Sunteți pe pagina 1din 78

EXPLORATION AND MINING

EXPLORATION AND MINING REPORT 734R

MAPPING SKARN AND PORPHYRY ALTERATION


MINERALOGY AT YERINGTON, NEVADA, USING
AIRBORNE HYPERSPECTRAL TIR SEBASS DATA

T.J. Cudahy1, K. Okada2, Y. Yamato2, M. Maekawa3,


J.A. Hackwell4 and J.F. Huntington5.

1
CSIRO Exploration and Mining, Floreat Park, WA, Australia
2
Sumitomo Metal Mining Ltd, Shimbashi, Tokyo, Japan
3
ERSDAC, Kachidoki, Tokyo, Japan
4
The Aerospace Corporation, El Segundo, CA, USA
5
CSIRO Exploration and Mining, North Ryde, NSW, Australia

March 2000

CSIRO Exploration and Mining


Underwood Avenue, Floreat Park, WA, Australia, 6014
ii CSIRO Exploration and Mining Report 734R
CONTENTS

SUMMARY ............................................................................................................ 1

1. INTRODUCTION ..............................................................................................5
1.1 Hyperspectral Thermal Infrared Mineral Mapping .................................................. 5
1.2 Skarns .................................................................................................................... 7

2. OBJECTIVES...................................................................................................8

3. YERINGTON TEST SITE .................................................................................8


3.1 Location and Physiography .................................................................................... 8
3.4.1 Field photographs......................................................................................................10
3.2 Geology ................................................................................................................ 12
3.3 Structure............................................................................................................... 14
3.4 Skarn Alteration.................................................................................................... 14
3.5 Skarn Hydrothermal Evolution.............................................................................. 16
3.6 Porphyry Alteration............................................................................................... 16
3.7 Porphyry Hydrothermal Evolution......................................................................... 17
3.8 Mineralisation ....................................................................................................... 18

4. TIR SPECTRAL MINERALOGY ....................................................................18


4.1 Introduction........................................................................................................... 18
4.2 Garnets................................................................................................................. 19
4.3 Pyroxenes ............................................................................................................ 20
4.4 Feldspars.............................................................................................................. 20

5. INSTRUMENTATION.....................................................................................22
5.1 SEBASS ............................................................................................................... 22
5.2 AVIRIS.................................................................................................................. 24
5.3 MicroFTIR............................................................................................................. 24
5.4 PIMA-II ................................................................................................................. 24
5.5 X-Ray Diffraction .................................................................................................. 24
5.6 Electron MicroProbe............................................................................................. 24

6. DATA ACQUISITION .....................................................................................24


6.1 Airborne................................................................................................................ 24
6.2 Field Sampling...................................................................................................... 25

7. SEBASS DATA REDUCTION........................................................................25


7.1 Atmospheric Correction........................................................................................ 25
7.1.1 Introduction ...............................................................................................................25

CSIRO Exploration and Mining Report 734R i


7.1.2 AEROSPACE ISAC In-scene Correction...................................................................27
7.1.3 Validation of the ISAC Results...................................................................................28
7.1.4 Path-Length (Altitude Effects) ...................................................................................32
7.1.5 Scene Atmospheric Heterogeneity .............................................................................33
7.1.6 Surface and Atmospheric Effects ...............................................................................33
7.1.7 Scene-Independent Mineral Identification.................................................................35
7.2 Instrument Noise and Spectral Subsampling........................................................37
7.3 Temperature-Emissivity Separation......................................................................40
7.3.1 Background................................................................................................................40
7.3.1 Is there need for TES? ...............................................................................................41

8. EXTRACTION OF MINERAL INFORMATION............................................... 42


8.1 Introduction ...........................................................................................................42
8.2 Spectral Indices ....................................................................................................42
8.3 Image Endmember Analysis and Partial Unmixing...............................................43
8.4 Comparison of Results .........................................................................................43

9. SEBASS MINERAL MAPS OF SKARN ALTERATION ................................ 45


9.1 Douglas Hill Line...................................................................................................45
9.2 Ludwig Line...........................................................................................................47
9.3 Garnet Solid-Solution Chemistry ..........................................................................49
9.4.1 Laboratory Study .......................................................................................................50
9.4.2 SEBASS Garnet Chemistry Mapping.........................................................................52

10. SEBASS MINERAL MAPS OF PORPHYRY ALTERATION ......................... 54


10.1 Douglas Hill Line .................................................................................................54
10.2 Macarthur Line ....................................................................................................57

11. AVIRIS MINERAL MAPS............................................................................... 59

12. DISCUSSION ................................................................................................. 61


12.1 Mineral Alteration Mapping and Exploration........................................................61

13. CONCLUSIONS............................................................................................. 63

14. ACKNOWLEDGMENTS ................................................................................ 64

15. FURTHER WORK.......................................................................................... 65

16. REFERENCES............................................................................................... 65

17. APPENDICES ................................................................................................ 70


A1. SEBASS Spectral Band Subsetting......................................................................70
A2. XRD Mineralogy....................................................................................................71
A3. Log Residuals .......................................................................................................73

ii CSIRO Exploration and Mining Report 734R


SUMMARY

The potential of hyperspectral thermal infrared remote sensing for enhanced mineral
exploration was evaluated using airborne SEBASS (Spatially Enhanced Broadband Array
Spectrograph System) image data (128 pixel wide swath @ ~3 m pixel resolution)
collected from skarn-type and porphyry-type alteration in the Yerington district, Nevada.
The 124 SEBASS bands between 7.6 and 13.4 µm enabled the spectral identification of a
range of minerals, including: quartz; grandite garnet (andradite to grossular);
clinopyroxene (diopside-hedenbergite); plagioclase feldspar (albite-oligoclase to
oligoclase-andesine); carbonate (calcite and dolomite); epidote; muscovite; kaolinite and
gypsum. Of these minerals, grandite was critical in defining the extent of the exoskarn
alteration while the plagioclase chemistry was critical in defining the calcic versus sodic
alteration associated with the meteoric versus magmatic metasomatism within the
porphyry intrusives.

The grandite Fe-Al solid-solution chemistry was found to be useful in targeting Cu-skarn
mineralisation. This parameter was measured with the SEBASS data using the
wavelength of a diagnostic grandite feature near 11.5 µm that was found in the laboratory
to be correlated with XRD and electron microprobe analyses. The SEBASS image of this
parameter showed that all the intersected Cu skarn-related mines are associated with near
pure andradite (see the figure below).

In addition, the Fe content of grandite associated with the early skarnoid phase increases
south-west from the contact with the granodiorite (main source of metasomatic fluids).
Clinopyroxene was also mapped though the associated laboratory TIR spectra. XRD
results do not conclusively identify diopside from hedenbergite, which also has a Fe-Mg

CSIRO Exploration and Mining Report 734R 3


solid solution relationship, with diopside potentially important for identifying Cu versus
Zn skarns.

The preliminary SEBASS study of the porphyry alteration in the Yerington granodiorite
mapped a zonation pattern comprising: epidote-quartz at the contact with the adjacent
limestones; then andesine-oligoclase feldspar; and finally albite-oligoclase feldspar
consistent with the published mapping, which found a progressive Ca enrichment
(relative decrease in Na) towards the contact with the carbonate wall rocks (epidote and
quartz is possibly indicative of oxidising conditions). The other abundant minerals in the
altered granodiorite, including biotite and hornblende (largely muscovite-poor), were not
recognised in the SEBASS data.

A preliminary comparison of the complementary information available from


hyperspectral AVIRIS VNIR-SWIR and SEBASS TIR airborne data found that, even
though the AVIRIS data could be used to discriminate most of the skarn alteration
zonation, the associated spectral signatures could be used to conclusively identify the
critical mineralogy. This included the alteration mineral groups: grandite; pyroxenes;
feldspars as well as quartz. Chlorite, epidote, amphibole and white micas with different
solid solution chemistries were identified using AVIRIS (as well as dolomite, calcite,
kaolinite and gypsum). The altered granodiorite is essentially white mica poor which
limited the utility of the AVIRIS data for mapping the porphyry alteration.

