International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

Contents lists available at ScienceDirect

International Journal of Applied Earth Observation and

Geoinformation
journal homepage: www.elsevier.com/locate/jag

Regional-scale mineral mapping using ASTER VNIR/SWIR data and

validation of reflectance and mineral map products using airborne
hyperspectral CASI/SASI data
Cui Jing a,b , Yan Bokun b,c,∗ , Wang Runsheng b,c , Tian Feng d , Zhao Yingjun d , Liu Dechang d ,
Yang Suming b,c , Shen Wei a
a
China University of Geosciences (Beijing), Beijing 100083, China
b
Institute for Remote Sensing Method, China Aero Geophysical Survey and Remote Sensing Center for Land and Resources, Beijing 100083, China
c
Key Laboratory of Aero Geophysics and Remote Sensing Geology, Ministry of Land and Resources of the People’s Republic of China, Beijing 100083, China
d
Beijing Research Institute of Uranium Geology, Beijing 100029, China

a r t i c l e i n f o a b s t r a c t

Article history: ASTER data have been widely and successfully used in lithological mapping and mineral exploration for
Received 29 September 2013 decades. The errors due to atmospheric water vapor and the characteristics of the photoelectric sensor
Accepted 17 April 2014 could lead to the anomalous characteristics of band 5 and 9 in the ASTER standard reflectivity product.
These anomalies could result in the spectroscopic misidentification of minerals. This study proposed a
Keywords: simple method of atmospheric correction for converting radiance-at-sensor to ground reflectance. The
ASTER
ASTER VNIR/SWIR reflectance correction factor was derived to correct the spectral shape bias resulting
Atmospheric correction
from the radiometric calibration error using airborne hyperspectral CASI SASI data. The ASTER VNIR/SWIR
Geological remote sensing
Seamless minerals mapping
reflectance correction factor was derived to correct the spectral shape bias resulting from the radiometric
calibration error. After applying the reflectance factor to the atmospheric-corrected ASTER L1B data, a
band combination mapping method was proposed for identifying minerals more quickly and accurately.
The results indicate that this method for atmospheric correction of ASTER data produces very good
results in the arid and bare areas. It is still unknown whether the method is suitable for humid and rainy
areas where atmospheric water vapor varies spatially more than in arid and bare areas. After applying the
reflectance factor to the atmospheric-corrected ASTER L1B data, the mean error of all reflectance bands
decreased from 0.0256 to 0.002, and the standard deviation decreased from 0.04251 to 0.0007. The errors
of the 2/1, 5/6 and 9/8 band ratios decreased from 2.38%, 4.102%, and 4.28% to 1.26%, −0.162%, and 0.31%,
respectively. The radiometric calibration error of the ASTER band 1–9 data can lead to the overestimation
of kaolinite. A band index of 2/1 for retrieving Fe3+ cannot produce a reliable Fe3+ distribution map, and
a new index should be developed.
© 2014 Elsevier B.V. All rights reserved.

Introduction radiation (SWIR) and thermal infrared radiation (TIR). The advent
of the ASTER with six channels in the shortwave infrared and
The Advanced Spaceborne Thermal Emission and Reflection five channels in the thermal region allows for the production of
Radiometer (ASTER) remote sensor (Abrams, 2000; Yamaguchi qualitative surface mineral maps of clay minerals (kaolinite, illite),
et al., 1999; Abrams and Hook, 1995), which is installed on the sulfate minerals (alunite), carbonate minerals (calcite, dolomite),
Terra satellite that was launched on December 18, 1999, meas- iron oxides (hematite, goethite) and silica (quartz), which allows
ures reflected and emitted electromagnetic radiation in 14 bands alteration facies (propylitic, argillic, etc.) to be mapped (van der
covering visible-near infrared radiation (VNIR), shortwave infrared Meer et al., 2012). Compared to other multispectral sensors such
as ETM+/TM and ALI, ASTER has a higher spectral resolution and
enhanced mineral mapping capabilities, and it has been used to dis-
criminate and map the minerals containing Fe3+ , Al OH, Mg OH,
∗ Corresponding author at: China Aero Geophysical Survey and Remote Sensing
CO3 2− , and SO4 2− , among others (Aboelkhair et al., 2010; Bedell,
Center for Land and Resources, 31 Xueyuan Road, Haidian District, Beijing 100083,
China. Tel.: +86 10 62060305; fax: +86 10 62060248. 2001; Crósta et al., 2003; Clark et al., 1990; Di Tommaso and
E-mail address: [email protected] (Y. Bokun). Rubinstein, 2007; Ninomiya and Fu, 2003a; Zhang et al., 2007;

http://dx.doi.org/10.1016/j.jag.2014.04.014
0303-2434/© 2014 Elsevier B.V. All rights reserved.
128 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

