Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223550481

Detecting lithology with Advanced Spaceborne Thermal Emission and Refection


Radiometer (ASTER) multispectral thermal infrared ‘radiance-at-sensor’ data

Article  in  Remote Sensing of Environment · November 2005


DOI: 10.1016/j.rse.2005.06.009

CITATIONS READS

358 2,189

3 authors:

Yoshiki Ninomiya Bihong Fu


National Institute of Advanced Industrial Science and Technology Chinese Academy of Sciences
47 PUBLICATIONS   1,776 CITATIONS    25 PUBLICATIONS   1,452 CITATIONS   

SEE PROFILE SEE PROFILE

Thomas Cudahy
C3DMM Pty Ltd
116 PUBLICATIONS   2,549 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Airborne Hyperspectral for coastal environment monitoring View project

Mineral Mapping with Field Spectrometers for Exploration 1997-99 View project

All content following this page was uploaded by Yoshiki Ninomiya on 19 July 2016.

The user has requested enhancement of the downloaded file.


* Manuscript

Corrigendum

Detecting lithology with Advanced Spaceborne Thermal Emission and


Reflection Radiometer (ASTER) multispectral thermal iinfrared
nfrared
"radiance-
"radiance-at-
at-sensor" data

Remote Sensing of Environment 99(1-2):127-139 (2005), ASTER special issue

Yoshiki Ninomiyaa, Bihong Fub and Thomas J. Cudahyc

aGeological Survey of Japan, AIST, Tsukuba 305-8567, Japan


bLanzhou Institute of Geology, Chinese Academy of Sciences, Lanzhou 730000, China
cCSIRO Exploration and Mining, PO Box 1130, Bentley, WA 6101, Australia
Corresponding author: Yoshiki Ninomiya Email: Yoshiki.Ninomiya@aist.go.jp

A factor of π was inadvertently left out of equation 10 in the original article. This
factor is necessary to convert the total energy leaving a surface element to radiance (W
m-2 µm-1 sr-1), as shown in Equation 5. The corrected equation is:

 λ13  c1 
exp i ⋅ ln + 1 − 1
λ  π ⋅ λ13 5 × L 13 
i i   sen 
nLsen = Lsen × (10 )
 
 c2 
exp  −1
 λ i × nT ε 13 
 a 
Remote Sensing of Environment 99 (2005) 127 – 139

Detecting lithology with Advanced Spaceborne Thermal Emission and


Reflection Radiometer (ASTER) multispectral thermal infrared
‘‘radiance-at-sensor’’ data
Yoshiki Ninomiya a,*, Bihong Fu b, Thomas J. Cudahy c
a
Geological Survey of Japan, AIST, Tsukuba 305-8567, Japan
b
Lanzhou Institute of Geology, Chinese Academy of Sciences, Lanzhou 730000, China
c
CSIRO Exploration and Mining, PO Box 1130, Bentley, WA 6101, Australia

Received 1 October 2004; received in revised form 7 June 2005; accepted 20 June 2005

Abstract

The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) aboard NASA’s Terra satellite measures multispectral
thermal infrared (TIR) emission from the Earth’s surface to space. Based on analysis of TIR spectral properties of typical rocks on the Earth,
several mineralogic indices including the Quartz Index (QI), Carbonate Index (CI) and Mafic Index (MI) for detecting mineralogic or
chemical composition of quartzose, carbonate and silicate rocks with ASTER-TIR data are proposed. These indices are applied to the
ASTER-TIR data scenes for selected study areas in China and Australia. The results show that ASTER-TIR can discriminate quartz and
carbonate rocks as well as mafic – ultramafic rocks, even with atmospherically uncorrected radiance-at-sensor data. Lithologic interpretations
agree well with published geologic data and field observations. The mineralogic indices applied to ASTER-TIR provide one unified approach
for lithologic mapping in arid and semi-arid regions of the Earth.
D 2005 Elsevier Inc. All rights reserved.

Keywords: Quartz; Carbonate; Silicate; Mafic; Felsic; Ophiolite; Mineralogic indices; Emissivity spectra; ASTER; Thermal infrared; Geology; Lithologic
mapping

1. Introduction Remote-sensing for lithologic mapping using TIR


spectral signatures was first demonstrated using the
In a pioneering study of spectroscopy, Lyon (1965) airborne Thermal Infrared Multispectral Scanner called
demonstrated that silica and silicate minerals, the major TIMS (Kahle & Goetz, 1983; Kahle et al., 1980; Kahle &
components of the Earth’s crust, show strong fundamental Rowan, 1980). Similar airborne systems (e.g., Fu & Chou,
spectral bands corresponding to the Si – O bond length in the 1998) confirmed the usefulness of TIR multispectral
thermal infrared (TIR) atmospheric window (8 – 12 Am), remote sensing. These TIR systems were able to measure
although they do not cause prominent spectral features in the changes in wavelength of the broad emissivity low
the visible to shortwave infrared region of the spectrum related to the Si – O bonds. Systems with higher spectral
(0.4 –2.5 Am). Various workers (e.g., Hunt & Salisbury, resolution, such as MIRACO2LAS (Cudahy and others,
1974; Salisbury et al., 1988) have shown that TIR 1999) and SEBASS (e.g., Cudahy et al., 2000), are able to
emissivity spectra of igneous rocks are correlated with the map more detailed TIR spectral signatures related to the
bulk (‘‘chemical’’) SiO2 content. abundances and chemistries of specific silicate, sulphate
and carbonate minerals.
The Advanced Spaceborne Thermal Emission and
* Corresponding author. Reflection Radiometer (ASTER) sensor was developed
E-mail address: Yoshiki.Ninomiya@aist.go.jp (Y. Ninomiya). based on the success of TIMS, and was launched onboard
0034-4257/$ - see front matter D 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2005.06.009
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

