Improving atmospheric correction methods for aerosol optical thickness retrieval supported by in-situ observations and GIS analysis

Abstract

Atmospheric correction is an important pre-processing step required in many satellite remote sensing applications. Although there are several available atmospheric correction algorithms, there is limited literature available that examines their effectiveness using in-situ measurements from spectroradiometers and sun-photometers to retrieve reflectance and atmospheric properties. The aims of this thesis are to improve the Empirical Line Method of atmospheric correction, apply the improved Empirical Line Method and the Darkest Pixel method to retrieve the aerosol optical thickness (AOT), and to conduct an accuracy assessement of the results using in-situ measurements from spectroradiometers and sun-photometers. Five commonly located surface materials (gray asphalt, black asphalt, concrete, black sand and compacted sand) were identified as meeting the criteria for pseudo-invariant targets and their spectral signatures were measured from January 2010 to May 2011, in order to assess that their reflectance values did not change over time. The empirical line method was improved by using several pseudo-invariant targets of varying spectral characteristics. The Darkest Pixel method utilized the actual reflectance values of the darkest pixel in the selected area of interest. Eleven satellite images were atmospherically corrected using the improved empirical line method and the darkest pixel method with the actual reflectance values. The study found a strong agreement between the AOT values derived from the algorithm developed in this thesis and the in-situ AOT measurements from the sun photometers. A GIS analysis was conducted using the algorithm developed in this to produce thematic maps to identify AOT distribution in the greater area of Limassol. The proposed AOT retrieval algorithm is a fully image based method that explicits global applicability. The innovation of this thesis is the retrieval of AOT by developing an image-based algorithm based on the radiative transfer equation to derive AOT using the reflectance values from the empirical line and darkest pixel method of atmpspheric correction. Novel contributions to this thesis include the employment of pseudo-invariant targets for conducting accuracy assessment of both atmospheric correction methods, the improvement of the empirical line method using five pseudo-invariant targets of varying spectral characteristics and assessing the effects of wet and dry conditions on the pseudo-invariant targets. It is the first time that a GIS analysis has been conducted to automatically retrieve AOT using the radiative transfer equations. The integrated use of field spectroscopy, GIS and remote sensing for atmospheric correction of satellite imagery and AOT retrieval is an innovative contribution to the field of remote sensing.