| 2004
sensor
ectral
ysical
Boca
Jse of
ACS
Chen,
band
ntent.
6. pp.
€ into
Essex,
2000.
'eport.
d TM
pheric
OSAT
S.M.,
\[berta
y, 611-
emoto
sé dos
rit.
bok h
curacy
atasets
ed by
Brazil
upport
(IRD),
Sensitivity Analysis of Spectral Indices to Ozone Absorption Using Physical Simulations in a
Forest Environment: Comparative Study between MODIS,
SPOT VEGETATION & AVHRR
A. Bannari and H. Asalhi
Remote Sensing and Geomatics of Environment Laboratory,
Department of Geography, University of Ottawa,
P.O. Box 450, Ottawa (Ontario), Canada KIN 6N5
Tel. (613) 562-5800 (ext. 1042), Fax (613) 562-5145
E-mail: abannari@uottawa.ca
Abstract: Considering a forest cover (Balsam Fir with various densities) we analyze the sensitivity of several vegetation indices to
ozone absorption using physical simulations in the spectral bands of MODIS, VEGETATION and AVHRR sensors. Except for the EVI,
MSR and GARI, the results indicate that all the vegetation indices used in this study normalize the ozone absorption in particular
TSARVI and GEMI. In addition, this study underlines the MODIS sensor contribution to minimize the ozone absorption effect.
1. Introduction
The processes of atmospheric diffusion by aerosols and
molecules, and absorption by gases (ozone, water vapour and
carbon dioxide) disturb terrestrial surface reflectance
measurements acquired by optical satellites. When designing
remote sensing sensors, the spectral bands are selected in
order to avoid as much as possible encroaching on the
absorption bands of atmospheric gases (Deschamps ef al,
1984). However, a relatively weak gas absorption effect
always persists, which results in an attenuation of a few
percentage of the reflectance measured at the sensor. In
general, aerosol and molecule diffusion and water vapour
absorption effects are corrected and removed from images
before extracting information on land use. In the literature, we
find a very rich and varied documentation demonstrating and
explaining the effects of these two atmospheric components
on optical remote sensing data (Deschamps et al, 1984;
Vermot ef al., 1996). Furthermore, investigations showing the
impact of ozone absorption on the data or on the extraction of
the biophysical parameters are rare.
Ozone is an important chemical component of the
atmosphere. It occurs naturally in the atmosphere in small
quantities as an unstable gas, but its levels are increasing at
the surface as a result of human activities. The measured
concentrations increase with altitude and are strong in forest
environments. Ozone concentrations have been increasing in
the atmosphere at a rate of approximately 1 % per year, with a
strong dependence on the amount of sunlight and prone to
strong inter-annual variations. The correction of optical
remote sensing images for ozone should be based on actual
measurements derived, for example, from the Total Ozone
Mapping Spectrometer (TOMS) or other appropriate sensors
(McPeters er al, 1996). However, access and utilization of
these data can be difficult. Using the concentration values
from standard climatic tables with latitudinal and seasonal
dependence is acceptable and is the most useful approach. In
this study, we analyze the sensitivity of several vegetation
indices to ozone absorption considering a forest cover
(Balsam Fir) and using physical simulations. To achieve our
goal, spectroradiometric measurements were acquired above a
forest cover with various densities. All of the measurements
801
were resampled and convolved in the solar-reflective spectral
bands of the MODIS, VEGETATION and AVHRR sensors.
The 6S (Second Simulation of the Satellite Signal of the Solar
spectrum) atmospheric radiative transfer code was used with
variable ozone concentrations and different forest cover
density using the spectral bands of these three sensors.
2. Materials and Methods
2.1 Study Site and Ground Measurements
The study site is a forest area of the Cape Breton Island, News
Scotland, Canada, 46°44’N, 60?40'O. The Western part of the
site is in the National park of Prince Albert and the Eastern
area is part of the provincial forest. It's an experimental zone
used since 1976 for the development of remote sensing
techniques and application in a forest environment. The site is
composed mainly of a stand (Balsam Fir) plantation with
different densities and various stages of maturity. In this
study, spectroradiometric measurements were acquired above
the forest canopy with various densities: 0, 25, 50, 70 and 100
percent. The measurements were carried out with a portable
GER-3700 spectroradiometer (350 to 2500 nm). To take into
account the bi-directional effect of the target reflectance,
which depends both on the illumination angle and the viewing
angle, we carried out measurements around the zenith hour
following a vertical view direction. All of the
spectroradiometric measurements were resampled and
convolved in the solar-reflective spectral bands of the
MODIS, VEGETATION and AVHRR sensors. Percent green
cover was estimated from 35-mm slides obtained
simultaneously from a camera mounted to the
spectroradiometer.
2.2 Simulations conditions
In order to highlight the ozone absorption effects on the
spectral bands (blue, green, red and near infrared) and,
consequently, on the vegetation indices, the transfer radiative
code 6S was used (Vermot et al, 1996). An atmospheric
model US-standard 1962 and an atmosphere without aerosols
were considered. We fixed the view angle in the nadir and the
solar angle at 45° with an azimuth angle at 30° and a view
zenith angle at 0°. Simulations were carried out considering a