Terra in December 1999. Terra was the first of NASA’s 2. Basics and definition
Earth Observation System (EOS) series of satellites. It
obtains multispectral image of the Earth (Yamaguchi et 2.1. Radiance at sensor
al., 1998) not only in the visible to near-infrared (VNIR;
three bands between 0.5 and 0.9 Am, 15-m resolution, The DN of ASTER-TIR band i in Level-1B or Level-3A
stereoscopic capability for the NIR band) and in the data is linearly related to radiance at sensor (Wm 2 sr 1
shortwave infrared (SWIR; six bands between 1.6 and 2.5 Am 1) by
Am, 30-m resolution), but also in the TIR (five bands  
between 8 and 12 Am, 90-m resolution, NEDT < 0.3 K at Lisen ¼ coef i  DN i  1 ; ð1Þ
300 K). An ASTER image covers about 60  60 km. where coef10 = 0.006882, coef11 = 0.006780, coef12 = 0.006590,
ASTER-TIR is the first satellite-borne multispectral TIR coef13 = 0.005693, and coef14 = 0.005225.
remote-sensing system with sufficient spectral, spatial and On the other hand, L sen is related to the radiance leaving
radiometric resolutions for geological applications. the surface (L sur), the transmissivity of the atmosphere (s)
ASTER typically observes more than 500 scenes per day. and the atmospheric path radiance (L Aj) by the equation
Most of the images are processed to Level-1A. All the
Level-1A images satisfying a condition on detected cloud Lsen ¼ Lsur Is þ LAj : ð2Þ
coverage are processed for nominal radiometric and geo- L sur is related to the radiance emitted from the surface by
metric corrections to generate the Level-1B ‘‘radiance-at-
sensor’’ data product. Level-2B data products of physical EA,
Lsur ¼ LBB Ie þ Iq; ð3Þ
parameters including surface radiance data with nominal k
atmospheric corrections (Level-2B01) are generated on the where L BB is the radiance emitted from a blackbody surface
basis of user request. Also, Level-4 DEMs are generated as (given by Planck function), e is surface emissivity, E A, is
from the along-track NIR stereo images (band 3N and 3B). the downwelling irradiance emitted by the atmosphere, and
A geometrically well-corrected orthorectified Level-3A q is surface reflectance, which for Lambertian surfaces is
product with ASTER-driven DEM is generated, which is related to surface emissivity by the Kirchhoff’s law,
radiometrically equivalent to the Level-1B data (Yamaguchi
q ¼ 1  e: ð4Þ
et al., 1998).
Most of the algorithms proposed so far for thematic From Eqs. (2) – (4) and Planck’s function, radiance at
mapping with multispectral TIR remote-sensing data, for sensor is given as
example, an algorithm to estimate the chemical SiO2 !
i
content in the surface silicate rocks (Ninomiya, 1995), i c1 1 i
EA,  i
 i i
Lsen ¼ 5 I
 i
 Ie þ I 1  e Is þLAj ;
assume the input is spectral emissivity data (e.g., ASTER kki exp c2 =k IT  1 k
Level-2B05 product) calculated after atmospheric-correc-
ð5Þ
tion and temperature separation. Radiance at sensor was
2
thought not suitable for quantitative analysis. However, where the radiation constants c 1 = 3.742  10 (Wm Am4)
8

the potential applicability of Quartz Index (QI), Carbonate and c 2 = 1.439  104 (Am K), T is the surface temperature
Index (CI) and Mafic Index (MI) defined for ASTER-TIR (K), k i is the center wavelength (Am) of ASTER band i,
data in scene-independent lithologic detection is suggested (i.e., k 10 = 8.3, k 11 = 8.65, k 12 = 9.1, k 13 = 10.6, k 14 = 11.3).
through theoretical and practical analyses of the behavior
of the indices, taking into consideration temperature 2.2. Spectral properties of rocks
and atmospheric parameters (Ninomiya, 2002). One
advantage over methods utilizing Level-2B05 data is that Fig. 1 shows directional emissivity spectral properties of
the lower-level data are free of functional uncertainties the selected typical rocks (Ninomiya & Fu, 1999) collected at
introduced by the atmospheric-correction and temperature the Yushishan study area described in Section 3.1. The
separation algorithms. Another advantage is that the spectra were measured with a portable thermal infrared
lower-level data are more readily available. Additionally, Fourier Transform Interferometer manufactured by Design
the straightforward applicability of the index approach to and Prototypes, Ltd., (Hook & Kahle, 1996) and convolved
orthorectified Level-3A data enables easy fusion with into the responsivity functions of ASTER-TIR (Fujisada,
GIS. 1995).
This paper focuses on the usefulness of these mineral Major carbonate minerals show spectral emissivity
indices calculated from ASTER-TIR Level-1B and Level- minima in ASTER band 14 (near 11.4 Am for calcite and
3A ‘‘radiance-at-sensor’’ data. Three test sites at different 11.2 Am for dolomite) due to the C– O bending mode, as
elevations are selected to show the stability of the indices shown in Fig. 1a. This feature in the emissivity spectra of
against temperature and atmospheric conditions, and to other carbonate minerals shifts to shorter wavelengths (e.g.,
define the exact threshold values of the indices for lithologic to ¨11.0 Am for magnesite). Quartz and unaltered igneous
mapping. rocks composed of silicate minerals typically do not show
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

oxides (Salisbury et al., 1992) and halite (Crowly & Hook,


1996) show similar spectral shape to ultramafic rocks (i.e.,
emissivities are high in ASTER bands 10 to 12; low in
ASTER bands 13 and 14).

