Porównanie wyników korekcji atmosferycznej danych satelitarnych CHRIS/Proba przeprowadzonych w oprogramowaniach BEAM/Visat oraz ATCOR

A comparison of output of two absolute atmospheric correction methods (ATCOR by R. Richter, 1996, and an algorithm by L. Guanter et al., 2005, implemented in the BEAM/Visat framework) is presented. Analyses are based on satellite data acquired by CHRIS (Compact High Resolution Imaging Spectrometer) sensor onboard the PROBA (Project for On-Board Autonomy) satellite. For comparison, a set of in situ spectral measurements obtained by the Norwegian NIVA Institute was taken as reference data. The area of study was the Vistula Lagoon in Northern Poland. All analyses presented here are based on comparison of results of atmospheric correction methods with in situ reference data. Alterations between ground and satellite spectral measurements can be caused by changes of humidity or solar zenith angle, as well as fluctuations of water masses, aerosols and air masses, all of which phenomena occur with time passage. In order to minimize the influence of this element, a set of simultaneous ground and satellite measurements was analyzed. Observations were collected on the same day, 18th August 2008. The best atmospheric correction was obtained in ATCOR with a ground model calibration, and the mean relative difference in spectral reflectance between the results obtained with this method and the reference data was 0,18%. The drawback of this method is that it requires results from in situ spectral measurements to reinforce the reflectance derivation, while such data is usually unavailable. Hence, only methods independent of ancillary data are treated as authoritative. In this case, the output of two methods – ATCOR without ground model calibration and an algorithm by L. Guanter et al., (2005) implemented in BEAM/Visat framework – were compared against the reference data. The comparison yields 2,30% and 2,10% reflectance mean difference between ATCOR, an algorithm by L. Guanter et al., (2005) and the reference data, correspondingly. This leads to conclusion that an algorithm by L. Guanter et al., (2005), provided better results in our case.