Monitoring Seasonal Inundation in Arctic Wetlands
Using RADARSAT ScanSAR Imagery

G. P. Livingston, D. G. Williams, and L. A. Morrissey

Abstract

We examine the use of medium resolution Radarsat ScanSAR imagery to monitor seasonal inundation and emergent vegetation in the Siberian Arctic.  An hierarchical multistage classifier based on seasonal SAR returns was used to differentiate open water (without emergent vegetation), inundated wetlands (flooded soils with emergent vegetation), and non-inundated areas.  ScanSAR-based classification accuracies ranged from 84 to 96% as estimated from 1:200,000 scale topographic maps.  Comparison to the 1:1 million scale Digital Chart of the World (DCW) database indicated the DCW underestimated open water and wetlands by 38 and 100% respectively.  Seasonal changes in radar return followed the rapid shift between open water and inundated wetlands early in the season and the gradual decline in surface waters throughout the remainder of the growing season.  The results demonstrate the potential to seasonally monitor the Arctic land surface and to significantly improve existing land cover databases commonly employed in global biogeochemical and hydrological modeling studies.