2.3. Definition of indices (Ninomiya & Fu, 2002)

From the spectral emissivity property of a carbonate rock


composed of calcite and dolomite, the two major carbonate
minerals on the Earth, shown in Fig. 1a and described in
Section 2.1, the Carbonate Index (CI) for ASTER-TIR data
is defined as
D13
CI ¼ ; ð6Þ
D14
where D i is any kind of ASTER data related to ASTER
band i. In this paper, we use ‘‘radiance-at-sensor’’ data
without atmospheric corrections for D. CI is expected to be
high for calcite and dolomite. No peculiar response is
expected for other carbonate minerals.
From the spectral emissivity property of quartz shown in
Fig. 1b and described in Section 2.1, the Quartz Index (QI)
is defined as
D11  D11
QI ¼ : ð7Þ
D10  D12
Fig. 1. Emissivity spectra of (a) carbonate rock, (b) quartzose rock, (c)
granite, (d) diorite, (e) gabbro, (f) peridotite, with the convolved data to QI is expected to be high for quartz and low for K-feldspar
ASTER bandpasses. Each tick in Y-axis registers 1.0 / 0.75 in emissivity and gypsum.
except for (b): 1.0 / 0.5.
As described in Section 2.1, the broad spectral emissivity
low shifts to longer wavelengths as the chemical SiO2
any spectral features in VNIR to SWIR. In contrast, they content in silicate rock decreases, as shown in Fig. 1c to f.
have prominent spectral features in TIR region due to This introduced the definition of the Mafic Index (MI) as
fundamental asymmetric Si – O – Si stretching vibrations. D12
Quartz, the most common mineral on the Earth, shows MI ¼ : ð8Þ
D13
absorption features (i.e., emissivity minima) in ASTER
bands 10 and 12, resulting in higher emissivity in band 11 MI is correlated to the SiO2 content in silicate rocks,
than in bands 10 and 12, as shown in Fig. 1b. The series of typically igneous rocks, but it is also sensitive to carbonates.
alkali feldspars (K-feldspars), which often coexist with To eliminate this unexpected property of MI, a series of
quartz in felsic igneous rocks, have a strong emissivity peak Mafic Index separated for carbonates, MIn , is redefined as
in band 11, resulting in lower emissivity in band 11 than in D12 D12 ID14n
bands 10 and 12, contrary to the property of quartz MI n ¼ n ¼ : ð9Þ
D13 ICI D13nþ1
described above. For silica and silicate minerals and rocks,
the broad spectral emissivity low corresponding to Si – O The original MI is the case for which n = 0. Comparing
bond length shifts to longer wavelength as the chemical images of different versions of MIn series shows good
SiO2 content (weight percent) decreases. After this property, separation of carbonates and silicates in MI3 (Ninomiya,
the ratio of the emissivity at band 12 to band 13 for silicate 2002). Therefore, in the present paper, we use MI3 for MI.
rocks (typically igneous rocks) increases as the SiO2 content MI is expected to correlate negatively with the SiO2 content
decreases (i.e., as the rock type changes from felsic to in silicate rocks. That is, it is expected to be high for
mafic), as shown in Fig. 1c, d, e and f. In addition, some ultramafic rocks, and systematically lower as the rock type
sulfate minerals including gypsum have a very strong changes to felsic and finally quartzose rock. MI is expected
absorption at band 11 spectral region (i.e., near 8.7 Am) to be high for halite and some iron oxides with the spectral
due to stretching fundamentals, as a result, it exhibits lower property described in Section 2.1. For theoretical blackbody
emissivity in band 11 than in bands 10 and 12, likely the and natural graybody materials, typically vegetation,
property of K-feldspars described above (Ninomiya & Fu, MI å 0.89, which is similar to index values for intermediate
2003). According to published spectral properties, some rocks with chemical SiO2 content ¨65% (Ninomiya, 2002).
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

This MI value is expected to be a robust boundary between


mafic and felsic rocks, with minimum influence of other
factors on spectral contrast, for example, atmospheric
downwelling irradiance and topographic effects, because
blackbodies have low spectral contrast.

2.4. Stability analysis and improvement

The Carbonate Index (CI), Quartz Index (QI) and Mafic


Index (MI) were calculated from ASTER Level-1B data.
Fig. 3. The Quartz Index (QI) calculated on simulated radiance-at-sensor
Scene-dependent qualitative analyses were made of the gray- data vs. atmospheric water-vapor content (kg/m2) assigned in MODTRAN
scale images of each index and the false-color composite at the elevation of 1000 m asl.
image of the three indices. Ninomiya and Fu (2002) pointed
out the potential usefulness of these indices for discriminat- normalization processing is not important for QI and MI;
ing rock types. A theoretical analysis of the stability of the however, here the normalized radiance at sensor is used for
indices with respect to surface temperature and atmospheric all the indices for the uniformity of the data processing.
parameters indicates that QI and MI are insensitive to Hereafter, the indices applied for the normalized radiance
temperature, provided atmospheric conditions are good, but calculated with Eq. (10) are expressed as CI, QI and MI,
that CI is heavily affected by temperature differences even in respectively.
good atmospheric conditions. We have confirmed these For analyzing the sensitivity of the indices to the
properties of the indices by analysis of multi-temporal atmospheric parameters, simulated ASTER-TIR radiance-
images of known study areas (Ninomiya, 2002). Normal- at-sensor data were generated for a 300-K blackbody and
ization of the brightness temperature for band 13 to a fixed typical rock samples shown in Fig. 1. Spectral atmospheric
temperature reduces the heavy dependency of CI to surface transmissivity, path radiance and downwelling irradiance
temperature. The normalized radiance at sensor at band i is were derived using an atmospheric radiative transfer
defined as model, MODTRAN, a moderate-resolution version of
( !) LOWTRAN 7 (Kneizys et al., 1988), applied to the
k13 c1
exp Iln 135 þ1 1 ‘‘mid-latitude summer’’ model atmosphere. The measured
i i
ki k  L13 sen emissivity spectra in Fig. 1 and the calculated spectra of
nLsen ¼ Lsen    ;
c2 atmospheric parameters are convolved into responsivity
exp i 1
k  nT =ea13 function of each band in ASTER-TIR (Fujisada, 1995) to
generate e i , s i , L Aji and E A,i in Eq. (5). The spectral
ð10Þ
contrast of emissivity for the surface rocks in remote
where L seni is radiance at sensor in band i, k i is the center sensing is usually degraded by various factors, for
wavelength (Am) of band i, e a13 is the assumed emissivity in example, weathering, topography and mixing with gray-
band 13, nT is the fixed temperature (K) to be normalized, body materials like vegetation. The degraded emissivity,
and c 1 and c 2 are the radiation constants given in Eq. (5). e d, can be estimated as
Here in this study, e a13 is adopted as 1.0, and nT is adopted as
ed ¼ ðe  1ÞIa þ 1; ð11Þ
300. Case studies with the indices applied to the normalized
radiance-at-sensor data suggested successful improvement where the degradation ratio, a, is between 0 and 1. Here,
on the ability of CI in mapping carbonate rocks (Ninomiya, degradation is not considered in generating simulated
2003; Ninomiya, 2004; Ninomiya & Fu, 2003). The ASTER-TIR radiance-at-sensor data, so a = 1. (Downwel-