The SEBASS data were found to be of very high radiometric, geometric and spectral
quality, though strong column-dependent instrument effects are evident in the data,
especially for the shorter wavelength bands. The restriction of conducting blackbody
calibration of the SEBASS data at the beginning and end of each flight-line did not result
in any apparent scene artefacts. The scene-dependent, “InScene Atmospheric Correction”
(ISAC) algorithm proved effective in correcting for both additive (up-welling radiance)
and multiplicative (transmission) atmospheric effects. Minor problems with this ISAC
method include no correction for reflected down-welling radiance, which results in less
surface spectral emissivity contrast, and uncorrected atmospheric-line
absorption/emission lines for “contrasting temperature” pixels. No major problems were
apparent for the lack of a path-length correction or for assuming a homogenous
atmosphere.

4 CSIRO Exploration and Mining Report 734R


1. INTRODUCTION

1.1 Hyperspectral Thermal Infrared Mineral Mapping


Hyperspectral thermal infrared (TIR – 7.5 to 14 µm wavelength region) remote sensing
has rapidly advanced with the recent development of airborne systems like
MIRACO2LAS (Cudahy et al., 1999) and SEBASS (Hackwell et al. 1996). This follows
years of laboratory studies of the TIR spectral properties of geological materials (Launer
1952, Lyon 1965, Hunt and Vincent 1968, Conel 1969, Vincent and Thomson 1972,
Logan et al. 1973, Salisbury et al. 1987, Salisbury and Walter 1989, Nash and Salisbury
1991). These have shown the potential of this of wavelength region for mineral mapping,
especially non-OH-bearing silicate mineralogy. However, capitalising on this potential
using remote (and field) systems has been slow compared with the rapid developments at
shorter wavelengths (visible, near infrared to shortwave infrared or VNIR-SWIR – 0.4 to
2.5 wavelength region). The delay in the development of hyperspectral TIR remote
sensing is for a number of reasons.

First, there have been engineering constraints on remote TIR system design. For
example, there is a requirement for frequent hot and cold blackbody calibration of the
sensor, as fluctuations in instrument temperature will yield different responses for the
same ground target. Another issue relates to the small amounts of electromagnetic
radiation (EMR) available for terrestrial remote sensing at these wavelengths compared
with the VNIR-SWIR. To obtain sufficiently small pixels with sufficiently large signal to
noise ratio (SNR), most multispectral TIR systems based on liquid-nitrogen cooled
mercury cadmium telluride (MCT) detectors have been designed with only a few spectral
bands (2-10) with bandwidths generally greater than 0.5 µm.

Using this low spectral-resolution design, a number of TIR remote sensing systems have
been active in the scientific community for a number of years, including: the 6-channel
TIMS (Kahle et al., 1980; Palluconi and Meeks, 1985); the 6 TIR channel Geoscan Mk
(Agar 1994); the 6 channel DIAS (Chang et al., 1993); and the 10 channel MIVIS
(Bianchi and Marino, 1994). Recently, the satellite-borne ASTER system with 5
calibrated TIR bands was successfully launched and will provide global coverage of
multispectral TIR at 90 m pixel resolution (ASTER, 2000). However, these low spectral
resolution TIR systems do not allow mapping of specific minerals but they do provide
measurement of the bulk SiO2 content (Lang et al., 1987; Sabine et al., 1994; Crowley
and Hook, 1996, Ninomiya, 1995). These systems can be used to discriminate and
separate a wide range of materials but lack the power of detailed spectral identification.

Another challenge confronting passive remote sensing at TIR wavelengths is the


relatively complicated nature of thermal infrared physics. This challenge includes
separating the desired surface emissivity information (related to the surface composition),
which is typically a very small component of the measured signal, from the dominating
surface kinetic temperature information (temperature-emissivity separation or TES).

In addition to atmospheric transmission attenuation, which affects all wavelengths, the


atmosphere generates down-welling (surface reflected - multiplicative) and up-welling

CSIRO Exploration and Mining Report 734R 5


(additive) radiance components at TIR wavelengths. These components both have a
temperature-dependent blackbody shape as well as narrow gas self-emission lines. In the
situation where the sky and ground temperatures are the same, this results in no
“apparent” surface emissivity contrast, essentially because the reflected down-welling
component cancels the emitted surface emissivity component.

Unlike the VNIR-SWIR, the spectral properties of geological materials at TIR


wavelengths are dominated by “reststrahlen” bands (Lyon, 1964). Superimposed on these
reststrahlen bands there can also exist absorption bands, Christiansen Frequencies (Henry,
1948) and Transparency Features (Conel, 1969). That is, in the VNIR-SWIR geological
materials are transparent such that volume scattering dominates, resulting in relatively
“narrow” mineral absorptions superimposed on an “invariant” background. In the TIR,
surface scattering dominates such that the entire shape of the TIR spectrum varies in
accordance with the mineralogy. This surface scattering also means that surface coatings
and particle size are of even greater effect than the mineral absorptions (Lyon 1964,
Aronson et al. 1966, Hunt and Logan 1972, Salisbury and Eastes 1985, Takashima and
Masuda 1987, Wald and Salisbury 1995).

There has been a focus within geological remote sensing to map potassic-propylitic
alteration, typical of porphyry-epithermal alteration (Marsh and McKeon 1983, Fraser et
al. 1986, Feldman and Taranik 1988, Kruse 1988, Rowan et al. 1991, Zaluski et al. 1994)
and volcanic massive sulphide alteration (Cudahy et al., 1999) where OH-bearing
minerals like white mica and chlorite are important and the VNIR-SWIR is optimum for
mapping this information. However, the TIR wavelength region also merits attention for
mineral exploration applications. For example, the TIR can in theory be used to identify
different types of silica, including high temperature/pressure crystobalite and tridymite as
well as low temperature quartz, silica glass (Conel 1969) and possibly different forms of
“hydrated” silica. This is important in separating silica that formed through hydrothermal
alteration from silica formed through weathering processes (Gunnesch et al. 1994).
Garnet is another TIR-active mineral group that is potentially important in skarn systems
(Einaudi and Burt, 1982), high-grade metamorphic SEDEX (Spry and Wonder, 1989) and
diamondiferous kimberlites (Deer et al. 1976). For example, Mn-rich (Ca-poor) garnets
appear to be associated with high metamorphic grade Broken Hill-style base metal
deposits (Spy and Wonder, 1989). In contrast, Ca-rich garnets appear to be associated
with skarn base metal mineralisation in Nevada (Harris and Einaudi 1982). Similarly,
pyroxene composition can be used as an indicator for skarn deposits including the type of
metallogeny (Einaudi and Burt 1982, Nakano et al. 1994). Feldspar mineralogy is
important in mapping potential granite host rocks or the granite source of mineralising
fluids (Ishihara 1981). For example, the Proterozoic Cu-Au-U mineralisation at Olympic
Dam (Haynes et al. 1995), rare earth mineralisation (Pollard 1995) and Archaean lode Au
deposits (Wyman and Kerrich 1988) are all associated with specific types of granites.
There is also the potential for the TIR remote sensing to provide improved mapping of
“difficult” minerals like carbonates, which are often difficult in the SWIR.

Answers to some of these issues were provided with the development of an active
hyperspectral TIR remote profiling system called MIRACO2LAS (Whitbourn et al.
1987). This system has approximately 100 channels between 9.2 and 11.2 µm and was
used to map a range of silicates, including quartz, feldspars (both microcline and Na-
plagioclase), garnets (almandine and spessartine), talc, amphiboles (actinolite and
tremolite), kaolinite, carbonates (dolomite and magnesite) and dry and green vegetation
(Cudahy et al. 1999). The identification of different feldspar minerals allowed mapping

6 CSIRO Exploration and Mining Report 734R


of granite-type whereas the mapping of different types of garnets (solid solution
variations in Fe and Mn) appears to separate the succession of rocks associated with the
“world class” Broken Hill Pb-Zn-Ag deposit from similar sequences in the region.
Importantly, many of the diagnostic spectral features that allowed identification of these
materials are narrow (<0.2 µm), making them difficult to detect with low spectral-
resolution TIR systems like the airborne TIMS and satellite-borne ASTER.
Unfortunately, the restricted wavelength coverage of MIRACO2LAS did not test the full
utility of this wavelength region, especially those wavelengths covering the Christiansen
Frequencies in the 7.5 to 9.2 µm wavelength region.

