Lithological mapping by remote sensing using emittance information in the thermal infrared region

Abstract

The aims of this study were to assess existing techniques, and develop new methods for extracting the geologically significant parameter spectral emittance from passive multispectral scanner data in the Thermal Infrared. Emittance cannot be extracted directly from the radiance data due to the underdetermined nature of the resultant equation set. Quantitative analysis of the Thermal Infrared data also requires the application of calibration and atmospheric correction algorithms. A calibrated and atmospherically corrected Thermal Infrared Multispectral Scanner (TIMS) scene from Halls Creek, Western Australia, was selected as a test area on which to apply the techniques. However, detailed mapping of the Halls Creek area was not an objective of this research. Existing image enhancement techniques namely the Decorrelation-Stretch and Thermal Log Residuals (TLR) as well as the alpha coefficients calculation (a simplification of the TLR method developed during this research) were applied to the TIMS data and each created useful images for delineating boundaries. A new emittance estimating technique, the Alpha Derived Emittance method, developed during this research was applied to the data for comparison with existing algorithms, the Model Emittance and Maximum Temperature methods. The Alpha Derived Emittance method utilised the Wien approximation to the Planck function and solved the underdetermined equation set by assuming the existence of a relationship between the mean and variation of emittance spectra. A Thermal Infrared spectral library was created from a representative collection of Halls Creek lithologies in order to assess the applicability of each emittance estimating method. Analysis of these spectra indicated that each of the methods were equally accurate in estimating emittance for Halls Creek lithologies. Further, each technique estimated the emittance to within 0.02 of emittance for approximately 65% of the lithologies. TIMS image statistics suggested that the Alpha Derived Emittance method most effectively separated the emittance and temperature from the radiance signal. This research indicates that the variability within samples may hinder the ability to discriminate lithologies.