127
TWO-VIEW, SINGLE-CHANNEL ATMOSPHERIC CORRECTION FOR ATSR-2
ictive
the errors
errors
uracy and
odology
:re and
asurement
P.R, NORTH, S.A. BRIGGS, SE. PLUMMER, JJ. SETTLE
British National Space Centre
Remote Sensing Applications Development Unit
Monkswood, Cambridgeshire PEI 7 2LS
U.K.
Tel: 44 4873 381 Fax: 44 4873 277
tvid
ABSTRACT:
This paper presents preliminary work towards utilising the dual look capability of the Second Along-Track
Scanning Radiometer (ATSR-2) to correct for the effects of atmospheric scattering. A novel and efficient
method is presented for simultaneous retrieval of atmospheric aerosol loading and surface spectral reflectance
ion of
using information in each optical channel independently. A simulation of radiative transfer is used to study the
sensitivity of the algorithm to realistic variation in surface cover and atmospheric profile and to sensor noise.
Results show highly accurate retrieval for a near Lambertian surface reflectance. The correction appears robust
suits for
vironment,
to deviation from expected aerosol type. Vegetation indices derived from the correction are largely independent
of atmospheric aerosol loading. Limitations of multi-angle parameter retrieval are discussed.
i in remote
KEY WORDS: Atmospheric correction, aerosol retrieval, ATSR-2, multi-angle imagery
1 - INTRODUCTION
This paper presents preliminary work towards utilising the dual look capability of the ATSR-2 sensor to correct
for the effects of atmospheric scattering. The ATSR-2 instrument is due to be launched on ERS-2 in 1995 and
will succeed the ATSR sensor. The sensor will match the existing ATSR specifications with the addition of
three optical channels of bandwith 20nm centred on 555, 659 and 865 nm. These optical channels will provide
an alternative source of information to AVHRR for regional and global vegetation analysis. The sensor has a
unique dual look capability, measuring radiance at nadir and approximately 50° along track, from pixels with
ground EFOV of the order 1km 2 and 4km 2 respectively. Use of the existing channel centred at 1.6pm is also
considered, referred to subsequently as channel 4.
Atmospheric correction of satellite data is important for surface energy budget estimation and spectral signature
analysis. Current methods for atmospheric correction of optical domain data largely rely on the identification of
objects in an image with an approximately known, low reflectance, or identification of a set of pixels with high
mutual contrast. The use of radiative transfer codes is also possible, although accurate atmospheric correction
relies on knowledge of the aerosol optical thickness at the time of imaging. An extension of the exploitation of
image contrast to multi-angular satellite imagery is suggested by Martonchik and Diner with application to the
MISR instrument (Martonchik 1992). However the method is inappropriate for correction of ATSR-2 imagery
since the ATSR-2 spatial sampling frequency is small compared to the characteristic scale of variation of
tropospheric aerosol loading. A review of current atmospheric correction methods is given by Kaufman
(Kaufman 1989).
The ATSR-2 conical scan allows two measurements of the same area of the earth's surface to be made through
widely differing atmospheric path lengths within a short time interval. The principal of using differential path
lengths to correct for atmospheric distortion has been successfully applied to the retrieval of sea surface
temperature using the ATSR instrument (Llewellyn-Jones 1985). Here we attempt to use the same principal to
retrieve bi-directional reflectance over land. The problem is complicated however by the unpredictable angular
variation of both the surface bi-directional reflectance and the aerosol single scattering albedo and phase
function.
In this paper we consider the case of measurement of the same area of surface from two view directions for a
single waveband. In this case we can retrieve only two parameters to describe the atmospheric content and the
surface bi-directional reflectance. The two parameters used are the absolute hemispherical albedo of the land
surface, approximated as Lambertian, and the atmospheric aerosol loading, assuming a known aerosol type
distribution. A novel method for atmospheric correction is derived which combines radiative transfer modelling
with multi-angle correction. Sensitivity of the correction and derived products to surface bi-directional
reflectance and atmospheric profile is analysed using simulated data.