Rockwell and Hofstra, 2008; Rajendran et al., 2011; Haselwimmer multiplicative factors were calculated to correct the anomalous
et al., 2011). characteristics of ASTER reflectance data.
ASTER data have been both widely and successfully used Finally, we constructed several spectral indices for retriev-
in lithological mapping and mineral exploration for decades, ing minerals from the ASTER reflectance data. The spectral
with most of the applications focused on hydrothermal systems indices were applied in the study area, and three categories
(Kratt et al., 2010; Mars and Rowan, 2010; Pour and Hashim, of minerals, including ferric iron, muscovite/illite and chlo-
2011, 2012a,b; Rockwell and Hofstra, 2008; Rowan et al., 2003; rite/epidote/calcite/dolomite were retrieved and validated using
Yamaguchi and Naito, 2003; Galvão et al., 2005; Rowan et al., CASI SASI hyperspectral data.
2006), barite mineralization systems (Madani and Emam, 2011),
and the evaporate systems (Kavak, 2005; Melendez-Pastor et al.,
2010; Öztan and Süzen, 2011). Hydrothermal alterations of sev- Study area
eral types of ore deposits, which include porphyry deposits
(Di Tommaso and Rubinstein, 2007; Tangestani et al., 2008; The study area, which covers 700 km2 , is located approximately
Pour and Hashim, 2011, 2012a,b), ophiolite complex deposits 15 km north of the town of Liuyuan, in the Beishan mountains
(Tangestani et al., 2011), and epithermal deposits (Pour and area of northwestern Gansu Province (Fig. 1). The region’s lati-
Hashim, 2012a,b), have been mapped. The above case studies tude and longitude ranges are 41◦ 08 32.92 N to 41◦ 15 32.34“N and
demonstrated that it is valuable to retrieve the hydrothermal alter- 95◦ 8 32.29 E to 95◦ 46 30.46 E, respectively. The surface of this
ation using the ASTER data within ore deposit mineralization zones. study area is arid and bare stone desert, with an average annual
Most of the above studies used a single ASTER scene. However, temperature of 8.8 ◦ C and an annual rainfall of 45.7 mm, with the
there is a large demand for fast regional-scale mineral mapping most rainfall occurring in July and August. Surface runoff does not
using multiple scenes. Two problems must to be solved before regularly occur in the area, except for flooding during the rainy
retrieving hydrothermal alterations using multiple ASTER data season.
scenes. The tectonic position of study area is in the intracontinental
First, it is necessary to develop a simple method of atmospheric rift zone of Yujingzi and Liuyuan, located in the southern mar-
correction for converting radiance-at-sensor to ground reflectance. gin of the Beishan epicontinental active belt, which is between the
Although some mineral mapping methods have been developed by tarim and sino-korean plates. The rift zone is a polymetallic met-
processing radiance-at-sensor data other than ground reflectance allogenic belt of copper, gold, silver, iron, lead and zinc. The study
data (Hewson et al., 2005; Ninomiya, 2003b, 2004; Rowan and area has undergone a very complex tectonic evolution, and strati-
Mars, 2003; Aboelkhair et al., 2010; Matar and Bamousa, 2013), graphic sequences from the Ordovician through the Quaternary
a reliable ASTER ground reflectance product is very important for were deformed.
retrieving minerals accurately; spectral reflectance deviations from In the field, the mineralization and alteration are widely dis-
correct spectral reflectance data result in false absorption features tributed. The primary alterations are carbonation, chloritization,
and distortion of diagnostic signatures, which results in the spec- epdidotization, ferritization, sericitization and silicification.
troscopic misidentification of minerals (Mars and Rowan, 2010).
An ASTER standard reflectivity product (Level-2B) was produced
on the basis of atmospheric parameters such as water vapor, which Data and methods
were calculated using MODIS data (Gao and Kaufman, 1998; Mars
and Rowan, 2010). Although the Level-2B data are available from The two ASTER scenes used in this study were acquired on
the EROS Data Center (EDC) in the USA and the Earth Remote September 29th, 2002 (ASTER L1B-012) and on June 30th, 2000
Sensing Data Analysis Center in Japan (ERSDAC), the data prod- (ASTER L1B-011) (Fig. 1). Twelve airborne imaging spectrometer
uct is produced using non-concurrently acquired MODIS water CASI SASI data scenes acquired on September 8th, 2010 were also
vapor data for atmospheric correction. This algorithm is too compli- used in this study (Fig. 1).
cated for most users, and the anomalously high band 9 reflectance The CASI SASI sensors, which were produced by the Cana-
is partly caused by an atmospheric water vapor error (Mars and dian company ITRES, include CASI-1500 and SASI-600 sensors. The
Rowan, 2010), which causes problems in generating seamless CASI-1500 and SASI-600 scanners acquired spectrums in the wave-
mosaics of mineral maps. In this paper, we propose an alternative length ranges of 380–1050 nm and 950–2450 nm, with 36 and 101
method of atmospheric correction for the ASTER VNIR/SWIR data bands, respectively. The scanners were flown over the study area on
using a few ground spectra, based on the hypothesis that the atmo- September 8th, 2010, between 13:00 and 16:30, and twelve lines
spheric water vapor in the spatial range of a single ASTER dataset of data were acquired during the overpasses. The survey was flown
is constant. at a height of 2 km. The spatial resolution of CASI data is 0.9 m by
Second, it is very important to quantitatively assess the ASTER 0.9 m, and that of the SASI is 2.25 m by 2.25 m.
reflectance product, including Level-2B and our product. Most pre- The ground spectra were measured under cloudless conditions
vious works compared ASTER reflectance ground spectra taken between 12:00 and 16:00 in this area, using an ASD Field Spec
in situ or qualitatively compared hyperspectral data (Rowan and Pro spectrometer (Analytical Spectral Devices, Boulder, Co., USA).
Mars, 2003; Mars and Rowan, 2006). Mars and Rowan (2010) This spectrometer was fitted with 10◦ field of view fiber optics,
quantitatively compared ASTER reflectance spectra to spectra from operated in the 350–2500 nm spectral regions with sampling inter-
field samples and ASTER-resampled AVIRIS reflectance data. Their vals of 1 nm. Surface reflectance measurements were taken at
work revealed the anomalous characteristics of band 5 and 9 20 sites over each plot, and for every measurement, the ASD
and demonstrated that the anomalous data would cause spec- spectrometer was kept stationary and the average of 100 consec-
troscopic misidentification of minerals. The radiance correction utively acquired spectra was recorded to reduce the noise level.
factors, which were computed by Biggar et al. (2005), corrected the We used the average of 20 sites acquired spectra as the measured
low band 5 reflectance anomaly and led to a high band 5 reflectance spectra.
anomaly for minerals with strong band 5 absorption, such as alunite The purpose of this study was to develop a new method for
(Mars and Rowan, 2010). In this paper, ASTER reflectance prod- atmospheric correction and to calculate the vicarious calibration
ucts were quantitatively assessed on the basis of ground spectra factor for ASTER band 1–9 data to reduce radiometric errors in the
taken in situ and airborne CASI SASI hyperspectral data, and the ASTER data. The flow chart for this study is shown in Fig. 2.
 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141 129

Fig. 1. Location map of study area. Sixty-three cross points were used for the calculation of spectral reflectance correct factor. Fifty-five purple circle points were used for
the spectral reflectance validation. The green points were used for the CASI/SASI spectral reflectance validation. (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)

Spectral construction and assessment Following a spatial resample, the SASI data were geometrically
corrected to register to the CASI data by picking ground control
Spectral construction of the CASI SASI data and validation points (GCP) and using a second-degree polynomial warp with
nearest neighbor resampling. The RMS error of registration is lower
The CASI data were spatially resampled to 2.25 m by 2.25 m, than 2 pixels. Then, the CASI 380–969 nm bands were layered
which is equal to the SASI data, using the nearest neighbor method. with the SASI 980-2450 nm bands. The layered CASI SASI data

Fig. 2. Data processing flow chart.

130 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

Fig. 3. Comparison between CASI SASI reflectance spectra and in situ field spectra. The distribution of these points is shown in Fig. 1.

