Atmospheric Correction for Flat and Rugged Terrain
The ATCOR Models
The task of earth observing sensors is the mapping of surface properties.
However, the surface information is masked, since the signal recorded by
spaceborne and airborne optical sensors consists of several components
and their magnitudes depend on atmospheric conditions. In addition, topographic
effects strongly influence the recorded signal. The objective of an atmospheric/topographic
correction is the elimination of atmospheric and illumination effects to
retrieve physical parameters of the earth's surface, e.g. surface reflectance,
emissivity and temperature.
This information can be used for monitoring, change detection, surface-vegetation
atmosphere transfer (SVAT) modeling, and surface energy balance investigations
for climatic modeling and upscaling. Therefore, atmospheric correction
is an essential part of preprocessing and a prerequisite for the derivation
of certain value added products.
For satellite sensors with a small field-of-view (FOV) the solar and
view geometry can be treated as approximately constant. Since satellite
sensors operate outside the earth's atmosphere a database of atmospheric
correction functions can be compiled for a common altitude based on calculations
with a radiative transfer code. This database was compiled for the satellite
sensors supported by ATCOR. However, airborne sensors operate in a range
of tropospheric altitudes and typically have a large field-of-view. Therefore, a compilation
of the corresponding airborne atmospheric database has to account for the altitude dependence and the
scan angle dependence.
Therefore, the ATCOR models have been split into separate codes optimized
for satellite sensors and airborne sensors.
An integral part of all ATCOR versions is a large database containing the results of
radiative transfer calculations based on the MODTRAN-4 code.
While ATCOR uses the AFRL MODTRAN code to calculate the database of atmospheric look-up tables
(LUT), the correctness of the LUTs is the responsibility of ATCOR.
The satellite ATCOR supports only small to moderate FOV sensors and
consists of separate codes for flat and rugged terrain. It includes a large
database of atmospheric correction functions (look-up-tables computed with
the MODTRAN 4 radiative transfer code) covering a wide range of weather
conditions, sun angles, and ground elevations.
The airborne ATCOR (called ATCOR-4 because of the 4 geometric degrees-of-freedom:
x, y, z, and scan angle) treats small and wide FOV sensors. For computational efficiency,
there are separate modules for flat and rugged terrain imagery.
A large "monochromatic" database of atmospheric correction functions comes with the ATCOR-4 model.
The database was compiled with the MODTRAN-4 code (DISORT, 8 stream option). It comprises the
altitudes 1 km to 10 km and 20 km (1 km increment, with occasional larger gaps).
For a given sensor and range of operating altitudes the corresponding altitude files from the
monochromatic database have to be resampled with the spectral filter functions of all channels.
This is the only preparation before starting with the atmospheric correction. Unlike with previous
ATCOR-4 versions (versions below 3.0), a run of a radiative transfer code is no more necessary.
Examples
The following mosaic shows part of a Landsat-5 TM scene in a flat terrain
recorded on 20 August 1989. The left shows the original scene, the right
is the processed surface reflectance scene employing ATCOR's haze removal
algorithm in the hazy regions.
(65k)
Another example shows part of an Ikonos scene of Dresden, 18 August 2002,
©Space Imaging Europe 2002, demonstrating a successful haze removal.
Ikonos bands 4, 2, 1 (NIR, Green, Blue) are color coded RGB.
(54k)
The next image is a small Landsat-5 TM sub-scene of Makhtesh-Ramon (Israel)
acquired 21 September 1995. The image was processed in the framework of
a joint German-Israeli Foundation project. The top part presents the original
data of TM bands 7/4/1 (coded RGB), the bottom part is surface reflectance
after combined atmospheric/topographic correction. Most of the topographic
features that can be seen in the original image are eliminated in the surface
reflectance image.
(90k)
Starting in 2005, the ATCOR versions include an algorithm for the removal of cloud shadow effects. An example of a Landsat-7 ETM+ sub-scene is shown below. The scene was recorded on 1 May 2000, and covers part of a rural area from Mecklenburg, Germany. The solar zenith angle was 41 degree, the
RGB is displayed with ETM bands 4/5/3 (830/16050/660 nm).
(50k)
Besides atmospheric correction, the surface reflectance cube retrieved from airborne data may be
used to simulate top-of-atmosphere (TOA) radiance scenes with an ATCOR-4 add-on module.
This is a convenient way to provide
realistic data for a spaceborne version of an airborne instrument. Atmospheric parameters
and the solar geometry can be varied and the resulting effect can be observed in the
TOA radiance scene. After calculating the TOA radiance cube from hyperspectral imagery
there is the possibility of synthesizing new hyper- or multi-spectral scenes with approriate
spectral resampling.
Example
The image shows a low altitude AVIRIS scene acquired 3 October 1998 (RGB channels
35, 20, 10 at 682, 556, 457 nm, respectively). The image was first ortho-rectified (with PARGE) to a USGS
digital elevation model resampled from the original 30 m resolution to
the 3.6 m AVIRIS resolution. The elevation of the scene ranges between
500 and 800 m. The lower part of the mosaicked image shows the result after combined
atmospheric & topographic correction. The residual topographic structures
are mainly caused by the 30 m DEM resolution which is not able to follow
rapid terrain changes in the 3.6 m pixel scale of AVIRIS.
(102k)
An example of de-shadowing of an airborne scene is presented next. In this case, the clouds are
located outside of the acquired scene, still the cloud shadow removal works fine.
The scene was recorded by the HyMap sensor on July 12, 2003, near Chinchon, Spain. The solar
zenith and azimuth angles were 18.3 degree and 124.8 degree, respectively. The color coding is
RGB = 878/646/462 nm.
(104k)
Last Updated: 23-Jan-2008
