Geography Reference
In-Depth Information
It is not always necessary to apply an atmospheric correction technique for each
remotely sensed study, since the necessity for that depends on the goals of the
analysis and the expected results or products. For clarification, it is very important
to be applied when a remotely sensed data of a certain region are to be evaluated
over a time period—e.g., over a period of a crop growing—(Liang
2004
).
Atmospheric correction is necessary for classifying a multi-sensor (especially
when integrated for an image classification) or multi-date imagery. It is moreover,
essential for mapping of change detection over a time, since it used to guarantee
that gray values of pixels are comparable in both images in a temporal sequence
(Liang
2004
), since atmospheric effects are one of the error sources in change
detection studies (e.g., Chavez and Mackinnon
1994
; Coppin et al.
2004
).
In general, if a single-date image is used in LULC-classification, it may not
require atmospheric correction as long as the atmospheric effects are consistent
over the whole scene, since their impacts are similarly on the spectral vectors of
training and unknown pixel, and their relative positions in spectral space are
unaffected. However, if the atmospheric conditions varies largely within the study
area (e.g., due to haze, smoke or dust storm), then spatially-dependent correction is
needed (Song et al.
2001
; Schowengerdt
2007
).
A lot of techniques were founded to normalize and, if possible, to correct the
radiometric distortions of the data and the atmospheric effect related to atmosphere
conditions. These include, for example: the simple relative calibration approaches
(e.g., the dark-object subtraction); and the complex approaches (e.g., 6S) (Mark-
ham and Barker
1987
; Canty et al.
2004
). These methods include: (1) Invariant-
Object Methods (Moran et al.
1992
; Chavez
1996
) (2) Histogram Matching
Methods (Richter
1996a
,
b
); (3) Dark-Object Methods (Chavez
1988
; Kaufman
et al.
2000
), which is frequently used; (4) Contrast Reduction Methods (Tanre
et al.
1988
); (5) Cluster Matching Method (Liang et al.
2001
); (6) The MODTRAN-
code (Berk et al.
1998
); and (7) The Second Simulation of the Satellite Signal in
the Solar Spectrum 6S-code (Vermote et al.
1997
).
In the study presented here, the simplified and fast correction approach using
the software program ATCOR-2 (Richter
2011
) was used to atmospherically cor-
rect the images when needed.
The ATCOR-2/ATmospheric CORrection program was developed by the Ger-
man Center for Aerospace (DLR/Deutschen Zentrum für Luft- und Raumfahrt)
(see: Richter
1996b
,
2011
). It provides spatially adaptive and fast algorithm. It
supports the remote sensing sensors LANDSAT-MSS/TM and SPOT from SPOT-
4. It works with a set of functions for atmospheric correction. This set was
developed based on MODTRAN-2 and SENSAT-5 code. ATCOR-2 assumes that
the target objects have an isotropic reflection behavior, where the error effect is
taken in account by the blooming effect. The program uses the comparative
analysis of the measured reflectance of a target object on the sensor with the back-
calculated reflection of the same target, which it derived from models. It is also
implemented in ERDAS IMAGINE (
http://www.geosystems.de
;
http://
www.atcor.de
)
.
The software has been available since 1996/2002, and is a part
Search WWH ::
Custom Search