Mapping Mangrove Species with High Resolution Optical and Polarimetric SAR Satellite Data

Mangrove forests are the most productive and important ecosystems which provide significant ecosystem services, such as carbon sequestration, flood mitigation, water quality protection and a wide range of habitats to support a rich variety of flora and fauna. Remote sensing provides a relatively easy and much less costly approach to monitor the dynamics of mangrove. However, remote sensing of mangrove is challenging due to the nature of mangrove and limitations of spectral and spatial resolutions of remote sensing data. Nowadays, advances in the high spatial and spectral resolutions of sensors now available to ecologists are making the direct remote sensing of certain aspects of biodiversity increasingly feasible. Very high spatial resolution optical data are more and more frequently used in the monitoring of mangrove forest. Additionally, the backscattering information for plants using active microwave remote sensing is also useful for identifying various mangrove species, since microwave is sensitive to the surface roughness and biophysical properties of mangrove forest. Nevertheless, the potential of synthetic aperture radar (SAR) is still under explored and discussion. Moreover, polarimetric SAR (PolSAR) data offer better capability for distinguishing different scattering mechanisms of ground targets, which will greatly help overcome the spectral confusion problems in optical data, and thus may provide complementary information for identifying various mangrove species. In this research, high spatial resolution satellite images from Worldview-3 and Radarsat-2 were employed to classify the mangrove species in the Mai Po wetland of Hong Kong. Spectral and textural features were extracted from the optical images and polarimetric features were extracted from the Radarsat-2 image. Support vector machine was applied to classify the mangrove species with the spectral and textural information from the optical image and polarimetric information from the PolSAR image. Subsequently, all these features were fused to identify the species with a supervised classification process. Statistical data of field survey from the Agriculture, Fisheries and Conservation Department of Hong Kong government was used to validate the results in this study. Experimental results indicated that 1) mangrove species classification with very high resolution satellite images was generally low (less than 80% of the overall accuracy) due to the highly spectral confusion between different mangrove species; 2) polarimetric features did not provide a generally positive result to the accuracy, while it was able to improve the accuracy of some species; 3) the combination of very high resolution optical and polarimetric SAR images is necessary to improve the accuracy of mangrove species mapping.