Strengths and weaknesses of multi-year Envisat ASAR backscatter measurements to map permanent open water bodies at global scale

• Published in: Remote Sensing of Environment
• Main Authors: Santoro, Maurizio, Wegmüller, Urs, Lamarche, Céline, Bontemps, Sophie, Defourny, Pierre, Arino, Olivier
• Other Authors: UCL - SST/ELI/ELIE - Environmental Sciences
• Format: Article in Journal/Newspaper
• Language: English
• Published: Elsevier Inc. 2015
• Subjects: Backscatter; C-band; Envisat ASAR; Land cover; SAR; Water body; Asar; Greenland; Antarc*; Antarctica; Ice Sheet
• Online Access: http://hdl.handle.net/2078.1/173259; https://doi.org/10.1016/j.rse.2015.10.031

The mapping of water bodies at global scale has been undertaken primarily using multi-spectral optical Earth Observation data. Limitations of optical data associatedwith non-uniform and temporally variable spectral signatures suggested investigating alternative approaches towards a more consistent and reliable detection of water bodies. Multi-year (2005-2012) observations of SAR backscattered intensities at moderate resolution from the Envisat Advanced Synthetic Aperture Radar (ASAR) instrument were used in this study to generate an indicator of open permanent water bodies (SAR-WBI) for the year 2010 time frame and for all land surfaces excluding Antarctica and the Greenland ice sheet. A first map of potential water bodies with a spatial resolution of 150 m was obtained with a global detection algorithm based on a set of thresholds applied to multi-temporal metrics of the SAR backscatter (temporal variability, TV, and minimum backscatter, MB). Local refinements were then used to reduce systematic commission and omission errors (4.6% of the total area mapped) due to the similarity of TV and MB over open water bodies and other land surface types primarily in cold and arid environments. The refinement rules are here explained by means of a detailed signature analysis of the SAR backscatter in such environments. The accuracy of the SAR-WBIwas 80%when compared against 2078 manually interpreted footprints with a size of 150 × 150 m2. Omission errors were primarily observed along coast- and shorelines whereas commission errors were associated with (i) ephemeral water bodies, (ii) seasonally inundated areas, and (iii) an incorrect choice of the local refinement.