The analysis of sensibility to the change of the input parameters in the 6S model J. Bojanowski Remote Sensing Department, Institute of Geodesy and Cartography, Warsaw, Poland e-mail: jedrzej.bojanowski@igik.edu.pl Keywords: atmospheric correction, 6S, eta squared ABSTRACT: The input data necessary to correct the atmospheric effect are often of a very difficult access and their precision is quite low. For this reason, it is rather important to know how each of the input parameters influence the atmospheric correction process and the output image. In this study we report the results of the sensibility analysis of the 6S model (Second Simulation of the Satellite Signal in the Solar Spectrum), which is one of the atmospheric correction algorithms, to changes in the input parameters. We have analyzed four input parameters describing atmospheric conditions in which the image was taken. 72.000 atmospheric corrections have been made on the different sets of input parameters. Next, on the basis of the images that were the result of the atmospheric corrections, for each test field and spectral band a regression model has been created and analyzed with eta and partial eta squared measurements. Thus the importance of the input parameters as the function of the wavelength and digital number of the input image were obtained. 1 INTRODUCTION The great complexity of the gasses and aerosols influence on the electromagnetic waves transferring through the atmosphere causes that the atmospheric correction algorithms are very complicated. They need many input parameters describing the state of the atmosphere. This data is not easily accessible and it is not very precise. This is why it is interesting to see to what extent the outcomes of the atmospheric corrections carried out with the parameters set to their default values differ from the corrections that use the actual real-world values measured in the moment and place of the image acquisition. This information is especially important when there is no precise atmospheric data available and it is necessary to use some standard description of the atmospheric conditions. Such a comparison gives also an assessment of an error which could be done while performing the atmospheric correction with some input parameters missing. In the present study we present a sensibility analysis of the atmospheric correction model to the change of the input parameters: gasses concentration, aerosols concentration, ground altitude and visibility. In this analysis we use the 6S model (Second Simulation of the Satellite Signal in the Solar Spectrum) which is freely available. We treat the algorithm as a black box even though its source code is known. To study an effect of changes in the input parameters we employ multivariate statistical analysis techniques to compare 6S correction results for different parameter sets. New Developments and Challenges in Remote Sensing, Z. Bochenek (ed.) ß2007 Millpress, Rotterdam, ISBN 978-90-5966-053-3 551