Wetland mapping by using multi-band and multitemporal SAR images: A case study of Hong he National Natural Reserve

Abstract

Wetland ecosystems have many important functions as the regional environment stabilization, natural species protection and ecological resources facility. Investigation and monitoring on wetland urgently need the application of remote sensing. The optional remote sensing technology is sometimes difficult to receive the data due to the bad weather, or in specific light conditions. However, radar remote sensing has a great potential for technical application in wetland mapping with its all-weather, all-time capabilities. And SAR scattering intensity is sensitive to the soil moisture of land cover and land change, therefore it is also helpful to the wetland research. In this paper, multi-band, multi-temporal SAR technique was used for wetland mapping. Unifies the SAR polarization characteristic, the backscattering characteristic of multi-temporal ENVISAT ASAR HH, HV, VV polarization data and ALOS PALSAR HH polarization data of different land use type and different vegetation type in Honghe National Natural Reserve (HNNR) were analyzed. The multi-temporal SAR data was used for vegetation classification research in decision tree classification method. And the authors use actual samples to do accuracy assessment in the way of computing confusion matrix. The result from this study shows that the improvement of water and marsh classification accuracy owes to the summer data, and the classification accuracy can be improved when we used the multitemporal data, HV polarization data in winter is more advantageous for forest map. The recognition rate of water, marsh and forest was relatively high. The classification of grass and bush had confusion. The classification accuracy of multitemporal SAR reached 79.55%.