Using the Wishart maximum likelihood classifier for assessing the potential of TerraSAR-X and ALOS PALSAR data for land cover mapping

Abstract

Land cover is a fundamental variable that impacts and links with many parts of the human and physical environments. However, mapping land cover particularly in the tropical regions is problematic due to persistent cloud cover, large and inaccessible areas, political instability, poor access to mapping data and weak government support to mapping agencies and research institutions. The objective of this study was to assess the potential of the dual-polarised TerraSAR-X (TSX) and quad-polarised L-band ALOS PALSAR satellite data for land cover mapping in the complex terrain of the Bwindi Impenetrable National Park (BINP). Polarimetric analysis of the satellite images was carried out using the Wishart maximum likelihood classification (WMLC) algorithm. A total of nine land cover classes were selected for analysis. For each land cover class, representative samples were extracted and used for land cover classification and accuracy assessment based on the interpretation of the high-resolution IKONOS satellite images as well as the a priori knowledge of the study area. Overall land cover classification accuracies of 86% and 43.9% were obtained using ALOS PALSAR and TSX data, respectively. These results indicate a realistic potential of using ALOS PALSAR data for land cover mapping compared with TSX data.

Keywords:

• TerraSAR-X
• ALOS PALSAR
• WMLC
• land cover
• accuracy

Acknowledgement

The authors would like to thank the German Space Agency (DLR), the United Nations Educational Scientific and Cultural Organisation (UNESCO) and ESA for providing access to TSX and ALOS PALSAR data. The TSX and ALOS PALSAR data were provided through project numbers LAN0559 and CIP7583 respectively.