A New Algorithm for Land-Cover Classification Using PolSAR and InSAR Data and Its Application to Surface Roughness Mapping Along the Gulf Coast

During a flooding event, the ability of the terrain to dissipate water flow energy depends on its land-cover type and the associated surface roughness. In this study, we developed a new land-cover classification algorithm using repeat-pass polarimetric synthetic aperture radar (PolSAR) and interferometric synthetic aperture radar (InSAR) data. Through a two-level hierarchical approach, we classified nine land-cover types with distinct surface roughness coefficients (Manning’s $n$ ). We demonstrated the performance of this algorithm using available L-band ALOS PALSAR scenes acquired between April 2007 and April 2011 over the Houston area. The radar-based surface roughness estimates show a good agreement with those independently derived from NOAA’s 22-class Coastal Change Analysis Program (C-CAP) 2010 land-cover classification data. Our algorithm is robust, and the randomly selected training sets only account for 0.3% of the total multilooked radar pixels (30-m spacing). Furthermore, we were able to accurately map surface roughness over the New Orleans area using available ALOS PALSAR scenes without selecting any new training sets. We note that NOAA’s C-CAP data are currently used for estimating surface roughness in the operational storm-surge models, and a new version is typically released every five to six years. With the launch of the L-band NASA-ISRO Synthetic Aperture Radar (NISAR) mission in the near future, our algorithm can be used to fill the temporal gaps of the existing C-CAP-based surface roughness database and improve the accuracy of near real-time hydrodynamic modeling.