An improvement of simplified atmospheric correction

Abstract

Recently, application of satellite data increase and use in various fields. Because remote sensing of satellite technique has some advantage, such as fixed spatial resolution and high frequency of observation. In order to using various fields, achievement of accuracy data is very important which can highly influence accuracy of output. Remotely sensed imageries observed by satellites are mostly contaminated by gases (ie. H₂O, O₃, O₂ etc) and aerosol in atmosphere. Because many sensors of Earth remote sensing or meteorological satellites observe with solar illumination which reflected signal from target through atmosphere. Although many multi spectral sensors designed to avoiding atmospheric effect using atmospheric windows, attenuation by atmosphere is indispensableness. In order to estimate accurate surface reflectance, Atmospheric correction of satellite measurements is a major step in multi-spectral sensors. In this study, Simplified Method for the Atmospheric Correction (SMAC) Radiative Transfer Model (RTM) used to retrieve surface reflectance from MODIS (MODerate resolution Imaging Spectrometer) Top Of Atmosphere (TOA) reflectance. Initial SMAC code was based on 5S (Simulation of the Satellite Signal in the Solar Spectrum) RTM and previously yields many coefficients for sensors and satellites. Estimated surface reflectance valuated using comparison with MODIS surface reflectance. Comparison showed that low accuracy values such as, R2 : 0.6196, Root Means Square Error (RMSE) : 0.00775, bias : -0.859. Thus sensitivity test of input parameters and coefficients was conducted to searching error sources. Among the input parameters, Aerosol Optical Depth (AOD) at 550nm is the most influence input parameter. And coefficients about Τ_(p) (average AOD) are more influence than any other coefficients of Τ_(a550) (Aerosol Optical Depth at 550nm) from sensitivity test. Stepwise multiple regression was conducted with testing and remove variable in three stages. A dependent variable is Τ_(p), and independent variables are Τ_(a550), solar zenith angle, viewing zenith angle in stepwise regression. Modified coefficients of Τ_(p) were used to surface reflectance which showed that improve accuracy of surface reflectance. Comparison between MODIS surface reflectance and estimated surface reflectance using modified coefficients appear that R²: 0.827, RMSE : 0.00672, bias : -0.000762.