were processed using FLAASH atmospheric correction software to Furthermore, we quantitatively validated the airborne data on
produce the reflectance data. There are several atmospheric cor- the basis of spectral features such as absorption wavelength, depth,
rection algorithms for retrieving surface reflectance from imaging area and symmetry (Fig. 4). There exist two main spectral features
spectrometer data, such as High-accuracy ATmospheric Correc- in the SWIR wavelength range. The first is from 2135 to 2285 nm
tion for Hyperspectral Data (HATCH) (Qu et al., 2003), Atmosphere at an absorption wavelength near 2210 nm, and the second is from
CORrection Now (ACORN) (Kruse, 2004), Fast Lineof-sight Atmo- 2285 to 2390 nm at an absorption wavelength near 2335 nm. The
spheric Analysis of Spectral Hypercubes (FLAASH) (Adler-Golden former is caused by minerals containing Al-OH and SO4 2− , such
et al., 1999), Imaging Spectrometer Data Analysis System (ISDAS) as muscovite, kaolinite and alunite. The latter is from caused by
(Staenz et al., 1998), and a series of Atmospheric and Topo- minerals containing Mg-OH and CO3 2− , such as cholorite, epidote,
graphic Correction (ATCOR) codes (Richter, 1996, 1998; Ritcher calcite and dolomite.
and Schlaepfer, 2002). FLAASH is a MODTRAN-based atmospheric ASD in situ measured spectra were used to assess the CASI SASI
correction software developed jointly by the Air Force Phillips Lab- reflectance product. The CASI SASI reflectance was subtracted from
oratory, Hamscon AFB and Spectral Sciences, Inc. (Adler-Golden the ASD measured data, and the result represented the value dif-
et al., 1999). Comparisons of atmospheric correction using ATREM, ference. The relative error, which was used to assess the spectral
ACORN and FLAASH (Griffin and Burke, 2003; Kruse, 2004) have accuracy, was determined using the value differences that were
revealed that retrieved reflectances from the three algorithms are compiled from each band value of the CASI/SASI data divided by
quite similar. FLAASH software retrieved atmospheric water vapor the same band value of the ASD measured data. For the CASI SASI
and optical depth of aerosol pixel-by-pixel using the methods of spectral range (380–2450 nm), the relative error was under 2%. For
Gao et al. (1993, 2009) and Kaufmann et al. (1997), and the atmo- the visible-near infrared range (380–1300 nm), the value difference
spheric model was developed for the sub-arctic summer, which range was from −0.03994 to 0.00035 with a standard deviation
was selected on the basis of the latitude of the study area and the of 0.00801; for the shortwave infrared range (2000–2450 nm), the
date and time of data acquisition. The 1135 nm wavelength was value difference range is from −0.00423 to 0.00938 with a standard
selected to retrieve atmospheric water vapor. deviation of 0.00360.
To ensure to the accuracy of airborne data, we quantitatively
compared the airborne data to the field spectra (Fig. 3). The spectral Spectral construction of the ASTER data
angle mapper (SAM) scores of the image and field spectra are all
above 0.8, which indicates a good match between the large-scale The simulated transmission of atmospheric water vapor for six
spectral shape of the image and that of the field spectra. standard atmospheric models, using the MODTRAN 4 atmospheric
 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141 131

Fig. 4. Comparison of spectral absorption features between the CASI SASI image and spectra measured in situ. The closer the points are to the 1:1 line, the closer the CASI SASI
results are to the ASD measured spectra.

radiative transfer model, indicates that the transmission of bands 8 pixels, the effect is significant. SWIR crosstalk correction software
and 9 of the ASTER data are mostly affected by atmospheric water was developed by Iwasaki et al. (2001) and Iwasaki and Tonooka
vapor (Hewson et al., 2005). Band 9 is more sensitive to atmospheric (2005) and is used in this paper. The second step is layer stacking of
water vapor than band 8. We can deduce from Hewson’s work that bands 1–3 and bands 4–9. The band 1–3 data were spatially resam-
the difference between the atmospheric water vapor value input pled to 30 m by 30 m, which is equal to the band 4–9 data, using
into a radiative transfer model and the true value (e.g., error of the nearest neighbor method before band layer stacking.
atmospheric water vapor) would lead to an error of absorption
depth in band 8. The simulation results using MODTRAN 4 indicated Validation and correction of the ASTER reflectance data
that there exists a linear relationship between the error of atmo-
spheric water vapor (EAWV) and the error of band 8 absorption To compare the ASTER reflectance data to the integrated
depth (EAD) for six standard atmospheric models (Fig. 5). CASI SASI data, the CASI SASI data were georegistered to the ASTER
Hypothesizing that the atmospheric water vapor in the spatial data at 30 m resolution using Delaunay triangulation warping with
range of one single ASTER datum is constant, the water vapor value nearest neighbor resampling (De Berg et al., 2008). One hundred
could be calculated using the formulas in Fig. 5, if the ground spectra ground control points were selected, and the registration RMS
of at least one site is measured in the field. The calculation steps error was 1.2 pixels. The CASI SASI data were spectrally resam-
include (1) calculating both the ASTER reflectance using a standard pled to ASTER VNIR-SWIR bandpasses using an ASTER spectral
atmospheric model and the EAD, which is the difference between response function. Sixty-three points were chosen for the com-
the image and ground measured spectra, and (2) calculating the parisons between the CASI SASI and ASTER data (cross in Fig. 1).
EAWV and atmospheric water vapor using the formulas in Fig. 5. The spectral shape and features, rather than reflectivity value, are
For multiple images in one study area, if the atmospheric water used to retrieve minerals. Therefore, all of the selected spectra were
vapor of one image (known image) is determined, the water vapor normalized before comparisons were made. The normalization was
of other images that overlap with the known image could also be performed for ASTER-resampled CASI SASI data and ASTER data by
determined using the above method with the overlapping known dividing all of the band reflectances by a reference band reflectance,
image spectra as ground true spectra. using a formula similar to Eq. (1). Band 4 was chosen as the refer-
Before calculating the atmospheric water vapor, two main pre- ence because it is less affected by atmosphere than the other bands.
processing steps must be performed. The first is ASTER SWIR We computed the difference in the normalized reflectance spectra
crosstalk removal. SWIR crosstalk is an additive error of radiance of the ASTER-resampled CASI SASI data and ASTER data according
due to the leakage of photons from one detector element to another to Eq. (2):
(Iwasaki et al., 2001; Iwasaki and Tonooka, 2005). All SWIR bands
are affected by cross-detector leakage, and this leakage is most pro- bi
bi = (1)
nounced for bands 4, 5 and 9. For dark pixels adjacent to bright b4
132 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

Fig. 5. Correlation between EAD and EAWV based on different atmosphere models. EAD is the error of the band 8 absorption depth. EAWV is the error of atmospheric water
vapor.

band 7 and band 8. The band index of (b5 + b7 )/b6 is usually used
for estimating the abundance of Al OH containing minerals, and
the index of (b7 + b9 )/b8 is for minerals that contain Mg OH. We
used the two band indexes to qualitatively compare ASTER reflec-
tivity data obtained using our method (ASTER L1B) with the ASTER
standard reflectivity product (ASTER L2B) (Fig. 7). Fig. 7 presents
the reflectance data mosaicked with data for two images without
image matching for neighboring scenes, and the band indexes are
shown in the top left corners of the figures. The left pictures in Fig. 7
are the results using the Flaash module in ENVI software with the
default water vapor value, and the middle and right pictures in
Fig. 7 are the results using our method and the ASTER L2B results,
respectively. The comparison in Fig. 7 shows that our method is the
best for producing seamless regional-scale map products.
On the assumption that the above atmospheric correction elimi-
nates atmospheric effects completely, the mean difference in Fig. 6
is equal or close to zero if the ASTER data contain no radiomet-
ric calibration errors. The mean differences of band 5 and 6 are
Fig. 6. The ASTER Reflectance difference from the ASTER-resampled CASI SASI data.
ASTER L1B is the ASTER reflectivity data obtained from the ASTER L1B data using lower than zero, which means the anomalously low reflectance of
our method, and ASTER L2B is the ASTER standard reflectivity product (Level-2B). these two bands can be attributed to radiometric calibration error.
The anomaly of band 5 has been found in previous works (Mars
and Rowan, 2006; Rowan et al., 2003; Rowan and Mars, 2003). The
Difference = CASI/SASI(bi ) − ASTER L1B(bi ) (2)
mean difference of band 8 is higher than that of band 9 (Fig. 8),
where i is the band number of the ASTER bands, bi is the original which causes band 9 to be low. However, previous works (Mars
reflectivity value of band i, b4 is the reflectivity value of band 4, bi and Rowan, 2006; Rowan et al., 2003; Rowan and Mars, 2003) have
is the normalized reflectivity value of band i, CASI/SASI(bi ) is the bi shown that the ASTER band 9 reflectance is often too high. We
of the ASTER-resampled CASI SASI data, and ASTER L1B(bi ) is the deduce that the elevated values of band 9 observed in previous
bi ASTER data. The results (Fig. 6) indicate that our method pro- works are due to atmospheric effects, because this band is sensitive
duced better results than the ASTER standard products, except for to atmospheric water vapor (Hewson et al., 2005).
 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141 133

Fig. 7. Comparison of band math results between the ASTER L1B-derived data and the ASTER L2B data.