A passive, airborne hyperspectral TIR imaging system, called SEBASS (Spatially


Enhanced Broadband Array Spectrograph System), was developed by Aerospace
Corporation (Hackwell et al., 1994) and comprises two 128 by 128 area detector arrays
for the 3 and 5 µm and 7.6 and 13.4 µm wavelength regions. The “breakthrough” in
obtaining sufficient signal to noise for narrow spectral bands over a small footprint in an
imaging system was assisted by the use of liquid-helium cooling of the detector arrays.
Thus, SEBASS provided an opportunity to resolve many of the “difficult” issues
described above, including thermal calibration, atmospheric correction, TES, spectral-
resolution and imaging (versus profiling) for enhanced mineral mapping.

In September 1999, SEBASS data were collected from a number of geology test sites in
Nevada as part of a group shoot campaign involving Aerospace, Mackay School of
Mines, Stanford University, Perry and Associates, Sumitomo Metal Mining Company
Limited and CSIRO Exploration and Mining. The test sites included epithermal
alteration in felsic host rocks at Virginia City and Steamboat Springs (near Reno) and
porphyry and skarn alteration near Yerington. This study focuses on the SEBASS data
collected from the skarn alteration and to a lesser extent the porphyry alteration at
Yerington.

1.2 Skarns
Einaudi and Burt (1982) described skarns as being of two types, namely:

(i) Exoskarns: These represent the replacement of carbonates-rich rocks. In Mg-


exoskarns, the typical alteration product is forsterite or its alteration products,
serpentine, diopside and calcite. In Ca-exoskarns, (bulk of the world’s economic
skarns), the typical alteration minerals are garnet and pyroxene (also wollastonite).
The garnets are typically andradite-grossular (often called grandite) (<15 mole%
spessartine + almandine) for all Cu, Zn, Pb skarns whereas Sn and W skarns have
increased spessartine-almandine content. Nakano et al. (1994) proposed that
pyroxene compositions (Mn/Fe ratios and Zn contents) could be used as an
indicator in the classification of skarn deposits. For example, Cu skarns contain
diopside whereas Zn-Pb contain hedenbergite-johannsenite.

(ii) Endoskarn: These represent the replacement of “source” intrusive rocks that are
genetically related to the skarn-forming fluids. This reflects a progressive
addition of calcium with zoning towards the limestone involving biotite to
amphibole to pyroxene to garnet. Any K-feldspar disappears while biotite and
plagioclase remains throughout, except where it is replaced by garnet. In

CSIRO Exploration and Mining Report 734R 7


oxidising conditions, epidote-quartz is favoured while pyroxene-plagioclase is
common in Cu and Pb-Zn skarns.

2. OBJECTIVES

The main aim of this study is to generate a “benchmark” geological case history that
evaluates and demonstrates the potential of hyperspectral TIR remote sensing for
improved mineral mapping, especially for exploration of Cu-skarn mineralisation. The
specific objectives of this study include:

1. To determine the TIR spectral characteristics (and other analytical measurements) of


the skarn alteration and host rock minerals present in the Yerington region;

2. To determine whether these spectral characteristics can be identified and mapped


remotely using the SEBASS imagery;

3. To assess whether TIR remote mapping of critical alteration mineralogy can be used
to target skarn base metal mineralisation;

4. To assess the complimentary role of VNIR, SWIR and TIR wavelengths for mapping
skarn alteration mineralogy;

5. To assess the instrument performance of the SEBASS system (including hot and cold
blackbody calibration);

6. To evaluate the effectiveness of the Aerospace atmospheric correction methodology


(In Scene Atmospheric Correction program or ISAC); and

7. To establish if other preprocessing steps are required before proceeding to the final
step of mineral information extraction from the hyperspectral TIR data (for example,
temperature-emissivity separation).

3. YERINGTON TEST SITE

3.1 Location and Physiography


The Yerington test site (Figure 1) in western Nevada, USA, lies within the western Great
Basin on the east side of the Sierra Nevada. Main access to Yerington is via Highway 50
from the north with a network of unsealed roads/tracks covering the test site to the west of
the Yerington township. The study area is dominated by the north-south trending
Singatse Range, which peaks at 6373 ft (~2000 metres) and is 2000 ft (600 metres) above
the surrounding alluvial plains. The region is semiarid with little vegetative cover
(typically less than 5%).

8 CSIRO Exploration and Mining Report 734R


300000 302000 304000 306000 308000 310000 312000 314000

4330000 4330000

4328000 4328000

4326000 4326000
McA
rthu
r
McArthur # 4324000
4324000
4

4322000 4322000

4320000 4320000
ll
s Hi
u gl a
4318000 Do 4318000
Green

##

Ye Yerington
wood

4316000 rin 4316000


g
3 Gypsum
Ma ton
# so
n 5
#

4314000
Douglas Hill 4314000
# Lu
Casting #
#

dw 1B
ig
1A McConnell
#
#

4312000 4312000

4310000 4310000

4308000 4308000

N
4306000 4306000
300000 302000 304000 306000 308000 310000 312000 314000

# Mine
SEBASS line rhyolitic tuff
alluvium (gravels) quartz monzonite (nonporphyritic)
alluvium (sand/silt) granitie porphyry
basalt quartz monzonite (porphyritic)
alluvium metasedimentary rocks
andesite metavolcanic rocks

Figure 1: Published 1:100 000 scale geology map (NBMG, 1996) showing the
location of the SEBASS lines and the area covered by this study for the high
altitude AVIRIS scene (blue box).

CSIRO Exploration and Mining Report 734R 9


3.4.1 Field photographs

The sparse vegetation cover, high relief and excellent exposure of the area are shown in
Figures 2, 3 and 4. Views of the Casting Copper, Douglas Hill, McConnell, Yerington,
Macarthur and Ludwig mines are provided in Figures 2 and 4. Close-up photos of
selected exposed rock types are presented in Figures 3 and 4.

Figure 2: Field photographs: (a) Casting Copper Mine (302725mE 4313237


mN); (b) Douglas Hill Mine (303362mE 4313555mN); (c) McConnel Mine
(306803mE 4312558mN); (d) Yerington Mine; (e) Macarthur Mine (306634mE
4324548mN); (f) Ludwig Gypsum Mine (302629mE 4314741mN). The locations
of these mines are shown on Figure 1.

10 CSIRO Exploration and Mining Report 734R


Figure 3: Field photographs of different types of outcrop (a) “unaltered” grey limetsone
(302816mE 4313221mN); (b) “altered” buff coloured dolomite (303420mE
4313609mN); (c) epidote alteration (303898mE 4314147mN); (d) diopside alteration
(303225mE 4313360mN); (e) grossular garnet alteration (303764mE 4313830mN); (f)
andradite alteration with malachite-azurite veining (302812mE 4313272mN); (g)
quartzite (302498mE 4314853mN); (h) actinolite alteration (303108 mN 4313364 mN).

CSIRO Exploration and Mining Report 734R 11


Figure 4: Field photographs: (a) fresh outcrop of Northern Batholith
granodiorite (305015mE 4315084mN); (b) Fresh outcrop of quartz monzonite
porphyry dyke (303902mE 4314043mN); (c) Open pit Macarthur mine
(306634mE 4324548mN) and (d) Macarthur mine graded pit-floor surface
(306634mE 4324548mN).