Fig. 2. The Carbonate Index (CI) calculated on simulated radiance-at-sensor Fig. 4. The Mafic Index (MI) calculated on simulated radiance-at-sensor
data vs. atmospheric water-vapor content (kg/m2) assigned in MODTRAN data by atmospheric water-vapor content (kg/m2) assigned in MODTRAN
at the elevation of 1000 m asl. at the elevation of 1000 m asl.
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

ling atmospheric irradiance also degrades the spectral 3. Results


contrast of emissivity as shown in Eq. (5), but this term
is treated separately.) 3.1. Yushishan study area
As an example, the effect of water vapor content (kg/m2)
assigned in MODTRAN at the elevation of 1000 m above Mt. Yushishan is located in Beishan Mountains, north-
sea level (asl) on CI calculated from the simulated radiance- western Gansu Province, China. The study area is the entire
at-sensor data is shown in Fig. 2. By comparing the results ASTER Level-1B scene acquired on April 7, 2000. Its
shown in Fig. 2 with the result derived by changing the center is located at 41-25VN and 95-05VE. The study area has
elevation of scene, it became clear that the main atmos- a typical continental arid climate with sparse vegetation.
pheric factor affecting the indices is water vapor content. The elevation of this area ranges from 1500 to 2500 m asl.
Fig. 2 indicates that CI is > 1.04 for the carbonates only if Precipitable water at the time of the ASTER data acquisition
the atmospheric water vapor content is less than 15 kg/m2. is estimated as 5 kg/m2 with the archived NCEP Reanalysis
The corresponding relationships for QI and MI are shown in data.
Figs. 3 and 4, respectively. Fig. 6a displays a geological map (BGMRGP, 1989)
Figs. 2 – 4 suggest each index responds sensitively for the overlaid on the VNIR false-color composite image of the
targeted rock types. This indicates the possibility of Yushishan study area. Outcrops are exposed well in this
mapping the target rock types using fixed threshold values study area, and consist of a variety suite of intrusive rocks
independent of the specific scene, provided that the ranging in composition from felsic to ultramafic, and in age
atmospheric conditions are good enough. from Precambrian to Permian. Rocks of the upper Precam-
Fig. 5 shows for ASTER Level-1B images a scatter brian Sinian system are dominated by dolomitic limestone
diagram of the histogram peak of CI vs. precipitable water and quartzite. For all the other sequences, older or younger
vapor content drawn from NCEP Reanalysis data provided than Sinian sequence, major components are silicate
by the NOAA-CIRES Climate Diagnostics Center, sedimentary, metamorphic and volcanic rocks with minor
Boulder, Colorado, USA, from their Web site at http:// exposure of carbonate rocks. Quaternary sediments are
www.cdc.noaa.gov/. The closed dots in Fig. 5 are for mainly composed of debris.
highly vegetated ASTER images, which are expected to Fig. 6b shows a color-composite image made from the
represent CI of graybody vegetation. The open dots are for three indices, with QI assigned to red, CI to green, and MI to
sparsely vegetated ASTER images, which may be affected blue. The three components were each linearly stretched
by the distribution of rocks in the image. Fig. 5 suggests such that 99% of the data ranged between 0 and 255 DN.
the applicable threshold for mapping carbonate with CI The usefulness for lithologic study of the color composite,
would be 1.04 to 1.045 when the atmospheric water vapor together with gray-scale images of the individual indices,
content is low enough (as a guideline, < 10 kg/m2), which has been pointed out by Ninomiya and Fu (2002) in a
is consistent with the results shown in Fig. 2 and described photointerpretive study. Our present paper confirms the
above. predicted lithologic properties through the discussion of
qualitative results such as those in Fig. 6b, together with the
quantitative index values.
Horizontal stripes in the cross-track direction are
significant in Fig. 6b and c. They arise from residual system
noise resulting from insufficient corrections for sensitivity
differences of each detector element in the ASTER-TIR
array, and also from inaccurate chopper temperature
calibration.
As shown in Fig. 6a and b, clouds have similar high CI
index values to carbonates, and are displayed as greenish in
Fig. 6b. Even haze and thin clouds undetectable in VNIR
images can cause this kind of confusion in CI, and analysis
of multiple images of the target area is necessary to assure
reliability.
Quaternary deposits shown in Fig. 6b were divided into
several parts according to provenance. The high-QI, reddish
part (Qa) is expected to be quartzite debris; the high-CI,
greenish part (Qb) is expected to be carbonate; and the low-
QI, low-CI but high-MI, bluish part (Qc) is expected to be
Fig. 5. A scatter diagram of full-scene histogram peak of CI for ASTER
Level-1B images vs. precipitable water-vapor content from NCEP silicate rocks. The region ‘‘Qd’’ shows various colors, which
Reanalysis data from the NOAA-CIRES Climate Diagnostics Center, suggests complexity of rock types and source regions. The
Boulder, Colorado, USA. region ‘‘Qe,’’ represented as cyan, has MI values as high as
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

Fig. 6. (a) A compiled geological map overlaid on an ASTER VNIR false-color composite image of the Mt. Yushishan study area. Abbreviated names of map
units: Z, Precambrian rocks; C – O, Cambrian to Ordovician rocks; O – S, Ordovician to Silurian rocks; P, Permian rocks. (b) Color-composite image of the
indices: QI = red, CI = green and MI = blue. Index values linearly scaled to display 99% of the histogram between 0 and 255 DN. The alphabetic labels identify
the targets of discussion in the main text. (c) Detected pixels with the indices as: red, quartzite (QI > 1.05); dark red, siliceous rock (QI > 1.03); yellow, carbonate
rock (CI > 1.045); dark yellow, possible carbonate rock (CI > 1.035); purple, ultramafic rock (MI > 0.92). Display is of MI image with a fixed gray-scale range of
0.8 to 0.9.
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