Fig. 8. (a) Comparison of reflectance between corrected spectra and uncorrected spectra. The spectra are point k in Fig. 1; (b) comparison of the difference in reflectance
between the corrected spectra and uncorrected spectra.

We divided the 63 normalized CASI SASI spectra by the nor- the ASTER data by our factors and chose 55 other points (pur-
malized ASTER reflectance and took the average value of the ple circle points in Fig. 1) to validate the calibration result of
ratio as the multiplicative factors (Table 1) to eliminate the the calculated factors. The ASTER data multiplied by our fac-
radiometric calibration error, and the positions of 63 normal- tors are referred to as “corrected ASTER data” in the following
ized CASI SASI spectra are shown in Fig. 1. Then, we multiplied text.

Table 1
Reflectance correction factor for ASTER data.

Band 1 Band 2 Band 3 Band4 4 Band 5 Band 6 Band 7 Band 8 Band 9

0.96328 0.968986 0.959134 1 1.102168 1.05494 0.981149 0.960269 0.999673

134 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

Table 2
Difference values for the ASTER VNIR-SWIR band ratios.

Uncorrected This study factor corrected Biggar factor corrected

Diff. ratio b2 /b1 Average 0.028301 0.014384

Percentage 2.38% 1.26%
Diff. ratio b5 /b6 Average 0.043576 −0.00132 −0.042
Percentage 4.102% −0.162% −3.785%
Diff. ratio b9 /b8 Average 0.04625 0.00312 0.096425
Percentage 4.28% 0.31% 9.01%

We calculated the difference of normalized reflectance spectra Minerals mapping

of the ASTER-resampled CASI SASI data and the corrected ASTER
data according to Eq. (2). We found that the difference errors Minerals mapping from the ASTER data
decreased significantly when the ASTER data were multiplied by
our factors (Fig. 8). The mean error of all bands deceased from Principal components analysis (PCA) transformation and band
0.0256 to 0.002, and the standard deviation decreased from 0.04251 ratio analysis are two main methods that could be used for multi-
to 0.0007. spectral remote sensing data to map mineral distribution. When
Biggar et al. (2005) used radiance correction factors to produce the PCA method (Crósta et al., 2003) is performed on multiple
the RefL1b products, which is a new reflectance dataset produced data scenes, it is difficult to mosaic the principal components
using ASTER level 1B data, crosstalk correction and radiance correc- because the eigenvalues of two adjacent scenes of data are dif-
tion factors (Mars and Rowan, 2010). We also used their factors with ferent. Therefore, although the PCA method has been applied
our ASTER level 1B data (Fig. 8). The results corrected using Biggar’s successfully in many study cases (Pour and Hashim, 2012a,b),
factors have larger absolute difference than our results, which is we think the PCA method is not suitable for regional-scale min-
due to a high band 4. The difference of band 5 is larger than band eral mapping, and we therefore used the band ratio method in
6 in the corrected spectrum using Biggar’s factors, which means this study. A example of a comprehensive review of mineral
Biggar’s factors will produce anomalously high band 5 reflectance, band ratio indices of the ASTER data is the ‘ASTER Mineral Index
as observed by Mars and Rowan (2010). Meanwhile, we found that Processing Manual’ compiled by Aleks Kalinowski and Simon Oliver
Biggar’s factors will produce anomalously low band 9 reflectance, (http://www.ga.gov.au/image cache/GA7833.pdf).
which is interpreted as atmospheric correction error by Mars and The laboratory spectra of common minerals are shown in
Rowan (2010). Fig. 10. Intense Fe3+ absorption features were found near 0.55 ␮m
To further assess the above results, we used band ratios of and 0.9 ␮m, with a peak near 0.75 ␮m, which corresponds to
2/1, 5/6 and 9/8 to quantify relative variations in spectral shape ASTER band 1 and band 2 (Bedini, 2011). On the basis of these
of the ASTER reflectance data. The band ratio of 2/1 was evalu- absorption features, b2 /b1 was used to estimate the abundance
ated because those band positions define the 0.5 ␮m, 0.66 ␮m and of the ferric iron. Al OH minerals (e.g., kaolinite and muscovite)
0.9 ␮m absorption features of Fe3+ (Bedini, 2011; Hunt, 1977; Hunt exhibit an absorption feature near 2.2 ␮m, corresponding to ASTER
and Ashley, 1979; Mars and Rowan, 2010). The 5/6 band ratio was band 6. The Al OH abundance was estimated using the ASTER
evaluated because those band positions define the 2.165 ␮m and band combination, (b5 + b7 )/b6 (Rowan and Mars, 2003). How-
2.2 ␮m Al OH absorption feature, and the 9/8 band ratio was eval- ever, the absorption features of Muscovite/Illite are different
uated because the band positions define the 2.31–2.33 ␮m features from that of Kaolinite and alunite. Kaolinite exhibits an absorp-
of CO3 2− and Fe, Mg OH (Mars and Rowan, 2010). The average ratio tion feature at band 6 along with a subtle shoulder at band
difference can be calculated using Eq. (3) (Table 2): 5. The spectral index (0.04023 × b4 + 0.95977 × b6 − b5 ) can be
used to identify the weak absorption at band 5. If the spectral
index (0.04023 × b4 + 0.95977 × b6 − b5 ) is greater than zero, then