3.2 Geology
The oldest rocks in the Yerington area are metamorphosed volcanic and sedimentary
rocks of Triassic and early Jurassic age (Proffett, 1977). These include quartz monzonites
and granite porphyry as well as the metasediments and metavolcanics in Figure 1. A
major unconformity developed in the Cretaceous and early Tertiary that was followed by
deposition of basal conglomerates and sedimentary breccias as well as Oligocene silicic
ignimbrites (rhyolitic tuff in Figure 1) and late Oligocene-Miocene volcanic extrusives
and intrusives (andesite in Figure 1) with 8-11 Mya basalts (basalts in Figure 1) and
Quaternary deposits (alluvium in Figure 1) (long sentence).

The detailed geology (Figure 5) shows that the Triassic-Jurassic metasediments and
metavolcanics (1800 m thick) include basal arkose sandstones, andesite/dacite tuff and
massive, ridge-forming limestones, which are up to 300 m thick in the Ludwig area
(Harris and Einaudi 1982). These fine to medium grained limestones are white to grey in
colour with small quantities of quartz and pyrite (light blue in Figure 5). This unit is
locally dolomitised where it weathers to a distinctive buff colour. The dolomite occurs
both as stratiform layers and as crosscutting veins, suggesting it is the product of
hydrothermal alteration. The limestone in the Ludwig area grades upwards into thinly
interbedded black silty limestone (100-130 m thick) and contains quartz, calcite,

12 CSIRO Exploration and Mining Report 734R


tremolite. The overlying felsic volcanic unit is very fine grained and comprises quartz,
plagioclase and garnet (1-2% pyrite). Similar rocks are found to the east of Ludwig
associated with the McConnel and Mason Valley mines (Einaudi 1977). Stratigraphically
above these units are evaporative gypsum (Figure 2h) and aeolian quartzites at the top of
the Triassic-Jurassic meta-sedimentary-volcanic sequence. These rocks were part of a
volcanic-arc terrane.

Figure 5: Published 1:24,000 scale geology with the locations of the two SEBASS
lines examined in detail in this study (after Proffett and Dilles, 1984).

CSIRO Exploration and Mining Report 734R 13


The Jurassic-Triassic metasediments were intruded by two batholiths, namely the
Southern and Northern (Yerington) Batholiths (Figure 5). The older Mid-Jurassic (150
Mya) Northern Batholith is a medium-grained, equigranular hornblende-bearing
granodiorite cut by numerous dykes of quartz monzonite porphyry. It is highly altered
and partly mineralised. The primary mineralogy comprises plagioclase (40-50%), K-
feldspar (20%), quartz (10-15%) and mafic minerals (10-20%). The plagioclase is
oligoclase to sodic-andesine. Other igneous rocks in this batholith include diorite, quartz
monzonite, aplite, rhyolite and pegmatite. The contact of this Northern Batholith with the
older metavolcanics and metasediments is highly irregular because of the extensive (by as
much as 500 m) and intense alteration of the granodiorite to endoskarn (Harris and
Einaudi 1982). The younger (Jurassic) Southern Batholith is a homogeneous medium-
grained, equigranular quartz monzonite with a few cross-cutting, quartz monzonite
porphyry dykes (unaltered and unmineralised). Minerals (belonging to?) include:
plagioclase, K-feldspar, quartz, hornblende and biotite. It (?) is weakly altered with no
skarn mineralisation associated (Harris and Einaudi 1982).

There are at least four generations of quartz monzonite dykes exposed in the Yerington
open pit porphyry Cu mine (red in Figure 5), two of which are closely related to the Cu
mineralisation and two which are younger (Harris and Einaudi 1982).

These rocks are overlain by Tertiary Volcanics, including quartz-latite and rhyolitic ash
flows and tuffs, and finally by Quaternary alluvium-colluvium (Figure 5).

3.3 Structure
The structural history of the area following Triassic-Jurassic sedimentation and volcanism
involved early faulting and folding associated with the intrusion of the batholiths. This
culminated in the tilting of the entire packet of rocks by 30° westward prior to the
Oligocene. Since the Miocene, “Basin and Range” faulting further rotated the rocks by
another 50° westward such that the originally horizontal strata are now nearly vertical
with the current exposure representing a north-south section through the entire sequence,
including the Jurassic contact metamorphism and skarn formation.

3.4 Yerington Skarn Alteration


Harris and Einaudi (1982) recognised three stages of alteration in the Ludwig area
(Figure 6). The first involved non-metasomatic metamorphism and was most pronounced
in the andesite tuffs where there was widespread calc-silicate formation. No wollastonite
was formed even though the limestones contained some quartz.

The next two periods of alteration involved metasomatic skarn development, both in the
metasediments/metavolcanics (exoskarn) and in the granodiorite (endoskarn). The depth
of formation of the skarn is estimated to be 2 km (Einaudi, 1977). The early exoskarn
stage (skarnoid) involved the development of fine-grained garnet and subordinate
pyroxene (0-30% - generally diopside, at least in the limestones). In general, the
garnet/pyroxene ratio is 4:1 or greater. In the Ludwig area, the limestone is altered to
within 1 km from the contact with the northern batholith. Skarnoid is preferentially
formed along the limestone-andesite contact with its width decreasing southward (Harris

14 CSIRO Exploration and Mining Report 734R


and Einaudi 1982). The garnet becomes increasingly Fe-rich southwards (Ad41 near
contact with northern batholith to Ad81 near Douglas Hill) whereas the pyroxene is
consistently Fe-poor (averaging Hd13). The underlying andesite tuff was first albitised to
60% plagioclase and 40% quartz followed by garnetisation. The silty limestone
stratigraphically above the skarnoid near Douglas Hill has garnet composition of Ad85.

Figure 6: Published geology and alteration maps of the detailed study area
shown in Figure 5 (after Harris and Einaudi, 1982)

The late metasomatic skarn stage involved the development of extensive areas of
pyroxene and coarse grained garnet skarn accompanied by Cu-Fe sulphides, especially in
the limestone (Harris and Einaudi, 1982). Two types formed:

(i) Mg-rich skarn characterised by low garnet/pyroxene ratios - the pyroxene is


dominantly diopside with garnets of intermediate composition; and

(ii) Fe-rich skarn with high garnet/pyroxene ratios (typically 4/1 to 10/0). The
garnets are isotropic and unzoned and are extremely pure andradite (Ad97-Ad100).

The margins of the Mg skarn appears be associated with veins of tremolite and erratic
serpentine, which finger out into the dolomitised marble. The presence of associated
periclase and monticellite suggest temperatures of greater than 600° C (Harris and
Einaudi, 1982).

CSIRO Exploration and Mining Report 734R 15


The Fe skarns are localised at the contacts between marble and the skarnoid (for example,
Douglas Hill) and marble and silty limestone (for example, Casting Copper and Ludwig)
but always within the calcite bearing marble (Harris and Einaudi, 1982). In the Mason
Valley Mine, both compositions of both the garnets and pyroxenes shift abruptly to higher
Fe contents in the hangingwall skarns at the contact between the early garnet zone
(skarnoid) and the dolomitized limestone. This mineral Fe enrichment, reaching andradite
garnet compositions, was contemporaneous with the chalcopyrite mineralisation. Not
also that the Fe content of the diopside also increases systematically toward the marble
contacts from an average of 36 mole% hedenbergite to 56 mole% hedenbergite. The
early skarnoid development within the limestone was unaffected by this later Fe-skarn
(and Mg-skarn) but locally it is veined by these late skarns.

During this later skarn stage, skarnoid in the silty limestones was converted to idocrase
while andesite tuff, not previously altered, was replaced by epidote, though skarnoid in
andesite tuff was unaffected. There were also two generations of amphibole: (1) early
tremolite and then (2) later actinolite, which was coeval with the Cu mineralisation
(Einaudi 1977). Later phase coarse green salite (up to 10 cm) is restricted to garnet
zones near Mg skarn.

In the Northern Batholith plutonic rocks albite and quartz alteration initially developed,
followed by extensive (hundreds of metres) garnet-pyroxene endoskarn. Mineralogy is
highly varied but was dominated by pale-brown garnet. The timing of this endoskarn is
unclear but was probably during the skarnoid stage (Harris and Einaudi, 1982).

