Research > Satellite
Thermal Imanging >
Image Processing of ASTER Data
The
ASTER Sensor
ASTER
is the acronym for Advanced Spaceborne Thermal Emission and Reflection
Radiometer. It is a Japanese multispectral sensor carried on the TERRA
satellite, which was launched on December 18, 1999. TERRA orbits the
Earth with a period of 98.1 minutes at a height of 720 km. The
instrument is described at the following web site: http://asterweb.jpl.nasa.gov/instrument/instrument.htm.
ASTER is a narrow-field of view sensor (the swath angle of the sensor
is ±2.4°), which scans a swath of 60 km on the ground every
16 days. The sensor has nine (9) reflective bands and five (5) bands in
the thermal infrared, providing thus a wealth of spectral information
to map geological surfaces. In the TIR region, ASTER has 12-bit
quantization, a radiometric precision (NEDT) of 0.3 °K, and one
standard deviation calibration uncertainty of 1 °K between 270-340
°K. The ASTER level 1B scenes of Nisyros are registered radiance at
sensor products, produced by the Japanese Ground Segment and delivered
by EROS Data Centre in HDF format. The thermal infrared telescope
operates in "whiskbroom" mode and collects emittance with a 90 m
spatial resolution.
Data
Processing
ATCOR2 (Flat Surface Model)
The
ASTER data were imported into GeomaticaTM
v8.2.1 as 16-bit unsigned integers. The data were first processed using
ATCOR2 assuming a constant emissivity value of 0.98. We assumed a
visibility of 15 km for all passes based on the meteorological data
supplied to us by the Hellenic Meteorological Service (EMY). These data
(Air Temperature, Dew Point, Relative Humidity, Atmospheric Pressure)
were collected at the Kos and Rhodes airports, every hour, at distances
25 and 80 km, respectively. TERRA passed over Nisyros at 09:11:55 GMT
on April 26, 2002 (local time 12:11:55) and at 20:15:08 GMT on June 13,
2002 (local time 23:15:08).
The
ATCOR2 processing results are as in the following:
The day scene (April 26, 2002) was processed using the US standard
atmospheric model (ATCOR3 Manual page 73) that matched better the
meteorological ground data. The maximum value (44 °C) was located
on a sunlit slope of the volcano on the northeastern side, outside the
caldera. The sea surface temperature was 17 °C. The Stefanos crater
surface temperatures were 32-33 °C, however, no thermal anomaly
could be detected, because of solar heating. The first night scene
(June 13, 2002) was processed using the Mid-latitude summer atmospheric
model (ATCOR3 Manual page 74). The maximum value (26 °C) was
recorded inside Stefanos Crater. Minimum values (17 °C) were
recorded on the northwestern flanks of the caldera. The sea-surface had
temperatures ranging between 21-22 °C. The second night scene
(October 26, 2002; Fig 1) was processed using the
Sub-arctic summer atmospheric model (ATCOR3 Manual page 74), which
matched better our meteorological data on the ground. The maximum value
(24 °C) over land was recorded inside Stefanos Crater. Minimum
values (12 °C) were recorded on the western top of the caldera. It
is noted that those values were also calculated on the October 2000
LANDSAT7 ETM+ scene of Nisyros using a different atmospheric model.
Because the same processing software and emissivity value (0.98) on
both epochs was used, it is suggested that the temperature increase of
2 °C indicates thermal instability of the Stefanos Crater region
over this two-year period.
Figure 1: Colour overview of the processed ASTER
scene of the Nissyros volcano acquired on 26 October 2002 GMT. A smooth
colour pattern blue-green-red-purple corresponding to cold-medium-hot
temperatures is shown. White line indicates the coastline and broken
black line indicates the caldera rim. The Stefanos crater rim is
indicated by the small black circle in the south of the caldera. Note
the thermal anomaly (22-23 °C) inside the crater. Grid is UTM35.
The
ATCOR3 (Topographic Influence)
This
work is arranged in three steps. In step A, the ATCOR3 software
produced the topographically corrected temperature image of the
volcano. The ASTER day scene (April 26, 2002) was orthorectified using
the OrthoEngineTM software package.
The orthorectified image was then imported to the ATCOR3 software. The
algorithm removes the topographic effects using a DTM to obtain
information about surface elevation, slope and aspect. A US standard
atmospheric model was used together with 15 km visibility. In step B,
the co-registration of images takes place. To be able to compute the
relative temperature difference per pixel for the April 26, 2002 scene,
the ATCOR2 image was geo-referenced to the ATCOR3 image by the use of a
1st order polynomial transformation. Then mean the error of
registration was less than 1 pixel. In step C, the two images are
subtracted in the following order: ATCOR3 image - ATCOR2 image (Fig. 2). Finally, a low pass, 3x3 filter is applied to
remove outliers. Although the mean difference is close to 0 °C,
large differences exist on both ends of the histogram. This suggests
that large differences are associated with steep topography.
Figure 2: ATCOR difference image of Nissyros
volcano during the ASTER day overpass on the 26-4-2002. Areas with
large positive temperature differences (> 2 °C) are located on
the western part of the caldera (area A), on the northeast part of the
volcano (area B) and along the caldera rim (area C).
Conclusions
The
results show a good comparison with previously processed LANDSAT7 ETM+
thermal data (Ganas & Lagios, 2003). The overall agreement of these
results with previous campaigns in Nissyros, includes the mapping of
the "orographic effect" (Fig. 1), so as it is
proposed that ASTER data may be successfully used to monitor the
thermal status of the volcano. In addition, the comparison of the ASTER
thermal map for October 2002 to the LANDSAT7 thermal map for October
2000 (Ganas & Lagios, 2003) show that surface temperatures inside
Stefanos Crater have risen on average by 2 °C within the period
2000-2002.
The
method of image differencing to investigate the topographic influence
in our day processing results was also used. The computed temperature
differences indicate that
- the
flat assumption model may underestimate temperatures on sun-lit slopes
up to 6 °C and
- there
is no difference over the flat crater region (Fig. 2).
It is
noted that the real differences in surface temperature may be larger
because of the averaging effect of the coarse ASTER pixel (90 m).
REFERENCES:
Ganas A. & Lagios E., 2003. LANDSAT7 thermal imaging of the Nisyros
Volcano.
International Journal of Remote Sensing, 24 (7), 1579-1586.
|