Assessment of Decadal Change in North American Wetlands Based on JERS and PALSAR Space-Based L-band SAR Data

Northern wetlands are believed to have sequestered about one third of the total global pool of soil carbon. With warmer, drier conditions occurring throughout the arctic as a consequence of global warming, they have the potential to evolve into major sources of atmospheric carbon. The ability to characterize long-term changes in the condition of northern wetlands is therefore essential to the development of accurate global carbon budgets. Space-based synthetic aperture radar (SAR) produces high-resolution imagery of vast inaccessible regions regardless of clouds or solar illumination. Especially at L-band, it presents a unique tool for monitoring changes in the characteristics of vegetated wetlands due to its inherent ability to sense vegetation structure and moisture content, biomass, and standing water underneath vegetation canopies. We have, therefore, been using L-band SAR imagery from two different spaceborne sensors separated in time by approximately one decade, JERS and PALSAR, to produce a thematic map of change in the types and extent of wetlands in Alaska and Canada. Two-season JERS imagery characterizes the wetlands status for the 1997 time frame while dual-polarized PALSAR imagery captures the wetlands status for the 2007 time frame. The SAR imagery is supplemented with ancillary information such as texture, image collection dates, slope, and latitude. The classification algorithm applied to each set of imagery is based upon the Random Forests technique. To classify, it first forms a multitude of decision trees based on characteristics of the imagery and ancillary data within designated training regions, then assigns each pixel in the raster to the wetlands class chosen for that pixel by the most decision trees. We have been generating a PALSAR-based wetlands map of Alaska along with an associated JERS-to-PALSAR decadal change map. We have also been extending our JERS-based wetlands map into Canada. Since our classifications are based on SAR, ancillary, and ground reference data from disparate data sources extending over large geographic regions, the ability to maintain good geographic registration has been a major focus. Additionally, we continue to explore options for including optical data sources in the classification and for upgrading to a segmented implementation. The accuracy of each thematic map is verified using ground reference data. For the parts of the PALSAR-based wetlands map completed so far, for example, the overall producer accuracy is about 90%. The most prevalent decadal change identified has been “emergent” wetlands changing into “scrub/shrub” wetlands. Our results are expected to demonstrate the utility of multi-platform satellite L-band SAR imagery for characterizing transitions in the extent and type of vegetated wetlands. This work has been undertaken in part within the JAXA ALOS Kyoto & Carbon Initiative. It was carried out at the University of Michigan and at the Jet Propulsion Laboratory, California Institute of Technology, under contract to the National Aeronautics and Space Administration