Alteration in the Southern batholith was less pronounced, and limited alteration occurs
within 100 metres from the contact with the sediments/volcanics. The alteration included
formation of pyroxene, chlorite and muscovite (Harris and Einaudi, 1982).

3.5 Skarn Hydrothermal Evolution


Within the sedimentary and volcanic rocks occurs a sequence of early skarnoid formation,
then Mg-rich skarn formation at relatively high temperatures followed by iron-rich skarn
formation, when the bulk of the mineralisation occurred. The general enrichment of Fe
content southward during the skarnoid phase suggests a source of iron associated with the
silica-bearing fluids that emanated from the northern batholith system as the unaltered
metasedimentary rocks contains much less Fe than the Fe skarnoid (Harris and Einaudi,
1982). Overall, the diverse alteration mineralogy assemblages are typical of low XCO2
fluids within an open system leading to highly variable assemblages (Harris and Einaudi,
1982)

3.6 Porphyry Alteration


Porphyry-style Cu mineralisation is found at Yerington (Figure 2d), Macarthur (Figures
2e and 4c) and Ann-Mason (Dilles and Einaudi, 1992). These are all genetically linked
with the middle Jurassic Northern (Yerington) Batholith and are associated with porphyry
dyke swarms. Dilles and Einaudi (1992) recognised that the history of hydrothermal
alteration-mineralisation can be divided into:

16 CSIRO Exploration and Mining Report 734R


(i) Pre-main stage (endoskarn);

(ii) Main stage (propylitic, sodic-calic and potassic); and

(iii) Late stage (sodic, chloritic and sericitic).

The pre-main stage endoskarn developed plagioclase + salite ± garnet in quartz


monzodiorite and was localised close to the metasedimentary wall rocks at palaeodepths
of 3-6 km.

The type of main stage alteration depended on depth of formation. The propylitic
alteration developed albite + epidote + actinolite + chlorite at depths above 4 km.
Stratigraphically below this unit (3.5 to greater than 6 km), the sodic-calcic main stage
alteration resulted in oligoclase + actinolite + sphene, but no sulphides, through the
addition of Na and leaching of K, Fe and Cu). The “crosscutting” potassic main stage
alteration (2.5-4 km depth) is characterised by replacement of hornblende by biotite and
K-feldspar and is associated with Cu mineralisation.

Late stage alteration superimposed over all dykes swarms involved early sodic (albite)
and chlorite and later sericite overprints.

3.7 Porphyry Hydrothermal Evolution

The 6 km vertical reconstruction of the wall rock alteration and hydrothermal fluid flow
paths of the Ann-Mason porphyry copper deposit involves four fluids of different origins
reacting with rocks through alkali exchange and H+ metasomatism (Dilles and Einaudi,
1992), namely:

(i) Low salinity garnet endoskarn fluids (XCO2 ≤0.02) at >500°C;

(ii) Moderate salinity (31-41%) sodic-calic main stage fluids at 250>400°C;

(iii) High salinity (32-62%) potassic main stage fluids at 700 to <400°C; and

(iv) Low-moderate salinity (2-13%) late stage fluids at 200-240°C.

The saline fluids involved in sodic-calcic alteration, were non-magmatic and flowed from
the wall rock into the batholith. These fluids (?) mixed with cooling, saline magmatic
fluids within convective cells coincident with the dyke swarms. Late meteoric fluids
possibly exchanged Na for K and leached Cu.

The Ann-Mason porphyry system at Yerington is characterised by Na-metasomatism


where saline fluids circulated near igneous heat sources (Dilles and Einaudi, 1992).

CSIRO Exploration and Mining Report 734R 17


3.8 Mineralisation
The copper-bearing ore bodies in the Yerington district are essentially of two types
(Moore 1969): (i) contact metamorphic-replacement in carbonate rocks (skarn); and (ii)
porphyry deposits in plutonic rocks. Important skarn deposits include: Bluestone Mine
(400,000 tons of ore); Mason Valley (1,701,794 tons of ore); McConnel Mine (15,402
tons of ore); Ludwig Mine (183,000 tons of ore); Douglas Hill (>68,905 tons of ore);
Casting Copper Mine (454,637 tons of ore). Important Cu-bearing porphyry deposits
include: Yerington (60,000,000 tons of ore); Anne Mason deposit and Macarthur deposit.
There are also economic commercial mineral deposits in the area including the Ludwig
gypsum deposit.

4. TIR SPECTRAL MINERALOGY

4.1 Introduction
Laboratory TIR bidirectional reflectance spectra of selected skarn related pure minerals
show diagnostic signatures related to the shapes and wavelengths of their respective
reststrahlen bands (Figure 7). More is described on the spectral behaviour of garnets,
pyroxenes and feldspars below.

Figure 7: Stack plot of


directional hemispherical
reflectance spectra of the types of
minerals found in the Ludwig
area, Yerington.

18 CSIRO Exploration and Mining Report 734R


4.2 Garnets
In Figure 7, andradite and grossular spectra have the longest wavelength (longer than 9.4
µm) Christiansen Frequencies (reflectance approaches 0 because the refractive indices of
the transmitting medium and the material of interest are the same) compared with all the
other selected silicate minerals. These garnets also show major, broad reflectance peaks
in the 10-12 µm region related to Si-O v3 vibrations, with a pronounced narrow trough
centred near 11.4 µm between them.

Figure 8 presents in more detail, laboratory bidirectional reflectance spectra of selected


library (ASTER 2000) garnet minerals, including the Ca-rich grandites (grossular and
andradite) and the Ca-poor pyralspites (pyrope, almandine and spessartine). Note that
there exists complete solid solution within each of these series, for example, Al-rich
grossular through to the Fe0-rich andradite, though there is less between these two series.
The TIR spectra show a pronounced shift in the position of the Christiansen Frequency
and the other reststrahlen bands by up to 1.0 µm, though all share the same spectral shape.
For example, the Christiansen Frequency shifts from pyrope (8.9 µm) to almandine (9.2
µm) to grossular (9.4 µm) to andradite (9.6 µm) while the minimum between the two Si-
O v3 vibration-related peaks shifts from pyrope (10.6 µm) to almandine (10.8 µm) to
grossular (11.2 µm) to andradite (11.6 µm).

Laboratory bi-directional reflectance - pure samples


3.5
spessartine
3 andradite
mean-normlasied reflectance

Fe-Ca grosssularite
2.5 pyrope
Al-Ca
2
Mn-Al
1.5

1 Mg-Al

0.5

0
7.5 8.5 9.5 10.5 11.5 12.5 13.5
wavelength (µm)

Figure 8: Selected laboratory reflectance spectra of pure garnets (from the


ASTER 2000). Note the variations in wavelength, but similar spectral shapes.

These garnet spectral shifts shown in Figure 8 are related to the garnet solid solution
chemistry. This spectral shift is consistent with a similar pattern in the XRD 2-d spacings
(Table 1). That is, the garnet 420 hkl reflection shows a progressive shift related to the
increase in the combined ionic radii of the cations ((Al + Mg) to (Al + Fe) to (Al +Mn) to
(Ca + Al) to (Ca + Cr) to (Ca + Fe)).

CSIRO Exploration and Mining Report 734R 19


Table 1: XRD 2-d spacings for the 420 hkl reflection of “pure” garnet minerals.

Garnet mineralogy XRD 2-d spacing (420 hkl peak)


Angstroms?
Pyrope 2.562
Almandine 2.569
Spessartine 2.600
grossular 2.647
Manganoan andradite 2.620
Uvarovite 2.684
andradite 2.696

4.3 Pyroxenes
Figure 9 presents laboratory bidrectional reflectance spectra of the three clinopyroxenes,
diopside (CaMg[SiO3]2), hendenbergite (CaFe[SiO3]2) and augite
(Ca[Mg,Fe][SiO3]2{[Al,Fe]2O3}x). There is complete solid solution between the Ca-rich
pyroxenes, diopside (Mg-rich) and hedenbergite (Fe-rich) and even with the less Ca-rich
augite/ferroaugite. The TIR spectra (Figure 9) show diagnostic clinopyroxene TIR
spectral features (emissivity lows) at 8.8, 10.4 and 10.9 µm. However, unlike the garnets
(Section 4.2), there is little evidence from these library spectra that there exist variations
in spectral behaviour related to changes in solid solution chemistry.