those for mafic rocks. However, further investigation is values indicate that part of the mapped felsic intrusions (‘‘L’’)
needed to determine whether the suspected mafic debris is, as appears, to have mafic to ultramafic composition,
exists or not. One possible source of confusion and although it is not indicated as such in the published geologic
ambiguity is the presence of minerals showing similar map.
values of MI, for example halite or some iron oxides.
With respect to the older sequences, a part of Sinian 3.2. Mt. Fitton study area
system mapped as Precambrian and shown as ‘‘A’’ in Fig. 6a
has high QI values (reddish) and we expect it to be quartzite, The Mt. Fitton study area is in the eastern central part of
but most of the rest of the Sinian system (‘‘B’’ and the area South Australia. It lies between 29-45V and 30-00V S, and
around Mt. Yushishan itself) has high CI values (greenish) between 139-10Vand 139-30VE. The elevation there ranges
and we expect it to be limestone or dolomite. Except for the from 50 to 750 m asl. The climate is arid, and vegetation is
Sinian system, other areas appear bluish, suggesting that the sparse. We analyzed an ASTER Level-3A image acquired
local lithologies are dominated by silicates. In the Paleozoic over the study area on April 24, 2000, using the three mineral
region, some parts (‘‘C’’), represented as magenta, are high indices as for the Yushishan area discussed above. Fig. 7a
both in QI and MI, but low in CI. Others areas (‘‘D’’ –‘‘F’’) shows the geology compiled from a published geological
are represented as bluish, typical for silicate rocks. Also, the map (GSSA, 1965) overlaid on a VNIR false-color image of
Paleozoic region includes a greenish part (‘‘G’’) that we the ASTER scene. Precipitable water at the time of the
expect to be carbonate rocks. ASTER data acquisition was ¨6 kg/m2 as estimated from the
The rock types detected on the basis of the individual archived NCEP Reanalysis data. The Precambrian Adelaide
index values confirm the lithologies predicted from the System is developed well in the study area, with only minor
colors in Fig. 6b. That is, the regions expected from the false exposures of Jurassic and Cretaceous sequences. Fig. 7b, c
colors to be quartzose or carbonate rocks are demonstrated and d present the index images, CI (index values: 1.02–
to have high enough QI (> 1.05) or CI (> 1.045) to qualify. 1.045), QI (1.0 –1.06) and MI (0.8 – 0.9). Some rocks in the
Each pixel thus classified is red or yellow, respectively, in study area have been hydrothermally altered. To locate
Fig. 6c. A secondary threshold on CI (> 1.035 for this alteration minerals exhibiting Al –OH spectral absorption
image), dark yellow in Fig. 6c, complements the detection bands, two additional indices, OHIa and OHIb, were
of carbonate rocks, and a threshold on QI (> 1.03 in this generated (Ninomiya, 2003). OHIa is defined for ASTER
case), dark red in Fig. 6c, is effective for detecting siliceous SWIR data as D 4 * D 7 / D 6 / D 6, where D i is radiance-at-
rocks with relatively high quartz and low feldspar content. sensor data for ASTER band i. OHIa is used to detect minerals
The regions of Paleozoic silicate rocks are classified by MI having an absorption feature at 2.2 Am, typically montmor-
value as follows: the region ‘‘D’’ with average MI å 0.90 is illonite and micas. OHIb is defined for ASTER SWIR data as
expected to be intermediate to mafic; the region ‘‘E’’ with D 4 * D 7 / D 5 / D 5. It is used to detect minerals having an
average MI å 0.87 is expected to be felsic; and the region absorption feature at 2.17 Am, typically pyrophillite. Minerals
‘‘F’’ composed of sub-regions with average MI of 0.87 – with absorption features both at 2.17 and 2.2 Am, typically
0.90, is expected to be mixed felsic and mafic. kaolinite and alunite, are detectable in both indices. The
Most of the region of intrusions is displayed as bluish, results suggest that only altered minerals having absorption
suggesting silicate composition. The analyses on the colors feature at 2.2 Am occur in the Mt. Fitton study area. Together
combined with the relative tone in MI gray-scale image with the geologic map, this suggests that the detected
shown in Fig. 6c, clearly indicate the different rock types. alteration minerals are mostly micas. Pixels of alteration
Both units ‘‘H’’ and ‘‘I’’ are mapped as felsic intrusions (Fig. minerals (OHIa > 4.0) are displayed as cyan in Fig. 7f.
6a); however, ‘‘H’’ appears darker in the MI image (Fig. 6c), Alteration occurs in a variety of Precambrian sequences
indicating a higher chemical SiO2 content than unit ‘‘I.’’ and intrusions.
Many veins in unit ‘‘H’’ with relatively high MI values (i.e., The region ‘‘A’’ is expected from its high CI values to be
low SiO2 content) are recognized as linear features in the MI carbonate (Fig. 7b). ‘‘A’’ is displayed as cyan in Fig. 7e.
gray scale image, which is consistent with the field Usually, pure carbonate shows high CI and low QI and MI,
observation. For the analysis based on the value of MI, the but in this case it shows relatively high MI (Fig. 7d). This
gray scale of MI (Fig. 6c) was set between 0.8 (black) and 0.9 implies that carbonate and mafic minerals or rocks occur
(white), and the pixels with MI > 0.92 (colored purple) are together in the region, consistent with the published
considered to be ultramafic rocks. MI values for both units geological map and field observations of talc and tremolite
‘‘H’’ and ‘‘I’’ are < 0.9, indicating felsic to intermediate there. Other thin or small units are expected from their CI
composition. The average MI value for unit ‘‘H’’ is ¨0.85, values to be carbonates. The regions labeled ‘‘B’’ are an
and the value for unit ‘‘I’’ is ¨0.875, which suggests that the example; the southern region ‘‘B’’ is at Wildman Bluff (Fig.
chemical SiO2 content in ‘‘I’’ is the lower. Unit ‘‘K,’’ mapped 7a). Marginally high CI values for pixels in lines repre-
as ultramafic intrusions, is well-detected with MI > 0.92, and sented by ‘‘C’’ indicate the existence of carbonate-rich
unit ‘‘J,’’ mapped as mafic intrusions, is also well-detected layers unresolved in the 90-m ASTER-TIR pixels. Carbo-
with MI > 0.90. Unit ‘‘J’’ is displayed as white in Fig. 2c. MI nate content in the layers may be low. High CI values in
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

‘‘D’’ indicate the presence of some carbonate; however, regions seem to be in Cenozoic deposits; however, the
consideration of all the index values for ‘‘D’’ indicates locations of the regions in general coincide with Mesozoic
silicate composition. formations. Regions labeled ‘‘F’’ have marginally high QI
Several small regions (‘‘E’’ in Fig. 7c) with QI values values, but the composition cannot be specified without
> 1.05 display as reddish in Fig. 7e, indicating that they are consideration of the other index values discussed below.
quartz-rich and feldspar-poor stone or sand. Some of the The brightness of the MI image in regions ‘‘G’’ and ‘‘H’’ is

