Hyperspectral data classification and distribution of hydrothermally altered minerals in the Krafla region, Iceland
View/ Open
Janice Hunter MSc Dissertation 2015.pdf (16.76Mb)
Date
15//2/26/1Item status
Restricted AccessAuthor
Hunter, Janice
Metadata
Abstract
Hydrothermally altered minerals indicate potential locations of commercial ore deposits and geothermal energy, so developing techniques to identify such minerals is highly beneficial. Using airborne remotely sensed data means much larger areas can be covered quickly for geological survey and many more samples can be obtained than can be gathered through field work.
In this report hyperspectral analysis is carried out to identify and map the distribution of hydrothermal alteration minerals in the region around the Krafla volcano, Iceland. The mineral distribution is compared to the topography to assess what influence it has on the distribution. The hyperspectral data used was gathered by an AISA Hawk sensor during a NERC ARSF flight over Krafla volcano in September 2008. Following atmospheric correction by the empirical scene-based method, internal average relative reflectance, the data from the shortwave infrared part of the electromagnetic spectrum are analysed using the Spectral Hourglass method to identify hydrothermally altered minerals. Simultaneously collected LiDAR point cloud data is converted to a hill-shaded digital elevation model.
Mineral mapping results are ambiguous, typical alteration minerals have been identified but also minerals unlikely to be in the region. The problem is attributed to the data being too noisy to allow minerals to be accurately identified and mapped using the Spectral Hourglass method. No relationship between mineral distribution and topography is observed, however, a comparison to a geological map suggests the distribution is more related to the rock type, age and geothermal conditions. It is recommended that a more effective model-based approach for atmospheric correction is used to reduce data noise and provide smoother spectral response curves to achieve better matches to the reference data.