CHARACTERIZATION OF SIBERIA LARCH FOREST FROM PALSAR L-BAND RADAR AND LANDSAT VCF DATA TRAINED BY HIGH-RESOLUTION STEREO AND FIELD DATA

Abstract:

Larch (Larix spp.) dominant forests occupy about 70% of the permafrost areas in Siberia. Improved understanding of the fate of the larch forests in the face of climate change requires a concerted long term research effort to detect and quantify ecosystem responses. Satellite data analysis offers much improved information on changes in the forest-tundra transitional zone and on forest structure in this area.

The LANDSAT data and radar data such as ALOS/PALSAR provide extensive coverage at high resolution (30m pixel). In this study, the utilization of these data for regional canopy height and biomass mapping was examined. The major challenge for characterization of forest in a remote area such as Siberia is the lack of adequate number of training and validation data. The results of using GLAS waveform data to extend field measurements which has been successful in other ecosystems were found to be poor because most of the waveforms were acquired when the deciduous larch tree had no needles. In this study, the stereo optical data from high resolution spaceborne imagery (HRSI) were used to a generate canopy height model (CHM) by identifying both tree top and ground surface in these sparse forests. The regression models between the field measured canopy height, biomass and the stereo CHM were developed and the field data were extended to the forested areas within the entire stereo images, providing more data for training and validation. We compare the mapping results from two methods. One uses GLAS waveform data and the other uses high resolution stereo imagery to extrapolate field measurements. Results demonstrate the potential and limitations using high resolution stereo imagery to provide training information of sparse larch forest in Siberia.