Fig. 7. (a) A compiled geological map overlaid on an ASTER VNIR false-color composite image of the Mt. Fitton study area. Abbreviations for map units: Pw,
upper Proterozoic Wilpena group in Adelaide system; Pu, lower Proterozoic Unberatana group in Adelaide system. (b) Gray-scale image of CI, linearly
stretched to display values from 1.02 to 1.045. Alphabetic labels identify the targets of discussion in the text. (c) Gray-scale image of QI, linearly stretched to
display values of 1.0 to 1.06. Alphabetic labels identify the targets of discussion in the text. (d) Gray-scale image of MI, linearly stretched to display values of
0.8 to 0.9. Alphabetic labels indicate the targets of discussion in the text. (e) Color-composite image of the indices: QI = red, CI = green, and MI = blue. Index
values have been linearly scaled to display 99% of the histogram for each color. (f) Pixels detected with the indices: red, quartzite (QI > 1.05); dark red, siliceous
rock (QI > 1.04); yellow, carbonate rock (CI > 1.045); dark yellow, possible carbonate rock (CI > 1.04); cyan, Al – OH bearing altered rock (OHIa > 4.0); pink,
quartz-rich Al – OH bearing altered rock (QI > 1.04 and OHIa > 4.0). Display is of MI image with a fixed gray-scale range of 0.8 to 0.9.
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

Fig. 7 (continued).

similar, which suggests that the chemical SiO2 contents of some of the regions in ‘‘J’’ has been confirmed at the field and
the surface rocks exposed there are also similar. From the is consistent with the remote-sensing assessment. Also,
MI values (0.85 – 0.86), it appears that the rocks are felsic several thin layers represented by ‘‘K’’ are expected to be
silicates. On the other hand, the QI values differ: for ‘‘G,’’ relatively mafic. The region ‘‘I’’ has relatively high MI
QI å 1.005 to 1.015, whereas for ‘‘H’’ QI å 1.02 to 1.03. values, indicating relatively low SiO2 contents compared to
This indicates that the rock types are different, even if the massive units such as ‘‘G’’ and ‘‘H.’’ It is not certain if the
chemical SiO2 content is the same. We interpret ‘‘G’’ to high-frequency textural features in the CI and MI images at
contain felsic rocks rich in K-feldspar, such as granite, region ‘‘I’’ are topographic artifacts, or if they reflect the
whereas ‘‘H’’ may contain acidic rocks poor in K-feldspar. complicated distribution of carbonate and silicate minerals
This interpretation agrees well with the geologic map. there.
There are several small regions like ‘‘J’’ with very high The joint analysis of the different mineral indices applied
values of MI. Values of ¨0.89 to 0.90 indicate intermediate to to Fig. 7e revealed areas having MI values of 0.8– 0.9 that
mafic silicate rock composition. The presence of tremolite in subdivided the region ‘‘F,’’ with relatively high QI values,
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

into two sub-regions. The eastern part in the lower in the discussion below. QI values >1.05 characterize pixels
Proterozoic sequence (Pu1) seems to be siliceous, and the showing the outcrop in T3. These appear red and are labeled
western part consist of upper Proterozoic sequences (Pu3, ‘‘A’’ in Fig. 8c. The high QI values indicate almost pure
Pu4, Pw1 and Pw2) seems to be silicate rocks with lower quartz rock. The QI values around the outcrop itself are
SiO2 content. lower, which from the index values nevertheless appears to
The latest formation of Proterozoic (Pw3) in the south- be silicate. The region K2 (‘‘B’’) shows many bright and
western part of the study area, around Quartzose Peak (Fig. dark small flecks in the CI image, and small color patches in
3a), is indicated to be quartzose or siliceous rocks, with red the color-composite image (Fig. 8b). Further investigation is
pixels (QI > 1.05) or dark red pixels (QI > 1.04). A part of the necessary to understand the complicated lithologic informa-
region near ‘‘Dingo Hill’’ (Fig. 3a) appears to be mica-rich, tion represented here. Alternatively, the pattern may result
from high values of OHIa and QI. These pixels are pink in from some kind of topographic artifact. Comparing the MI
Fig. 3e. Most of the other regions in Pw3 seem to be silicate image to the VNIR image indicates that most of the outcrops
rocks with relatively high SiO2 content. Comparing the in the northern part in K2 have relatively high values of MI,
images of the indices in and around the detected altered indicating high mafic contents. Probably the source of these
regions shows that some areas appear from QI values to be sedimentary rocks is the nearby mafic– ultramafic rocks.
quartz-rich, but the remote-sensing indices alone are not Regions in the ophiolitic belt (‘‘C’’) shown as white in
sufficient to determine whether the quartz is from source Fig. 8c have MI values > 0.97 and correlate well to the
rock or generated by hydrothermal silicification processes. mapped ultramafic rocks (Fig. 8a). Some of the ultramafic
regions (‘‘D’’) in the geological map have lower values of
3.3. Xigaze study area MI. In part of one of the westernmost regions ‘‘D’’ (Fig. 8a),
possible carbonate rocks appear yellow (CI > 1.045) in Fig.
The study area is located on the Xigaze segment of 8c. This occurrence is not explained in the published
Yarlung Zangbo ophiolite belt, southern Tibet, China, from geological map. Region ‘‘E,’’ which has MI values lower
29-00Vto 29-20VN and from 88-45Vto 89-30VE. The elevation than for ultramafic rocks but high enough (MI > 0.9) for us
of this area ranges from 3700 to 5000 m asl. Xigaze has a to expect mafic rock compositions, agrees well with the
warm, semi-arid monsoon highland climate and vegetations distribution of mafic rocks in the geological map. A part of
are sparsely distributed along the river valleys. Short grasses region ‘‘E’’ has relatively high CI values (> 1.04) indicated
sparsely cover the mountain regions. Two ASTER images of as dark yellow in Fig. 8c. We interpret this to indicate
the Xigaze area were analyzed. The image of the western part carbonate content. This possibly reflects the carbonate
of the study area was acquired on December 13, 2001; the concentrations in pores in the basalt rocks occurring in
image of the eastern part was acquired on November 1, 2000. ‘‘E’’ that we observed in the field. There are several
Fig. 8a shows the compiled geological map (Wang et al., ultramafic or mafic layers detected in the MI index image.
1984) overlaid on the mosaicked ASTER VNIR false-color Some, labeled ‘‘F,’’ are not described in the published
images. Precipitable water at the time of the ASTER data geological map. The regions of radiolarian cherts in J3 –K1
acquisition is estimated from archived NCEP Reanalysis data have been identified as units with QI > 1.035, the pixels of
to have been nearly 0 kg/m2 for the western scene, and 3 kg/ which appear dark red or red in Fig. 8c. These are labeled
m2 for the eastern scene. The Xigaze ophiolite represents a ‘‘G.’’ The extent of the units may be grasped intuitively with
peculiar oceanic lithosphere, comprising from north (top) to the color-composite image of the indices (Fig. 8b).
south (bottom) marine sediments in stratigraphic contact over
pillow lavas or lava flows, to fresh harzburgites and
lherzolites (Nicolas et al., 1981). It is bounded by Upper 4. Discussion
Cretaceous flysch (K2) in the north and by Upper Triassic
flysch (T3) or Upper Jurassic – Lower Cretaceous abyssal The case studies reported here are at different elevations
sediments and basic lava (J3 – K1) in the south. The J3 – K1 and present a set of examples that demonstrate the stability
sequence along the boundary with ultramafic unit partly of the mineral indices to temperature and atmospheric
consists of radiolarian cherts. changes. The stability of the indices, especially CI, to
The labels on the mosaicked color-composite image of temperature is accomplished by normalizing the radiance-at-
the indices (Fig. 8b) together with the MI image scaled 0.85 sensor data to a fixed temperature as described in Section
(black) to 0.95 (white) (Fig. 8c) show locations of features 2.4. Analyses of the behavior of the indices with respect to