N
CASI SASIn (bi /bj ) − ASTER L1Bn (bi /bj )
(b5 + b7 )/b6 can be used to estimate the abundance of Kaolinite;
Diff ¯ratio = (3) otherwise, (b5 + b7 )/b6 can be used to represent the abundance of
N
n=1 Muscovite/Illite. The spectral index (0.04023 × b4 + 0.95977 × b6 )
represents the reflectance of band 5 by using linear interpolation
with the reflectance values of band 4 and band 6, which was calcu-
where n is the point number, i and j are the band numbers of the lated using Eq. (4) (Fig. 9):
ASTER bands, Diff ¯ratio is the average ratio difference, bi /bj is com- w5 − w4
piled from the band i and band j reflectance, CASI SASIn (bi /bj ) is b5 = b4 − (b4 − b6 ) (4)
w6 − w4
the ASTER-resampled CASI SASI band ratio, CASI SASIn (bi /bj ) is the
ASTER L1B-derived band ratio, and N is the number of points. where w4 , w5 and w6 are the wavelengths of band 4, band 5 and
The average ratio differences of b2 /b1 , b5 /b6 and b9 /b8 were band 6, respectively; the values of w4 , w5 and w6 are 1.165 ␮m,
calculated, and the average differences of above three spectral 2.167 ␮m and 2.209 ␮m, respectively; and b4 and b6 are the
indexes were 0.014384, −0.00132 and 0.00312, respectively. How- reflectance values of band 4 and band 6, respectively. In a similar
ever, by applying Biggar’s factors to the ASTER data, the average way, the abundances of the Mg OH (e.g., chlorite and epidote) and
ratio differences of b5 /b6 and b9 /b8 were −0.042 and 0.096425, carbonate (e.g., calcite and dolomite) groups of minerals were esti-
respectively. This result indicated that our result was more accurate mated using (b6 + b9 )/b8 , based on their absorption feature in the
than Biggar’s in this study area. This also indicated that the band spectral range of 2.33–2.35 ␮m (band 8). In total, five mineral iden-
2 and band 9 reflectances were anomalously low compared to the tification indices were used for mineral mapping (Table 3). These
ASTER-resampled CASI SASI data, and the band 5 reflectance was indices are based on the spectra shape of USGS laboratory rather
anomalously high compared to the ASTER-resampled CASI SASI than the absolute value. When we converted the spectral indexes
data. into classified mineral products, the thresholds for “If” terms did
 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141 135

Fig. 9. Spectra of several minerals in the USGS mineral spectral library. b5 is the reflectance of band 5 obtained via linear interpolation of the reflectance values of band 4
and band 6.

Table 3 data using the Minimum Noise Fraction transformation method

Mineral index for ASTER mineral mapping.
(MNF) to determine the inherent dimensionality of the image
Mineral feature Mineral index (this study) Mineral index data and reduce the computational requirements for subsequent
(previous studya ) processing; (2) obtaining the end-member spectra using the Pixel
Ferric iron b2 /b1 b2 /b1 Purity Index (PPI); and (3) mapping the mineral distributions
Kaolinite If (b5 > b6 ) & (b7 > b6 ) and b7 /b5 , (b4 + b6 )/b5 using the Mixture Tuned Matched Filtering (MTMF) technique.
((0.04023 × b4 + 0.95977 × b6 ) > b5 ) The MNF, essentially two cascaded Principal Components Anal-
then (b5 + b7 )/b6
yses (PCA), orders transformed data on the basis of the signal
Muscovite/illite If (b5 > b6 ) & (b7 > b6 ) and (b5 + b7 )/b6
((0.04023 × b4 + 0.95977 × b6 ) < b5 ) to noise ratio rather than variance (Green et al., 1998). The first
then (b5 + b7 )/b6 transformation, based on an estimated noise covariance matrix,
Alunite If (b4 > b5 ) & (b6 > b5 ) (b4 + b6 )/b5 de-correlates and scales noise in the data. The noise has unit
Then (b4 + b6 )/5
variance and no band-to-band correlations in the resulting trans-
Chlorite/epidote/ If (b7 > b8 ) & (b9 > b8 ) (b7 + b9 )/b8
calcite/dolomite then (b7 + b9 )/b8 formed data. The second step is a standard Principal Components
a
transformation of the noise-whitened data. The PPI method iden-
http://www.ga.gov.au/image cache/GA7833.pdf.
tifies the extremes of the dataset projected in n-dimensional
space and takes the spectra of spectrally pure pixels as the
not need to be set, but the thresholds for “Then” terms did need to end-member spectra (Gleeson et al., 2010). MTMF is based on
be set. The thresholds were set by changing them until the spectra an understanding of the limitations of Matched Filtering (MF)
shape of that class was consistent with the spectral laboratory. and Linear Spectral Unmixing and was developed by Boardman
The atmospheric correction and mineral mapping results were (1998). This method retains the strengths of the MF and Lin-
mainly evaluated using the airborne data, as no fieldwork was con- ear Unmixing techniques, but the inclusion of a feasibility index
ducted outside the area using airborne data. Therefore, the mineral adds ‘weight’ to the output for each end member, leading to an
mapping results from ASTER data are shown only for the extent of optimized technique that has the ability to detect subtle changes
airborne data (Fig. 10). in spectra to highlight the most feasible areas (Bishop et al.,
2011).
Minerals mapping from the integrated CASI SASI data For this study, the CASI SASI bands covering the 350–1300 nm
and 2000–2500 nm wavelengths were selected for processing to
The methods for the analysis of hyperspectral data were retrieve end-member spectra (Fig. 11). Eight end-members, includ-
developed by AIG, and these approaches were implemented and ing muscovite, illite, chlorite, epidote, calcite, dolomite, hematite
documented within ENVI software (Kruse and Boardman, 2000). In and geotite, were extracted. These end-members were renamed
this study, the mineral mapping step, using integrated CASI SASI through comparison with the USGS spectral library. The mineral
reflectance data, consists of three parts: (1) transforming the mapping result is shown in Fig. 12.
136 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

Fig. 10. Mineral mapping result from ASTER reflectance data using mineral indices. The ASTER reflectance data were derived from ASTER L1B data using our method.

The validation of mineral maps ASTER data result (Fig. 10). The validation sites were put in the
areas that were easily accessed and where the selected minerals
Hyperspectral data have been called a “virtual verification”, due are present. The mineral map of the CASI SASI data was vali-
to their high spectral resolution (King and Clark, 2000). The min- dated first, and the validation results were tabulated in a confusion
eral map (Fig. 12) of the CASI SASI data was used to validate the matrix (Table 4). The field pictures and ground spectra of some

Fig. 11. Endmember spectra from a CASI SASI image for MTMF processing.
 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141 137

Fig. 12. Mineral mapping result from CASI SASI data using MTMF method.

Table 4
Validation of the mineral mapping result.

CASI SASI image processing

M/I M/I + Ch/E/C/D Ch/E/C/D None Row total

XRD analysis M/I 5 5

M/I + Ch/E/C/D 2 17 1 20
Ch/E/C/D 3 1 4
Column total 7 17 4 1 29
Overall agreement = 25/29 = 86.2%

M: muscovite; I: illite; Ch: chlorite; E: epidote; C: calcite; D: dolomite.

