Close Window

Abstract Location ID: 132

Use of Space-Based L-Band SAR for Mapping and Change Detection of Boreal Wetlands in North America

Jane Whitcomb, University of Michigan, jbwhit@umich.edu (Presenting)
Mahta Moghaddam, University of Michigan, mmoghadd@eecs.umich.edu
Kyle McDonald, Jet Propulsion Laboratory, kyle.mcdonald@jpl.nasa.gov
Erika Podest, Jet Propulsion Laboratory, erika.podest@jpl.nasa.gov
Bruce Chapman, Jet Propulsion Laboratory, bruce.chapman@jpl.nasa.gov

Space-based L-Band synthetic aperture radar (SAR) can be used to efficiently map different wetlands throughout extensive ecoregions. Previously we classified HH-polarized L-band SAR imagery collected from the Japanese Earth Resources Satellite (JERS) in 1997-1998 to form a 100 m resolution wetlands map of Alaska. We are now classifying HH- and HV-polarized L-band SAR imagery collected from the Phased Array L-Band SAR (PALSAR) satellite in 2007 to develop a second 100 m resolution wetlands map. A thematic map of differences between the JERS-based map and the PALSAR-based map reveals how Alaskan wetlands changed during the 1997-2007 decade.

Supporting data used in the classification include image collection dates, elevation, slope, open water masking, proximity to water, and latitude. The classification algorithm is based on an open-source decision-tree classifier called “Random Forests”. It requires ground reference data of known wetlands/uplands class for training and validation, which are taken from standardized FWS/USGS products for US classifications and from Canadian study sites such as BOREAS and Mer Bleue for Canadian classifications (for regions lacking ground reference data, classifications are based on decision trees saved from nearby same-ecosystem regions that do have ground reference data). Producer and user error statistics are calculated based on confusion matrices generated by Random Forests. Averaging results across regions completed so far, our aggregate classification accuracy is 88.8% for JERS and 88.6% for PALSAR.

We continue to expand the 1997-1998 JERS-based wetlands map into Canada and assemble the 2007 PALSAR-based wetlands map. These products, with the corresponding map of 1997-2007 wetlands changes, offer new insight into the evolution of North American boreal wetlands that is expected to help clarify the relationship between wetlands and climate change.

Presentation Type:   Poster

Poster Session:  Ecosystems Science

NASA TE Funded Awards Represented:

  • Moghaddam, Mahta
    Mapping Wetlands Dynamics for Reducing Uncertainties in the Boreal North American Carbon Budget and Diagnosis and Prognosis of changes in Lake and Wetland Extent on the regional Carbon Balance of northern Eurasia

Close Window