RADIATIVE TRANSFER CODE BASED ESTIMATION OF ATMOSPHERIC 
PARAMETERS FOR INVERSION OF IMAGING SPECTROMETER MEASURED 
RADIANCE TO APPARENT SURFACE REFLECTANCE 
Robert O. Green 
JPL, California Institute of Technology . Pasadena, CA (USA) 
ABSTRACT 
Surface reflectance is required to quantitatively investigate molecular absorption and particle scattering 
properties of materials on the Earth’s surface. Atmospheric aerosol optical depth, surface pressure and water 
vapor are required to constraint a radiative transfer code for the inversion of measured spectral radiance to 
apparent surface reflectance. In this paper, a suite of algorithms using non linear least squares fitting 
techniques are described that directly estimate these atmospheric parameters from spectral radiance measured 
by the Airborne Visible/Infrared Imaging Spectrometer. The derived atmospheric parameters are used to 
constraint a radiative transfer code for the inversion of the imaging spectrometer radiance to apparent 
reflectance, the derived apparent reflectance is validated with respect to in situ measurement on the same 
target. 
KEY WORDS : Atmospheric parameters, Imaging spectrometer. Radiative transfer code, Reflectance 
1 - INTRODUCTION 
Data were acquired by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) over the Jasper Ridge 
ecological preserve on 6 June 1992. Jasper ridge is located near Stanford University on the southern portion of 
the San Francisco peninsula, California. The AVIRIS data cover a 10 by 11 km area with 20 by 20 m spatial 
resolution. Figure 1 shows a single channel image of the AVIRIS data for Jasper Ridge (top towards the 
northwest). For each AVIRIS spatial element 224 spectral channels are measured from 400 to 2500 nm at 
nominally 10 nm spectral sampling. These data are calibrated to the measured total upwelling spectral radiance 
(Chrien et al., 1990; Green et al., 1990a; Chrien et al., 1993). In figure 2 the calibrated upwelling spectral 
radiance for homogeneous portion of the Stanford polo field is given. Figure 3 shows the radiance spectrum for 
a portion of the Stanford golf course. The shape of these spectra result from the solar irradiance, molecular and 
aerosol scattering of the atmosphere, illumination geometry as well as the reflectance of the surface. Algorithms 
have been developed to directly estimate aerosol scattering, molecular scattering and well mixed gas absorption, 
and water absorption from the measured upwelling spectral radiance. These algorithms use a non linear least 
squares spectral fitting (NLLSSF) algorithm linked to the MQDTRAN radiative transfer code (Berk et al., 
1989). With estimates of these atmospheric parameters, the AVIRIS radiance spectra for each spatial element 
are inverted to apparent surface reflectance. Preliminary validation of this set of algorithms is provided through 
comparison of the inverted reflectance spectrum and a in situ measured spectrum for the polo field. 
2 - AEROSOL OPTICAL DEPTH 
Under low visibility conditions the radiance scattered from atmospheric aerosols may comprise a significant 
portion of the total radiance reaching AVIRIS in the 400 to 700 nm region of the spectrum. A NLLSSF 
algorithm has been developed to estimate the aerosol optical depth directly from the measured radiance in thi s 
spectral region. This algorithm optimizes the fit between the AVIRIS measured radiance and a MODTRAN 
modeled radiance with the aerosol optical depth as the primary fitting parameter. Parameters modeling the 
reflectance magnitude, the spectral reflectance slope, and the surface leaf chlorophyll absorption are also 
included. For the Jasper Ridge data set, the MODTRAN rural aerosol model was selected to initially constraint 
the algorithm. Figure 4 shows the resulting fit between the AVIRIS measured spectrum and the NLLSSF 
spectrum for polo field target. The aerosol optical depth required for this fit is reported in terms of the 
MODTRAN visibility for the rural model. The image of visibility for the complete Jasper Ridge data set shows 
193