Fig. 13. Field photos and the ASD measured spectra of the verification points.
138 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

Fig. 14. Fe3+ validation. The hematite and geotite spectra are from the USGS minerals spectral library, and ASTER L1B and CASI SASI are from the image. The ASTER L1B
reflectance was derived from ASTER L1B data using our method.

validation sites are shown in Fig. 13. The comparison shows an CASI SASI and ASTER data from three sites (Figs. 13 and 14). The
agreement of 86.2% between the mapping results and XRD analysis ASTER data did not indicate Fe3+ at site a, but the CASI SASI data
results. did. The ASTER data identified Fe3+ at sites b and c, where no Fe3+
The mineral distribution maps of the ASTER data and the exists in the CASI SASI map. The spectral check demonstrated that
CASI SASI data are shown in Figs. 10 and 12, respectively; the pat- the band index of 2/1 for retrieving Fe3+ cannot produce a reli-
terns of the image map units are similar, except for ferric iron. The able Fe3+ distribution map. The distribution features are similar for
Fe3+ distribution feature of CASI SASI data is clearly different from two other types of minerals: chlorite/epidote/calcite/dolomite and
that of the ASTER data. We checked the imaging spectra of the muscovite/illite.
 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141 139

Fig. 15. Comparison of the mineral mapping result between factor corrected ASTER L1B data and uncorrected ASTER L1B data.

Discussion the study area in this work can meet the requirements because the
composition and status of the ground are constant. The airborne
Remote sensing sensors are radiometrically calibrated in labo- hyperspectral data can provide us with a large amount of ground
ratories prior to delivery. However, the sensors can change during spectra for calculating the radiometric calibration factor to correct
launch or during extended operation in space. Vicarious calibration the spectral shape bias. The accuracy of the airborne hyperspectral
is a method of correcting the changes in radiometric calibration. The data is the key factor affecting the calculation of radiometric cor-
ASTER vicarious calibration method is a reflectance-based method, rection factors. Therefore, we deduce that our calibration factors
which requires measurement of the ground reflectance and atmo- are most likely suitable for other areas because the airborne data
spheric parameters at the time of image acquisition (Biggar et al., were validated qualitatively and quantitatively and were proven to
2005). The top of atmosphere (TOA) predicted radiance value for be highly accurate.
a given ground reflectance was computed using radiative transfer For this study, the precise radiation calibration needs a pre-
models and codes such as 6S or MODTRAN. Then, this calculated cise atmospheric correction. Atmospheric water vapor, which is a
radiance was compared to the average image radiance to compute required atmospheric parameter for spectral construction of the
a calibration coefficient. The calibration correction factor was also ASTER data, affects ASTER bands 3, 5, 8 and 9. In this paper, we
calculated using the ratio of the computed radiance with the mea- developed a new method for calculating this parameter using a few
sured image radiance. The concurrent measurement with image ground spectra based on the hypothesis that the atmospheric water
acquisition prevents measurement of the ground reflectance at a vapor in the spatial range of a single ASTER datum is constant. In
large number of sites. A few ground sites are not enough to reduce arid and bare areas, this method produced very good results, which
random error. Biggar’s radiometric factor error is most likely due are shown in Figs. 6 and 7. However, it is still unknown whether this
to the small number of ground sites, and the error will lead to the method is suitable for humid and rainy areas where atmospheric
anomalous high in the band 5 reflectance of the ASTER data, as water vapor varies much more significantly in space than in arid
observed in this study and that by Mars and Rowan (2010). and bare areas. We also found that this method needs to include at
The method adopted in this paper does not require concurrent least one measured ground spectra. It cannot be used if no ground
measurements, but it does require that the spectral reflectance spectra exist.
shape of the ground site measured at the time of ASTER data acqui- To evaluate the effect of our ASTER reflectance correction fac-
sition is consistent with that measured at the time of airborne data tors on mineral retrieval, we compared the mineral mapping results
acquisition. If we do not consider the effect of the bidirectional of the corrected data with those of uncorrected data (Fig. 15). The
reflectance distribution function of the ground on spectral shape, distributions of kaolinite identified by the two types of data are
140 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141