Fig. 8. (a) A compiled geological map overlaid on an ASTER VNIR false-color composite image of the Xigaze study area. Abbreviation of map units: T3,
Upper Triassic rocks; J3 – K1, Upper Jurassic to Lower Cretaceous rocks; K1, Lower Cretaceous rocks; K2, Upper Cretaceous rocks; E, Lower Tertiary rocks.
(b) Color-composite image of the indices: QI = red, CI = green, and MI = blue, linearly scaled to cover 99% of the histogram for each color. The alphabetic labels
indicate the targets of discussion in the text. (c) Pixels detected with the indices: red, quartzite (QI > 1.05); dark red, siliceous rock (QI > 1.035); yellow,
carbonate rock (CI > 1.045); dark yellow, possible carbonate rock (CI > 1.04). Display is of MI image with a fixed gray-scale range of 0.85 to 0.95.
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

atmospheric parameters indicated that the dominant atmos- expect revisions to the ASTER Level-1 standard data-
pheric factor affecting the indices is water-vapor content. processing software will reduce the amount of striping.
The relative insensitivity of the indices to temperature and We find that it is better to use ASTER data resampled
atmospheric water vapor suggests that fixed threshold with a nearest-neighbor (NN) algorithm, rather than bilinear
values for the indices can be applied widely to detect rock or cubic-convolution algorithms, for applications using
types, provided that atmospheric water-vapor content is low spectral data. However, NN resampling in ASTER often
enough. In the case studies presented above, precipitable cause curved linear and zonal artificial patterns in the index
water-vapor content was estimated to have been <10 kg/m2 images. We attribute this noise to some error in the ASTER
according to NCEP Reanalysis data. standard processing code. Because of this problem, it may
We conclude that: be better for now to use convolution re-sampling.
Carbonate rocks composed of calcite and/or dolomite are
well-detected by the CI index and exceed a threshold value
of 1.045. A secondary threshold of CI å 1.035 (the value is 5. Conclusions
determined case by case) is useful in complementing the
detection of carbonate rocks. It should be cautioned that The Carbonate Index (CI), Quartz Index (QI) and Mafic
index values for clouds or anomalous atmospheric regions Index (MI) defined by emphasizing the diagnostic features
overlap with values for carbonate rocks. Even thin clouds found in the TIR spectral emissivity data measured for
undetectable in VNIR images can introduce ambiguity into corresponding rock types are applied to the ASTER Level-
the CI image, so that analysis of multiple images of the 1B and Level-3A radiance-at-sensor data scenes observing
same area is desirable. selected study areas in China and Australia. The conceptual
Quartzite or quartzose rocks are detected using QI with a lithologic mapping approach proposed here is applicable to
threshold value of 1.05. A secondary threshold of radiance-at-sensor data without atmospheric corrections and
QI å 1.035 (case by case) is useful in detecting siliceous has advantages over ones that require as input the
rocks high in quartz but low in K-feldspar. It should be atmospherically corrected spectral emissivity data produced
noticed that rocks rich in both quartz and K-feldspar have by the ASTER Temperature-Emissivity Separation (TES)
low QI values, which is usual for igneous rocks such as algorithm. Our results indicate the robustness of the index
granite, but rare for sedimentary rocks. QI usually functions approach in detecting rock types in spite of climate and
well as an indicator of quartz-rich sedimentary rocks. elevation differences for each case study. This suggests the
Mafic rocks (here, silicate rocks with less than 60% potential of this approach for global lithologic mapping in
chemical SiO2 content) are well-detected using MI > 0.90. arid and semi-arid regions on a standardized basis,
Ultramafic rocks (here, silicate rocks with less than 50% especially in areas where field exploration is difficult for
chemical SiO2 content) are well-detected using MI > 0.92. political and geographical reasons.
MI å 0.89 is a stable boundary between mafic and felsic
rocks. There is little effect on spectral contrast by extraneous
factors because the blackbody spectrum yields similar Acknowledgements
values in MI as for intermediate rocks (Ninomiya, 2002),
as indicated in Fig. 4. MI values for vegetation, a typical This study was partly funded by the ITIT project of
graybody, are similar to values for intermediate rocks; AIST, Japan and Key Laboratory of Gas Geochemistry,
therefore, analysis is required to classify rock types with MI Lanzhou Institute of Geology, Chinese Academy of
in the presence of vegetation. Sciences, and benefited from discussion with the ASTER
Complicated spatial patterns of index values appear as science team members. The spectral emissivity data were
small patches in the images for some regions in the case measured with the Micro Fourier Transform Interferometer
study areas. Further investigation is needed to determine (AFTIR) owned by Earth Remote Sensing Data Analysis
whether these reflect complicated lithologic settings, or Center (ERSDAC). We thank Drs. Yutaka Takahashi,
image artifacts. Keyang Ma and Maobing Wu for their help during field
Artifactual cross-track striping in ASTER-TIR images is investigations. B. Fu would like to acknowledge the
emphasized in index images (CI, QI, MI). Striping support from Postdoctoral Fellowship (P 03150) of Japan
originates from the systematic noise of ASTER Level-1 Society for the Promotion of Science (JSPS) during the
data processing. That is, residual calibration errors in the paper work.
corrections for detector sensitivities, together with array and
chopper temperature calibration, lead to the striping.
Detecting rock types with the fixed thresholds from the
References
indices is confirmed to be feasible. However, sometimes the
striping makes it difficult to represent details in the spatial Bureau of Geology and Mineral Resources of Gansu Province (BGMRGP)
distribution of rock units. Rowan and Mars (2003) also (1989). Regional geology of Gansu Province. Beijing’ Geological
discussed this problem affecting ASTER-TIR data. We Publishing House. 692 pp. (in Chinese with English summary).
Y. Ninomiya et al. / Remote Sensing of Environment 99 (2005) 127 – 139