0.98

0.96
emissivity

0.94

0.92
Diopside
0.9 Hedenbergite
Augite
0.88
7 8 9 10 11 12 13 14
wavelength (µm)

Figure 9: Calculated emissivity spectra of selected clinopyroxenes derived from


laboratory directional hemispherical reflectance (ASTER 2000) .

4.4 Feldspars
The feldspars are a ubiquitous mineral group in the Earth's crust. There are 11 major
types of feldspar minerals each showing different structure and chemistry, the precise
nature of which is related to the environment of their crystallisation. As a consequence,

20 CSIRO Exploration and Mining Report 734R


feldspar mineralogy is useful in the classification of many rocks, especially igneous rocks
(Deer et al. 1977).

Most feldspars can be considered as part of a ternary system (Figure 10) with the three
apices of this diagram defined by Na2AlSi3O8, K2AlSi3O8, and CaAlSi3O8 1. Members of
the series between the Na and K endmembers are considered alkali feldspars (microcline,
orthoclase, sanidine, anorthoclase) and those between Na and Ca are considered
plagioclase feldspars (with increasing Ca the series comprises albite, oligoclase, andesine,
labradorite, bytownite and anorthite). The alkali feldspars can contain up to 10% Ca in
solid solution whereas the plagioclase can contain up to 10% K.

The mineralogy of the K endmember depends on the temperature of formation and hence
structure of the feldspar. The alkali feldspar that forms at the highest temperatures of
crystallisation is sanidine, which is monoclinic. Orthoclase, which is also monoclinic,
forms at moderate temperatures, while microcline, which is triclinic, forms at the lowest
temperatures. Adularia is a low temperature, K-rich (Or90Ab9An1) feldspar with variable
structure and confined to Alpine vein-type paragenesis (Deer et al., 1977). A more
complex relationship with temperature is found for the plagioclase feldspars with both
monoclinic and triclinic forms possible for all solid solution chemistries.

Figure 10: K-Na-Ca ternary diagram of the feldspar mineral group (after Deer et
al. 1977).

Various workers have measured the TIR spectral properties of feldspar minerals (Lyon
and Burns, 1963; Aronson et al., 1967; Farmer, 1974; Vincent et al., 1975; Nash and
Salisbury, 1991; Thomson and Salisbury, 1993). They have shown that the alkali-
feldspar minerals and plagioclase felspars have diagnostic spectral behaviour, especially
in the 8 to 11 µm wavelength region where there exist changes in the number and position
of the reststrahlen features.

1
Note also that Ba, NH3 and other ions and ion groups can substitute into the structure of the feldspars.

CSIRO Exploration and Mining Report 734R 21


Figure 11 presents directional hemispherical reflectance spectra of selected feldspar
minerals including Na-plagioclase (albite) and the alkali feldspars, microcline (K),
sanidine (K/Na) and orthoclase (K/Na). These spectra were of coarsely crystalline
samples (ASTER, 1999). All show similar a reflectance doublet near 8.5 µm, though at
slightly different wavelengths. The main differences exist in the 9-10 µm region. For
example, microcline and albite show a small peak at 9.2 µm whereas the other two
feldspars show only a trough. Sanidine and orthoclase show a single asymmetric peak
near 9.5 µm. In contrast, microcline has twin peak at 9.5 and 9.8 µm whereas albite has
twin peaks at 9.6 and 10.0 µm.

0.7
9.62 microcline (K)
8.5
0.6
9.98 orthoclase (K/Na)
9.2
0.5
sanidine (K/Na)
reflectance

0.4 albite (Na)


0.3

0.2
9.48
0.1

0
7.5 8.5 9.5 10.5 11.5 12.5 13.5
wavelength (µm)

Figure 11: Directional hemispherical reflectance spectra of selected feldspar


minerals (from ASTER, 2000).

5. INSTRUMENTATION

5.1 SEBASS
SEBASS is a push-broom imaging hyperspectral TIR “dispersive” system that comprises
two helium cooled 128 by 128 detector element arrays (Figure 10) operating within the
2.9 to 5.2 and 7.5 to 13.6 µm wavelength regions (Hackwell et al, 1994). That is, there
are 128 contiguous spectral bands for 128 pixels for each detector array. The angular
resolution of the fore-optics is 1.1 mrad per pixel with a total field of view of 141 mrad
(approximately 8.1°). This yields pixels of approximately 2.5 metres over a 320 metres
swath at an operational flying height of approximately 2000 metres.

The SEBASS system (Figure 11) includes hot and cold blackbodies, which are swung in
front of the fore-optics at the beginning and end of each flight line. Thermal calibration
of the system assumes linear drift between these blackbody measurements.

22 CSIRO Exploration and Mining Report 734R


Figure 12: : Schematic diagram of the SEBASS system.

Figure 13: Photographs of components of the SEBASS system: (a) blackbody


unit; (b) instrument mounting; (c) SEBASS sensor; (d) and (e) electronic
peripherals including data acquisition system.

CSIRO Exploration and Mining Report 734R 23


5.2 AVIRIS
A portion of high altitude AVRIS data (http://makalu.jpl.nasa.gov/aviris.html) was acquired for
the test site from the University of Colorado (Alex Goetz personal communication, 1999).
Although not examined in detail, these 220 channel VNIR-SWIR (0.45 to 2.5 µ) image
data were used to briefly examine the complementary role of the VNIR-SWIR (AVIRIS)
and the TIR (SEBASS) wavelengths, although there was a large disparity in pixel size
(2.5 m SEBASS versus 30 m AVIRIS).

5.3 MicroFTIR
The emissivity of field samples was measured using a portable FTIR spectroradiometer
designed and built by Designs and Prototypes (Hook and Kahle, 1996; Korb et al., 1996).
This instrument measures emissivity in the 3-5 and 8-14 µm wavelength region at
approximately 6 wavenumber resolution. The area sensed is approximately 50 mm
diameter. A correction for sky down-welling irradiance is implemented using a gold, zinc
or aluminium reference plate.

5.4 PIMA-II
The 1300-2500 nm directional-hemispherical reflectance of samples was measured using
an Integrated Spectronics field portable PIMA-II spectrometer. The spectral resolution is
approximately 8 nm in the 2.2 µm region. This instrument has its own internal light
source and internal gold reference and plastic wavelength calibration standards. Samples
are measured "in contact" with a transparent sapphire window, the area of which is
approximately 10 mm diameter.

5.5 X-Ray Diffraction


Laboratory X-Ray Diffraction (XRD) analyses of samples was undertaken using a
Phillips PW 1050 Diffractometer with Sietronics instrumentation and CuK α radiation.
Samples were scanned between 2 and 65° (2Θ angle) with a 0.02° step size over a 30
minute scan. XRD data were corrected for any instrument offset using the known peak
positions of constituent mineral, for example, quartz at 3.34 Å.

5.6 Electron MicroProbe


Electron Microprobe analyses of selected garnet samples were conducted in Japan.

6. DATA ACQUISITION

6.1 Airborne
Six lines of SEBASS imagery were acquired from the Yerington district (Figure 1) in
both August 1999 and September 1999. The August campaign was affected by moist,
cloudy conditions with possible wet ground. Inspection of these data revealed problems

24 CSIRO Exploration and Mining Report 734R


consistent with these effects. In contrast, the September data set was acquired under dry,
clear blue sky conditions at approximately midday. Under such “ideal” conditions there
is the least effect of down-welling radiance (Section 7) due to the higher temperature
contrast between the “warm” ground and “cold” sky. Aircraft flying height was
approximately 7000 feet (2000 metres) collecting pixels with a pixel size of ~2.5 over
swath of ~320 metres.

