METHODS TO ELIMINATE THE ATMOSPHERIC EFFECTS IN REMOTELY SENSED DATA 
J. Lichtenberger, Cs. Ferencz, D. Hamar, P. Bognär, E.O. Ferencz, G. Molnär, Sz. Päsztor, P. Steinbach, 
B. Szekely, Gy. Tarcsait, G. Timär 
Eötvös University, Space Research Group, Hungary 
(H-1117, Budapest, Päzmäny Peter setäny 2.) 
Commission VII, Working Group 1 
KEY WORDS: Atmospheric correction, Landsat MSS, TM, NOAA AVHRR, IRSIC LISS 
ABSTRACT: 
Quantitative applications of remotely sensed data (e.g. investigation of the canopy of the main crops) requires accurate calibration 
of row satellite data as well as correction of atmospheric and soil moisture effects. Simultaneous use of data from different sensors 
and/or different dates has to be based on physical quantities obtained by correction procedures. The paper presents different 
procedures that eliminate the effects of the atmosphere, discuss their results and applicability, mainly concentrated on a new method 
developed for Landsat TM. This atmospheric correction method is based on a heuristic tasseled cap-like transformation followed 
by additional nonlinear transformations to eliminate the soil moisture effects together with atmospheric effects. A comparative study 
has been done to verify and test the new method that yielded to a conclusion that the usage of the ACABA method provides a fast 
and reliable atmospheric correction. The new method has been applied to two subimages over different parts of Hungary, and has 
provided excellent corrected images. 
INTRODUCTION 
The presence of the atmosphere appears in the satellite data as 
noise. Removal or significant reduction of this noise is 
inevitable in applications based on remotely sensed data. This 
is particularly true for multitemporal and/or multisatellite 
(multisensor) applications, where a common physical quantity 
(e.g. reflectance factor) is the only possible way of 
incorporating heterogenous data into a homogeneous data base 
and the removal of atmospheric effects is necessary for the 
calculation of this physical quantity. 
There are several known methods for atmospheric correction of 
satellite data. The usage of a Radiation Transfer Code) (RTC) 
with measured input (normal optical depth, 6) proved to be very 
accurate (Holm et al. 1989). However, the in-situ measurement 
of à (in sufficient density) is very expensive for recent or 
planned applications and is impossible for historical data. 
The other solution for acquiring atmospheric parameters is the 
satellite image itself. The Dark Object Subtraction (DOS) 
method obtains the atmospheric path radiance (L,) from the 
pixels over deep clear water (Ahern et al. 1977) or from the 
darkest pixel of the image (the latter ones are not necessarily 
clear water pixels) (Chavez 1988). The atmospheric parameter 
obtained by DOS can be used as an input of a RTC. The 
problem of the DOS method is the possible lack of a real "dark 
object" in the scene and the spatial variations of the atmospheric 
conditions over area covered by the given image. The latter one 
means, that the L, value is not constant, but may vary 
significantly from place to place, in contrast to the constant 
value obtained from DOS. 
These problems led to the development of stand-alone 
atmospheric correction methods that do not require input data 
but only the image itself and can be applied to the measured 
pixels. This way, the two main disadvantages of the previously 
mentioned methods can be resolved: the necessity of the 
expensive (and frequently unobtainable) in-situ measurements in 
using RTC and the inadequate spatial resolution of the obtained 
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
L, in the DOS method. The X-STAR algorithm (Lambeck et al. 
1977) was developed for Landsat MSS data. It uses the 
Yellowness Tasseled Cap (TC) index for the calculation of ó. 
A similar approach was applied in the ATMOYE algorithm 
(Ferencz et al. 1993), but the ó was used in the followingg new, 
simplified version of the Radiative Transfer Equation (RTE): 
e 5,4080, 1 
pTi 
L, .=L (E =e 3 (1) 
mi Ai 
where Lni is measured, La; is the surface radiance, L,, is the 
path radiance of a full-haze atmosphere and à, is the normal 
optical depth in the i-th band; £ is a model parameter. 
The ATMOYE method was developed for Landsat MSS and 
(assuming an average reflectance in the visible band based on 
model calculation) for NOAA AVHRR data and was used in a 
pilot project for yield estimation at small and medium scale in 
Hungary (Hamar et al. 1995). Greenness profiles calculated 
from atmospherically corrected MSS and AVHRR data is shown 
in Figure 1 demonstrating the possibility of combination of data 
from two different sensors. 
In the following, we describe an atmospheric correction method 
developed for Landsat Thematic Mapper data over vegetated 
surfaces. It is not the modification of ATMOYE for TM, 
because the Yellowness concept cannot be applied for TM and 
because our results in the above mentioned pilot project (Hamar 
et al. 1995) and other efforts (see e.g. the SAVI by Huete 
1988) have indicated, that the atmospheric correction and the 
effects of soil moisture and soil brightness are not separable and 
they should be tied together. 
THE ABSTRACT TASSELED CAP (ATC) 
TRANSFORMATION 
The basis of the original Tasseled Cap (TC) transformation 
(Kauth and Thomas 1976) is the following: the satellite bands 
787