Discrimination of vegetation types in alpine sites with ALOS PALSAR-, RADARSAT-2-, and lidar-derived information

Abstract

Natural vegetation monitoring in the alpine mountain range is a priority in the European Union in view of climate change effects. Many potential monitoring tools, based on advanced remote sensing sensors, are still not fully integrated in operational activities, such as those exploiting very high-resolution synthetic aperture radar (SAR) or light detection and ranging (lidar) data. Their testing is important for possible incorporation in routine monitoring and to increase the quantity and quality of environmental information. In this study the potential of ALOS PALSAR and RADARSAT-2 SAR scenes' synergic use for discrimination of different vegetation types was tested in an alpine heterogeneous and fragmented landscape. The integration of a lidar-based canopy height model (CHM) with SAR data was also tested. A SPOT image was used as a benchmark to evaluate the results obtained with different input data. Discrimination of vegetation types was performed with maximum likelihood classification and neural networks. Six tested data combinations obtained more than 85% overall accuracy, and the most complex input which integrates the two SARs with lidar CHM outperformed the result based on SPOT. Neural network algorithms provided the best results. This study highlights the advantages of integrating SAR sensors with lidar CHM for vegetation monitoring in a changing environment.

Acknowledgements

This research was carried out in the framework of the SOFIA project (ESA AO-6280), which provided the RADARSAT-2 SAR scenes. Gaia Vaglio Laurin and Riccardo Valentini are thankful to the ERC Africa GHG Grant (project # 247349) for personal research support. Thanks to Prof. Qi Chen, University of Hawaii at Manoa, for constructive revision of the manuscript.