6.2 Field Sampling


A four day field sampling campaign to the Yerington test was conducted in October 1999
by Australian and Japanese scientists. Twin sets of approximately 100 samples from 50
sample localities were collected and shipped back to Australia and Japan for analysis.

7. SEBASS DATA REDUCTION

The data reduction of the SEBASS data involved correction for instrument and
atmospheric effects using software provided by Aerospace Corporation. No details are
provided here (Aerospace response?, is it possible to express it differently assuming Jack
Hackwells later supply of algorithms?) on the instrument correction other than to confirm
that it involved hot and cold blackbody calibration of the data assuming linear drift in
detector responsivity between the beginning and end of each run when such
measurements were taken.

7.1 Atmospheric Correction

7.1.1 Introduction
Atmospheric correction of remotely sensed TIR data should account for:

1. Path length
2. Gas concentrations;
3. Atmospheric layering;
4. Temperature; and
5. Cloud type and position.

The first three parameters affect atmospheric transmission, which comprises:

(i) Atmospheric-line features, which are sharp and related to water vapour and
carbon dioxide, ozone, methane and oxygen (Figure 14); and
(ii) Atmospheric-continuum absorption, which is largely related to water vapour.

Water vapour and the other gases also emit TIR radiation they absorb. The first type of
sky emission is a component of the up-welling radiance, which for a homogenous “warm”
atmospheric profile, results in a Planck “blackbody” radiation function as well as
superimposed atmospheric-line features, the intensity of which depends on the air
temperature and concentration of the gases. The second is down-welling radiance which
is also comprises blackbody behaviour as well as gas emission features, with the
combined effect being reflected from the Earth’s surface, the magnitude of which

CSIRO Exploration and Mining Report 734R 25


depends on the surface reflectance/emissivity (Figure 15). When the sky and surface
temperatures are the same, the reflected signal effectively cancels any surface spectral
contrast. Layered (composition and temperature) atmospheres are more complicated with
absorption/emission occurring with each modelled layer of the atmosphere.

Figure 14: Atmospheric TIR transmittance curves for different concentrations of


water vapour (shown as percentages) calculated using the FASCODE model and
HITRAN database for a 1 km thick column of air (after Hausknecht 1996). Both
the sharp atmospheric-line absorptions and the overall atmospheric-continuum
absorption are apparent.

Figure 15: Spectral plot showing the effects of sky down-welling radiance on
surface emissivity measurements. These passive TIR measurements were taken
under clear sky conditions (source ie instrument?, reference?).

Local sources of “high” radiant temperature, such as low cumulus cloud, will contribute
to the up-welling, if below the aircraft, and also contribute to down-welling radiance,
especially for those pixels near to these sources. Sun angle and cloud shadows do not
contribute directly to the “down-welling” radiance, but they affect the surface kinetic
temperature and hence the surface radiance. Thus, the optimum conditions for TIR data
acquisition are clear, cold skies and a warm, dry surface where there exists a large
thermal contrast between the sky and surface radiant temperatures.

26 CSIRO Exploration and Mining Report 734R


7.1.2 AEROSPACE ISAC In-scene Correction

Aerospace’s In-scene Atmospheric Compensation (ISAC) software was used for


atmospheric correction of the SEBASS data. A brief description of this method is given
here but more details can be obtained from Johnson (1998) and Young (1998).

ISAC is a scene-based correction that uses the SEBASS data itself to derive the required
input parameters for the estimation of atmospheric effects, namely: an additive effect
related to primarily to up-welling radiance (assuming a homogeneous atmospheric
profile); and a multiplicative effect related to attenuation by absorption of gases (related
to path length and gas concentrations). This can be expressed by:

Lobs ( λ , j ) = τ ( λ ) B ( λ , T j ) + Lu ( λ ) 1.
where
Lobs is the observed radiance at sensor;
λ is wavelength;
j is those pixels in the scene with blackbody behaviour;
τ is the atmospheric transmission;
Lu is the up-welling sky radiance;
B is the surface Blackbody radiance; and
Tj is the surface kinetic temperature.

From Equation 1, a scattergram of Lobs (j) versus B (Tj) for a given wavelength, λ, should
yield a linear expression where the slope of the line represents the atmospheric
transmission and the offset the up-welling radiance (Figure 16).

Figure 16: Schematic diagram


showing the theoretical
relationship between observed
and actual pixel radiance at a
given TIR wavelength assuming
blackbody behaviour (from
Young 1998).

To solve for τ and Lu for real TIR passive data, where pixels are not necessarily
blackbodies, the following steps were developed (Johnson, 1998):

1. Calculate and record from each pixel’s radiance spectrum, its maximum brightness
temperature T’i and its wavelength λI where Ti ' = max{TB [ Lobs ( λ , i )]}λ ;

2. Sort the maximum brightness temperatures by wavelength and count the number of
brightness temperatures (“hits”) that accumulate in each wavelength bin;

3. Find the wavelength with the most number of “hits” (maximum hit method) –
reference wavelength (λr);

4. Subset those pixels with a maximum brightness at λr and give them an index “j”;

CSIRO Exploration and Mining Report 734R 27


5. Let λn represent a particular wavelength from the spectrum where λn≠λr;

6. Calculate the Planck Radiation Function, B(λn, T’i), assuming no atmospheric effects
(no gains or offset values) at the reference wavelength for all j;

7. Plot the points defined by the coordinates Lobs(λn, j) and B(λn, T’i). Both the
“unscaled” atmospheric gains and offsets can be calculated from this data cloud
(Figure 17). A straight line resting above of the data cloud is fitted. Departure below
this line is a function of surface emissivity;

Figure 17: The relationship


between observed and actual
pixel radiance at a given TIR
wavelength with departure from
ideal blackbody behaviour
caused by emissivity contrast
(from Young 1998).

8. An additional transformation can be applied to convert the unscaled gains and offsets
to units that are less scene-dependent (closer to actual transmissivity and sky
emission) using scene derived estimates of water line absorption features (Young,
1998).

The assumptions inherent in the ISAC method include:

1. Scene-based and physically-based with no reliance on a radiative transfer model;

2. One gain and one offset for each wavelength;

3. No correction for heterogeneous atmosphere;

4. No correction for variations in path-length;

5. No correction for reflected down-welling sky radiance (Figure 15);

6. No compensation for atmospheric continuum-absorption (assumes reference


wavelength has zero offset and unity gain).

7.1.3 Validation of the ISAC Results

The ISAC atmospheric correction was applied to all the SEBASS lines to yield apparent
surface radiance. The result of this correction is shown in Figure 18, which compares the
radiance at sensor and ISAC corrected surface radiance spectra of a given pixel. This
shows that most, if not all, of the atmospheric-line absorptions have been removed

28 CSIRO Exploration and Mining Report 734R


successfully, for example, the major water vapour feature at 12.54 µm. It also shows that
the corrected spectrum appears to approximate a Planck Blackbody Radiation function.

1300

1200

1100
radiance units

1000

900

800 12.54 µm

700

600
radiance at sensor
radiance at surface
500
7.5 8.5 9.5 10.5 11.5 12.5 13.5
wavelength (µm)

Figure 18: Comparison of the radiance at sensor versus the radiance at surface
spectra for a selected SEBASS image pixel selected from the Ludwig line.

The estimated atmospheric transmittances (gains) and up-welling radiances (offsets)


derived using the ISAC method show a high degree of correlation (Figures 19 and 20).
This is similar to the results of Johnson (1998) and is consistent with the fact that
selective absorbers in the atmosphere (with respect to wavelength) are also selective
emitters.

1.5
vakue

0.5
atmopsheric gains
atmospheric offsets

0
7.5 8.5 9.5 10.5 11.5 12.5 13.5
wavelength (µm)

Figure 19: Scattergram showing the gains and offsets calculated using ISAC for
the Ludwig SEBASS line.

CSIRO Exploration and Mining Report 734R 29


1.2

0.8
atmospheric offsets

y = 1.0018x-0.6887
0.6
R2 = 0.9918

0.4

0.2

0
0 1 2 3 4 5 6 7 8
atmospheric gains

Figure 20: Scattergram showing the correlation between the ISAC generated
gains and offsets for the Ludwig line.