Crowly, J. K., & Hook, S. J. (1996). Mapping playa evaporate minerals Nicolas, A., Girardeau, J., Marcoux, J., Dupre, B., Xiao, X., Chang, C., et
and associated sediments in Death Valley, California, with multi- al. (1981). The Xigaze ophiolite: A peculiar oceanic lithosphere.
spectral thermal infrared images. Journal of Geophysical Research, Nature, 294, 414 – 417.
101, 643 – 660. Ninomiya, Y. (1995). Quantitative estimation of SiO2 content in igneous
Cudahy, T. J., Okada, K., Yamato, Y., Huntington, J. F., & Hackwell, J. A. rocks using thermal infrared spectral with a neural network approach.
(2000). Mapping skarn alteration mineralogy at Yerington, Nevada, IEEE Transactions on Geoscience and Remote Sensing, 33, 684 – 691.
using airborne hyperspectral TIR SEBASS imaging data. ERIM Ninomiya, Y. (2002). Mapping quartz, carbonate minerals and mafic –
Proceedings of the 14th International Conference on Applied Geologic ultramafic rocks using remotely sensed multispectral thermal infrared
Remote Sensing (pp. 70 – 79). ASTER data. Proceedings of SPIE, 4710, 191 – 202.
Cudahy, T. J., Whitbourn, L. B., Connor, P., Mason, P., & Phillips, R. N. Ninomiya, Y. (2003). Rock type mapping with indices defined for
(1999). Mapping surface mineralogy and scattering behaviour using multispectral thermal infrared ASTER data: Case studies. Proceedings
backscattered reflectance from a hyperspectral midinfrared airborne of SPIE, 4886, 123 – 132.
CO2 laser system (MIRACO2LAS). IEEE Transactions on Geoscience Ninomiya, Y. (2004). Lithologic mapping with multispectral ASTER TIR
and Remote Sensing, 37, 2019 – 2034. and SWIR data. Proceedings of SPIE, 5234, 180 – 190.
Fu, B., & Chou, X. (1998). Thermal infrared spectra and TIMS imagery Ninomiya, Y., & Fu, B. (1999). Potential applicability of ASTER thermal
features of sedimentary rocks in the Kalpin Uplift, Tarim Basin, China. infrared multispectral data on estimation of SiO2 content in surface
Geocarto International, 13, 69 – 73. rocks. Journal of Remote Sensing Society of Japan, 19, 102 – 115.
Fujisada, H. (1995). Design and performance of ASTER instrument. Ninomiya, Y., & Fu, B. (2002). Quartz index, carbonate index and SiO2
Proceedings of SPIE, 2583, 16 – 25. content index defined for ASTER TIR data. Journal of Remote Sensing
Geological Survey of South Australia (GSSA) (1965). S. A. geological Society of Japan, 22, 50 – 61.
atlas, 1:250,000 series, Marree, sheet H54-5. Ninomiya, Y., & Fu, B. (2003). Extracting lithologic information from
Hook, S. J., & Kahle, A. B. (1996). The micro Fourier transform ASTER multispectral thermal infrared data in the northeastern Pamirs.
interferometer (AFTIR)—A new field spectrometer for acquisition of Xinjiang Geology, 21, 22 – 30.
infrared data of natural surfaces. Remote Sensing of Environment, 56, Rowan, L. C., & Mars, J. C. (2003). Lithologic mapping in the Mountain
172 – 181. Pass, California area using Advanced Spaceborne Thermal Emission
Hunt, G. R., & Salisbury, J. W. (1974). Mid-infrared spectral behavior of and Reflection Radiometer (ASTER) data. Remote Sensing of Environ-
igneous rocks. Air force Cambridge research laboratory technical ment, 84, 350 – 366.
report, TR-74-0625. 142 pp. Salisbury, J. W., Walter, L. S., & D’Aria, D. (1988). Midinfrared (2.5 to 13
Kahle, A. B., & Goetz, A. F. H. (1983). Mineralogic information from a Am) spectra of igneous rocks. USGS open file report (pp. 88 – 686).
new thermal infrared multispectral scanner. Science, 222, 24 – 27. Salisbury, J. W., Walter, L. S., Verg, N., & D’Aria, D. (1992). Infrared (2.1
Kahle, A. B., Madura, D. P., & Soha, J. M. (1980). Middle infrared to 25 Am) spectra of minerals. Baltimore’ The Johns Hopkins
multispectral aircraft scanner data: Analysis for geological applications. University Press. 267 pp.
Applied Optics, 19, 2279 – 2290. Wang, X., Xiao, X., Cao, Y., Zheng, H. (1984). Geological map of the
Kahle, A. B., & Rowan, L. C. (1980). Evaluation of multispectral middle ophiolite zone along the middle Yarlung Zangbo (Tsangpo) river,
infrared aircraft images for lithologic mapping in the East Tintic Xizang (Tibet). Publishing House of Surveying and Mapping, Beijing.
Mountains, Utah. Geology, 8, 234 – 239. Yamaguchi, Y., Kahle, A. B., Tsu, H., Kawakami, T., & Pniel, M. (1998).
Kneizys, F. X., Shettle, E. P., Abreu, L. W., Chetwynd, J. H., Anderson, Overview of Advanced Spaceborne Thermal Emission and Reflection
G. P., & Gallery, W. O., et al. (1988). User’s guide to LOWTRAN, Radiometer (ASTER). IEEE Transactions on Geoscience and Remote
Vol. 7. Air Force Geophysics Laboratory. AFGL-TR-99-0137. Sensing, 36, 1062 – 1071.
Lyon, R. J. P. (1965). Analysis of rocks by spectral infrared emission (8 to
25 microns). Economical Geology, 60, 715 – 736.

View publication stats

You might also like