Browsing by Author "Van der Merwe, Hendrik Naude"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Remote sensing driven lithological discrimination within nappes of the Naukluft Nappe Complex, Namibia
    (Stellenbosch : Stellenbosch University, 2015-04) Van der Merwe, Hendrik Naude; Miller, Jodie A.; Kemp, Jacobus Nicholas; Stellenbosch University. Faculty of Science. Dept. of Earth Sciences.
    ENGLISH ABSTRACT: Geological remote sensing is a powerful tool for lithological discrimination, especially in arid regions with minimal vegetative cover to obscure rock exposures. Commercial multispectral imaging satellites provide a broad spectral range with which to target specific rock types. Landsat ETM+ (7), ASTER, and SPOT 5 multispectral images were acquired and digitally processed: band ratioing, principle components analysis, and maximum likelihood supervised classification. The sensors were evaluated on the ability to discriminate between sedimentary rocks in a structurally complex setting. The study focusses on the formations of the Naukluft Nappe Complex, Namibia. Previous work of the area had to be consulted in order to identify the main target rock types. Dolomite, limestone, quartzite, and shale were determined to make up the majority of rock types in the area. Landsat, ASTER, and SPOT 5 imagery were acquired and pre-processed. Each was subjected to transform techniques: band ratios and PCA. Band ratios were tailored to highlighted target rock types as well as a number of control ratios to ensure the integrity of important ratios. PCA components were inspected to find the most useful ones which were combined into FCCs. Transform results, expert knowledge, and a geological map were consulted to identify training and accuracy samples for the supervised classifications. All three classifications made use of the same set of training and accuracy samples to facilitate useful comparisons. Transform results were promising for Landsat and ASTER images, while SPOT 5 struggled. The limited spectral resolution of SPOT 5 limited its use for identifying target rock types, with the superior spatial resolution contributing very little. Landsat benefitted from good spectral resolution. This allowed for good performance with highlighting limestone and dolomite, while being less successful with shale. Quartzite was a real problem as the spectral resolution of Landsat could not cover this range as well. ASTER, having the highest spectral resolution, could distinguish between all four target rock types. Landsat and ASTER results suffered in areas where formations were relatively thin (smaller than sensor spatial resolution). The supervised classification results were similar to the transforms in that both Landsat and ASTER provided useful results, while SPOT 5 failed to yield definitive results. Accuracy assessment determined that ASTER performed the best at 98.72%. Landsat produced an accuracy of 93.29% while SPOT 5 was 80.17% accuracy. Landsat completely overestimated the amount of quartzite present, while all results classified significant proportions Quaternary sediments as shale. Limestone was well represented in even the poorest results, while dolomite usually struggled in areas where it was in close association with quartzite. Silica yields relatively strong responses in the TIR spectrum which could lead to misclassification of dolomite, which also has strong TIR signatures.