completely different. Some kaolinte was identified by uncorrected Abrams, M., Hook, S.J., 1995. Simulated Aster data for geologic studies. IEEE Trans.
data but not by corrected data. The distribution features of the mus- Geosci. Remote Sens. 33, 692–699.
Abrams, M., 2000. The Advanced Spaceborne Thermal Emission and Reflection
covite/illite identified by the two types of data are very similar, Radiometer (ASTER): data products for the high spatial resolution imager on
and more muscovite/illite was identified by corrected data than NASA’s Terra platform. Int. J. Remote Sens. 21, 847–859.
by the uncorrected data. We found that the spectra of uncorrected Adler-Golden, S.M., Mattchew, M.W., Bemstein, S., Levine, R.Y., Berk, A., Richtsmeier,
S.C., 1999. Atmospheric correction for short-wave spectral imagery based on
data indicated a kaolinite absorption feature, but the corrected and MODTRAN4. In: Grenn, R.O. (Ed.), Summaries of the Eighth JPL Airborne Earth
CASI SASI data did not show that feature. The spectral shapes of the Science Workshop JPL Publication, Bol. 99-17. Jet Propul. Lab. 1, Pasadena, CA,
corrected and CASI SASI data are very similar to each other. In this pp. 21–29.
Bedell, R.L., 2001. Geological mapping with ASTER satellite: new global satellite data
study area, kaolinite was not identified by CASI SASI either. The dis-
that is a significant leap in remote sensing geologic and alteration mapping. Spec.
tribution area of muscovite/illite identified by the corrected data is Publ. Geol. Soc. Nevada 33, 329–334.
47.4% larger than that identified by the uncorrected data. When we Bedini, E., 2011. Mineral mapping in the Kap Simpson complex, central East
Greenland, using HyMap and ASTER remote sensing data. Adv. Space Res. 47,
compared the distribution features of muscovite/illite identified by
60–73.
the ASTER and CASI SASI data, we found that distribution area of Biggar, S.F., Thome, K.J., McCorkel, J.T., D’Amico, J.M., 2005. Vicarious calibration of
the ASTER data is larger than that of the CASI SASI data in the west- the ASTER SWIR sensor including crosstalk correction. SPIE 5882, 55217–55218.
ern study region; however, the distribution of the CASI SASI data Boardman, J.W., 1998. Leveraging the high dimensionality of AVIRIS data for
improved subpixel target unmixing and rejection of false positives: Mixture
is larger than that of the ASTER data in the middle study region. Tuned Matched Filtering. In: Summaries of the 7th JPL Airborne Earth Science
It is difficult to interpret this difference because this phenomenon Workshop (Jet Propulsion Laboratory Publication), 1, p. 53.
may have several causes, such as data noise and imposed threshold Bishop, C.A., Liu, J.G., Mason, P.J., 2011. Hyperspectral remote sensing for miner-
als exploration in Pulang, Yunan Province, China. Int. J. Remote Sens. 32 (9),
settings. The exact reason needs to be identified in future research. 2409–2426.
Spectral indices have been proposed in this paper for identifying Clark, R.N., King, T.V.V., Klejwa, M., Swayze, G.A., 1990. High spectral resolution
muscovite/illite, kaolinite and alunite. However, the identification reflectance spectroscopy of minerals. J. Geophys. Res. 95, 12653–12680.
Crósta, A.P., De Souza Filho, C.R., Azevedo, F., Brodie, C., 2003. Targeting key alteration
results cannot be validated at this time because there is not any minerals in epithermal deposits in Patagonia, Argentina, using ASTER imagery
kaolinite or alunite in this study area. The identification of kaolin- and principal component analysis. Int. J. Remote Sens. 24, 4233–4240.
ite and alunite is very important because these two minerals are De Berg, M., Cheong, O., Van Kreveld, M., Overmars, M., 2008. Computational
Geometry: Algorithms and Applications. Springer-Verlag, Berlin Heidelberg, pp.
important alteration types for searching for epithermal deposits
191–218.
(Richard, 2010; van der Meer et al., 2012). Di Tommaso, I., Rubinstein, N., 2007. Hydrothermal alteration mapping using ASTER
data in the Infiernillo porphyry deposit, Argentina. Ore Geol. Rev. 32, 275–290.
Galvão, L.S., Almeida-Filho, R., Vitorello, Í., 2005. Spectral discrimination of
Conclusions hydrothermally altered materials using ASTER short-wave infrared bands: eval-
uation in a tropical savannah environment. Int. J. Appl. Earth Observ. Geoinform.
We propose an alternative method for atmospheric correction 7, 107–114.
Gao, B., Kaufman, Y.J., 1998. Algorithm Technical Background Document: The
of the ASTER VNIR/SWIR data without retrieving the atmospheric MODIS Near-IR Water Vapor Algorithm, Product ID: MOD05 – Total Precip-
water vapor using the MODIS data. The water vapor value could itable Water. Science Publications and Reports, 25. http://modis.gsfc.nasa.gov/
be calculated on the basis of the linear relationship between the data/atbd/atbd mod03.pdf
Gao, B.C., Heidebrecht, K.B., Goetz, A.F.H., 1993. Derivation of scaled surface
atmospheric water vapor and the ASTER band 8 absorption depth, reflectances from AVIRIS data. Remote Sens. Environ. 44, 165–178.
if the ground spectra of at least one site were measured in the Gao, B.C., Montes, M.J., Davis, C.O., Goetz, A.F.H., 2009. Atmospheric correction algo-
field. This method is simple and fast and shows good results during rithms for hyperspectral remote sensing data of land and ocean. Remote Sens.
Environ. 113, S17–S24.
application in arid and bare areas. Gleeson, D.F., Pappalardo, R.T., Grasby, S.E., Anderson, M.S., Beauchamp, B., Castano,
The ASTER VNIR/SWIR reflectance correction factors were R., Chien, S.A., Doggett, T., Mandrake, L., Wagstaff, K.L., 2010. Characterization
derived for correcting the spectral shape bias resulting from the of a sulfur-rich Arctic spring site and field analog to Europa using hyperspectral
data. Remote Sens. Environ. 114, 1297–1311.
radiometric calibration error. After applying the reflectance factor
Green, R.O., Eastwood, M.L., Sarture, C.M., Chrien, T.G., Aronsson, M., Chippendale,
to the atmospheric corrected ASTER L1B data, the mean error of B.J., Faust, J.A., Pavri, B.E., Chovit, C.J., Solis, M., Olah, M.R., Williams, O., 1998.
all reflectance bands decreased from 0.0256 to 0.002 for this study Imaging spectroscopy and the airborne visible/infrared imaging spectrometer
case, and the standard deviation decreased from 0.04251 to 0.0007. (AVIRIS). Remote Sens. Environ. 65, 227–248.
Griffin, M.K., Burke, H.K., 2003. Comparison of hyperspectral data for atmospheric
The errors of the 2/1, 5/6 and 9/8 band ratios decreased from 2.38%, effects. Lindoln Lab. J. 14, 227–248.
4.102%, and 4.28% to 1.26%, −0.162%, and 0.31%, respectively. Radio- Haselwimmer, C.E., Riley, T.R., Liu, J.G., 2011. Lithologic mapping in the Oscar II Coast
metric calibration errors in the ASTER band 1–9 data can lead to the area, Graham land, Antarctic Peninsula using ASTER data. Int. J. Remote Sens. 32
(7), 2013–2035.
overestimation of kaolinite. Hewson, R.D., Cudahy, T.J., Mizuhiko, S., Ueda, K., Mauger, A.J., 2005. Seamless geo-
The ASTER data have no band at approximately 1000 nm, the logical map generation using ASTER in the Broken Hill-Curnamona province of
main Fe3+ absorption feature, and the band index of 2/1 for retriev- Australia. Remote Sens. Environ. 99, 159–172.
Hunt, G., 1977. Spectral signatures of particulate minerals in the visible and near
ing Fe3+ cannot produce a reliable Fe3+ distribution map. A new infrared. Geophysics 42, 501–513.
index should therefore be developed. Hunt, G.R., Ashley, R.P., 1979. Spectra of altered rocks in the visible and near infrared.
Econ. Geol. 74, 1613–1629.
Iwasaki, A., Fujisada, H., Akao, H., Shindou, O., Akagi, S., 2001. Enhancement of spec-
Acknowledgements tral separation performance for ASTER/SWIR. Proc. SPIE 4486, 42–50.
Iwasaki, A., Tonooka, H., 2005. Validation of a crosstalk correction algorithm for
This research was supported by the grant from the National ASTER/SWIR. IEEE Trans. Geosci. Remote Sens. 43, 2747–2751.
Kaufmann, Y.J., Wald, A.E., Remer, L.A., Gao, B.C., Li, R.R., Flynn, L., 1997. The MODIS
Natural Science Foundation of China (Grant No. 41102208 and 2.1-␮m channel-correction with visible reflectance for use in remote sensing of
Grant No. 41172302) and Project of China Geological Survey (Grant aerosol. IEEE Trans. Geosci. Remote Sens. 35, 1286–1298.
No. 1212010660601). The authors wish to thank the anonymous Kavak, K.S., 2005. Recognition of gypsum geohorizons in the Sivas Basin (Turkey)
using ASTER and Landsat ETM+ images. Int. J. Remote Sens. 26, 4583–4596.
reviewers and F. D. van der Meer for their constructive comments King, T.V., Clark, R., 2000. Verification of remotely sensed data. In: Kuehn, F., King,
that helped to improve the scholarly quality of the paper. T.V., Hoerig, B., Peters, D. (Eds.), Remote Sensing for Site Characterization.
Springer, Berlin Heidelberg, pp. 59–62.
Kruse, F.A., 2004. Comparison of ATREM, ACORN, and FLAASH atmospheric correc-
References tions using low-altitude AVIRIS data of Boulder, CO. In: Summaries of 13th JPL
Airborne Geoscience Workshop, Jet Propulsion Lab, Pasadena, CA.
Aboelkhair, H., Ninomiya, Y., Watanabe, Y., Sato, I., 2010. Processing and interpreta- Kratt, C., Calvin, W.M., Coolbaugh, M.F., 2010. Mineral mapping in the Pyramid Lake
tion of ASTER TIR data for mapping of rare-metal-enriched albite granitoids in basin: hydrothermal alteration, chemical precipitates and geothermal energy
the Central Eastern Desert of Egypt. J. Afr. Earth Sci. 58, 141–151. potential. Remote Sens. Environ. 114, 2297–2304.
 C. Jing et al. / International Journal of Applied Earth Observation and Geoinformation 33 (2014) 127–141 141

