Effects of incidence angles and image combinations on mapping accuracy of surficial materials in the Umiujalik Lake area, Nunavut, using RADARSAT-2 polarimetric and LANDSAT-7 images, and DEM data. Part 1. Nonpolarimetric analysis

Abstract

This study assesses the use of multibeam RADARSAT-2 multipolarized synthetic aperture radar images (hereafter termed “RADARSAT-2 images”), in combination with LANDSAT-7 Enhanced Thematic Mapper (ETM +) and digital elevation model (DEM) data for mapping surficial materials (bedrock, boulders, organic material, sand and gravel, thick till, and thin till) in Arctic Canada. In particular we tested the effects of RADARSAT-2 incidence angles on classification accuracy. This research contributes to the geoscience framework for mineral exploration in Archean to Paleoproterozoic rocks of the northeast Thelon region of Nunavut. The RADARSAT-2 images were acquired in three west-looking descending beam modes (FQ1, FQ12, and FQ20) with increasing respective incidence angles. A maximum likelihood classification (MLC) was applied to different combinations of RADARSAT-2 and LANDSAT-7 ETM+ images, and DEM data. The incidence angle effect on classification overall accuracies is greatest when only the HH polarized images are used, but is reduced when the HV and (or) VV polarized images are added to the classifier. The best MLC overall accuracy of 85.1% is achieved by combining all polarizations and all incidence angles (beam modes) with LANDSAT-7 ETM+ images and DEM data. The influences of variable environmental conditions (moisture and temperature) on mapping accuracy require further research.