southern Sweden, where only up to about 40 % of the pixels could be correctly identified as changed for about 65 % of the clear-felled areas. For the south county, also much over classification of non changed areas occurred. It would still be possible to use the ALOS PALSAR data in a sampling routine, where changed areas are checked against cutting permits and samples of the remaining detected changes should be checked in situ for determining type of change. In the extension phase of the project, an up scaling of the mapping of clear-cuts, and possibly also biomass, to all of Sweden is planned. There is also a need for further algorithm development. Index Terms—ALOS PALSAR, K&C Initiative, boreal forest, forest theme, deforestation, clear-cuts, Sweden.