Kruse, F.A., Boardman, J.W., 2000. Characterization and mapping of kimberlites and Richter, R., 1996. Atmospheric correction of DAIS Hyperspectral image data. Comput.
related diatremes using hyperspectral remote sensing. In: Aerospace Conference Geosci. 22, 785–793.
Proceedings, 2000 IEEE, vol. 293, pp. 299–304. Richter, R., 1998. Correction of satellite imagery over mountainous terrain. Appl.
Madani, A.A., Emam, A.A., 2011. SWIR ASTER band ratios for lithological mapping Opt. 37, 4004–4015.
and mineral exploration: a case study from El Hudi area, southeastern desert, Ritcher, R., Schlaepfer, D., 2002. Geo-atmospheric processing of airborne imaging
Egypt. Arab. J. Geosci. 4, 45–52. spectrometry data. Part 2: atmospheric/topographic correction. Int. J. Remote
Mars, J.C., Rowan, L.C., 2006. Regional mapping of phyllic- and argillic-altered rocks Sens. 23 (13), 2631–2649.
in the Zagros magmatic arc, Iran, using Advanced Spaceborne Thermal Emis- Rockwell, B.W., Hofstra, A.H., 2008. Identification of quartz and carbonate minerals
sion and Reflection Radiometer (ASTER) data and logical operator algorithms. across northern Nevada using ASTER thermal infrared emissivity data implica-
Geosphere 2, 161–186. tions for geologic mapping and mineral resource investigations in well-studied
Mars, J.C., Rowan, L.C., 2010. Spectral assessment of new ASTER SWIR surface and frontier areas. Geosphere 4 (1), 218–246.
reflectance data products for spectroscopic mapping of rocks and minerals. Rowan, L.C., Hook, S.J., Abrams, M.J., Mars, J.C., 2003. Mapping hydrothermally
Remote Sens. Environ. 114, 2011–2025. altered rocks at Cuprite, Nevada, using the Advanced Spaceborne Thermal Emis-
Matar, S.S., Bamousa, A.O., 2013. Integration of the ASTER thermal infra-red bands sion and Reflection Radiometer (ASTER), a new satellite-imaging system. Econ.
imageries with geological map of Jabal Al Hasir area, Asir Terrane, the Arabian Geol. 98 (5), 1019–1027.
Shield. J. Taibah Univ. Sci. 7 (1), 1–7. Rowan, L.C., Mars, J.C., 2003. Lithologic mapping in the Mountain Pass, California,
Melendez-Pastor, I., Navarro-Pedreño, J., Koch, M., Gómez, I., 2010. Applying imaging area using Advanced Spaceborne Thermal Emission and Reflection Radiometer
spectroscopy techniques to map saline soils with ASTER images. Geoderma 158, (ASTER) data. Remote Sens. Environ. 84 (3), 350–366.
55–65. Rowan, L.C., Mars, J.C., Simpson, C.J., 2006. Lithologic mapping of the Mordor,
Ninomiya, Y., Fu, B., 2003a. Extracting lithologic information from ASTER multi- NT, Australia ultramafic complex by using the Advanced Spaceborne Ther-
spectral thermal infrared data in the northeastern PAMIRS. Xinjiang Geol. 21, mal Emission and Reflection Radiometer (ASTER). Remote Sens. Environ. 99,
22–29. 105–126.
Ninomiya, Y., 2003b. A stabilized vegetation index and several mineralogic indices Staenz, K., Szeredi, T., Schwarz, J., 1998. ISDAS-a system for processing/analyzing
defined for ASTER VNIR and SWIR data. In: Geoscience and Remote Sensing Sym- hyperspectral data. Can. J. Remote Sens. 24, 99–113.
posium, 2003. IGARSS’03. Proceedings. 2003 IEEE International, pp. 1552–1554. Tangestani, M.H., Mazhari, N., Agar, B., Moore, F., 2008. Evaluating Advanced Space-
Ninomiya, Y., 2004. Lithologic mapping with multispectral ASTER TIR and SWIR borne Thermal Emission and Reflection Radiometer (ASTER) data for alteration
data. In: Proc. SPIE 5234, Sensors, Systems, and Next-Generation Satellites VII, zone enhancement in a semi-arid area, northern Shahre-Babak, SE Iran. Int. J.
pp. 180–190. Remote Sens. 29, 2833–2850.
Öztan, S.N., Süzen, L.M., 2011. Mapping evaporate minerals by ASTER. Int. J. Remote Tangestani, M.H., Jaffari, L., Vincent, R.K., Sridhar, B.B.M., 2011. Spectral char-
Sens. 32, 1651–1673. acterization and ASTER-based lithological mapping of an ophiolite complex:
Pour, A.B., Hashim, M., 2011. Identification of hydrothermal alteration minerals for a case study from Neyriz ophiolite, SW Iran. Remote Sens. Environ. 115,
exploring of porphyry copper deposit using ASTER data, SE Iran. J. Asian Earth 2243–2254.
Sci. 42, 1309–1323. van der Meer, F.D., van der Werff, H.M.A., van Ruitenbeek, F.J.A., Checker, C.A., Bakker,
Pour, A.B., Hashim, M., 2012a. The application of ASTER remote sensing data to W.H., Noomen, M.F., van der Meijde, M., Carranza, E.J.M., 2012. Multi- and hyper-
porphyry copper and epithermal gold deposits. Ore Geol. Rev. 44, 1–9. spectral geologic remote sensing: a review. Int. J. Appl. Earth Observ. Geoinform.
Pour, A.B., Hashim, M., 2012b. Identifying areas of high economic-potential copper 14, 112–128.
mineralization using ASTER data in the Urumieh–Dokhtar Volcanic Belt, Iran. Yamaguchi, Y., Fujisada, H., Kudoh, M., Kawakami, T., Tsu, H., Kahle, A.B., Pniel, M.,
Adv. Space Res. 49, 753–769. 1999. ASTER instrument characterization and operation scenario. Adv. Space
Qu, Z., Kindel, B.C., Goetz, A.F.H., 2003. The high accuracy atmospheric correction Res. 23, 1415–1424.
for hypersepctral dta (HATCH) model. IEEE Trans. Geosci. Remote Sens. 41, Yamaguchi, Y., Naito, C., 2003. Spectral indices for lithologic discrimination
1223–1231. and mapping by using the ASTER SWIR bands. Int. J. Remote Sens. 24,
Rajendran, S., Thirunavukkarrasu, A., Balamurugan, G., Shandkar, K., 2011. Discrimi- 4311–4323.
nation of iron ore deposits of granulite terrain of southern Peninsular India using Zhang, X., Pazner, M., Duke, N., 2007. Lithologic and mineral information extrac-
ASTER data. J. Asian Earth Sci. 41, 99–106. tion for gold exploration using ASTER data in the south Chocolate Mountains
Richard, H.S., 2010. Porphyry copper systems. Society of Economic Geologists, Inc. (California). ISPRS J. Photogram. Rem. Sens. 62, 271–282.
Econ. Geol. 105, 3–41.