To validate the accuracy of ISAC derived surface radiance, ISAC-corrected SEBASS


pixels collected from over the Walker River at Mason (Figure 21) were examined for
blackbody type spectral shape. The hypothesis being that surface radiance from a body of
water should approximate blackbody behaviour.

Figure 21: Field photograph of the Walker River at Mason where the SEBASS
line passed.

The mean SEBASS surface radiance spectrum collected from the Walker River (Figure
21) is most closely modelled using a Planck Radiation blackbody curve for a surface
kinetic temperature of 293K (Figure 22). Overall, the shape and absolute units of the
SEBASS surface radiance data are in close agreement with the modelled response.

30 CSIRO Exploration and Mining Report 734R


1100

1000

Radiant excitance (W/m 2 µm-1 sr-1) * 100 900

800

700 SEBASS radiance at sensor


SEBASS surface radiance
calculated (283 degrees K)
600 calculated (288 degrees K)
calculated (293 degrees K)
calculated (298 degrees K)
calculated (303 degrees K)
500
7.5 8.5 9.5 10.5 11.5 12.5 13.5

wavelength (µm)

Figure 22:Comparison of average SEBASS spectra (radiance at sensor and


surface radiance) collected from pixels spanning a river (30 pixels) near
Yerington (Mason SEBASS run – Figures 1 and 21) with calculated radiance
spectra (at different temperatures) using the Planck Blackbody Radiation
function. Enlarged to read legend

Figure 22 also shows that, in detail, there is spectral mismatch after normalising the
measured response with the modelled response (Figure 23). The closeness of the fit
between the SEBASS spectrum and the calculated blackbody spectrum becomes
increasingly poorer at shorter (<10 µm) and longer (>11.5 µm) wavelengths. It is
unlikely that this is related to non-blackbody behaviour of the river. Instead, this
departure is probably related, at least in part, to uncorrected water vapour continuum-
absorption (Figure 14), which is not corrected in the ISAC method (Figure 18).

1.01

0.99

0.98
apparent emissivity

0.97

0.96

0.95

0.94

0.93

0.92
7.5 8.5 9.5 10.5 11.5 12.5 13.5

wavelength (µm)

Figure 23: Apparent emissivity spectrum for water derived using the calculated
Planck blackbody radiation curve for 293 K and the atmospherically corrected
SEBASS spectrum of river water (Figure 21).

CSIRO Exploration and Mining Report 734R 31


These results do not necessarily represent a problem for extracting mineral emissivity
information, as this slowly-changing, “uncorrected” atmospheric-continuum absorption is
potentially removed through normalising each pixel by its hull (or continuum) spectrum.
This normalisation procedure has the additional advantage of enhancing the higher
frequency emissivity contrast related to the surface composition by removing the
temperature component (Section 7.3).

7.1.4 Path-Length (Altitude Effects)

It was considered important to examine the effect of not correcting for atmospheric path-
length, which is not corrected using ISAC, especially as there is considerable topographic
relief (greater than 500 m) in the study area.

A section of the Douglas Hill line covering a limestone ridge with 200 m of local vertical
relief was selected for this purpose (Figure 24a). Seven training sites of 16 pixels each
and spaced from the top of the limestone ridge and down to the colluvial apron below
were collected. The mean spectra of each of these training sites (Figure 24b) shows that
all the atmospheric gas absorption-line absorptions have the same apparent intensities
across the full spectral interval from 7.8 to 13.4 µm. Only the narrow carbonate
emissivity low near 11.3 µm shows departure between the training sites, though this is
related to the surface composition (carbonate abundance) and not to the atmosphere.
Subsequent ISAC correction of this flight-line confirmed that all atmospheric-line
absorption were effectively removed for the range of altitudes considered.

Figure 24: Composite figure showing the effectivess of the ISAC atmospheric
correction is an area of 200 m local relief). Stacking the training site mean spectra
collected from the side of ridge shows no apparent differences in the relative in the
depths of the atmospheric line-absorptions. That is, local relief variations of 200 m
produce no spurious features in the atmospherically corrected data.

32 CSIRO Exploration and Mining Report 734R


7.1.5 Scene Atmospheric Heterogeneity

At VNIR-SWIR wavelengths, it is now routine to calculate and correct for water vapour
on a per-pixel basis as the atmosphere is typically heterogeneous. This is especially
important as water vapour has a major effect on a given pixel’s spectral behaviour. The
same appears to be true for the TIR (Figure 14). As a consequence, programs like
ATREM (http://www.cires.colorado.edu/cses/atrem.html) have been developed for
VNIR-SWIR data to calculate the water vapour on a per-pixel basis using the measured
depth/s of a number of water vapour line-absorptions.

The possibility of heterogenous atmospheric water vapour distribution in the SEBASS


data was evaluated using a ratio of the raw SEBASS band 11.8144 µm (base of water line
absorption) over band 11.7307 µm (shoulder of water line absorption). This results
(Figure 25) shows that the surface temperature effects still dominates the non-ISAC
corrected radiance data (Figure 25a), confirming the need for the up-welling correction.
Figure 25b shows the same ratio sensitive to water vapour bu applied to ISAC corrected
data. This image shows that surface temperature has been effectively cancelled.

Figure 25: (a) A ratio image of raw


SEBASS bands 11.8144 µm (base of
water line absorption) over 11.7307 µm
(shoulder of water line absorption)
showing that temperature effects are not
removed simply through normalisation;
(b) the same ratio combination of bands
as in (a) except that the SEBASS data
were corrected using the ISAC software
showing that most if not all of the
temperature effects have been removed
indicating that additive (up-welling)
component must first be removed prior
to normalisation. This image also shows
that there is no heterogenous
atmospheric water vapour variation
throughout the subarea, possibly
because of the dominating instrument
effects (see Section ….); and (c) A ratio
image of raw SEBASS bands 11.8144
and 11.8560 µm, both of which are
outside any line absorption showing that
normalisation alone is effective in
removing temperature.

Furthermore, Figure 25b shows that, apart from the enhanced instrument column-striping
noise, there is no apparent spatial variation along the flight-line that can be attributed to
heterogeneous water vapour. This is an important result and suggests there is not a
significant problem in assuming a homogeneous atmosphere with the ISAC correction, at
least for these SEBASS data under these atmospheric conditions.

CSIRO Exploration and Mining Report 734R 33


As a further test of the suitability of the ISAC estimate of the up-welling additive
component, a ratio image of two “close-spaced”, raw SEBASS bands positioned outside
any atmospheric-line absorptions was investigated, namely 11.8144 and 11.8560 µm.
These are located outside major gas-line features. The estimated ISAC offset corrections
for these wavelengths were effectively zero (unity for a blackbody) (Figure 19). The
results (Figure 25c) shows that effectively no residual temperature is apparent, unlike
Figure 25a. That is, there is no significant up-welling additive effect at these
wavelengths outside the atmospheric-line absorptions. Therefore, the additive correction
of ISAC successfully accounts for the up-welling effects of the atmosphere and that this is
critical prior to any normalisation of the data.

7.1.6 Surface and Atmospheric Effects


Figure 25b shows pronounced surface-related spectral contrast. Variations in surface
emissivity alone are not a suitable explanation for this spatial contrast because the
SEBASS bands selected for the normalisation (11.8144 and 11.8560 µm) are so to close
together in wavelength to be able capture significant mineralogical information. To
further explore this issue, the mean raw SEBASS spectra from training sites were
collected from different geological units (Figure 26). These mean spectra show the
relative intensity of the atmospheric-line absorptions change between ROIs, especially at

Figure 26: Composite figure examining the validity of the assumption that the
atmospheric profile is constant along the length of a given SEBASS run. The top
two strips show the training sites with the bottom two plots showing the surface
radiance and the continuum-removed surface radiance (apparent? surface
emissivity).

34 CSIRO Exploration and Mining Report 734R


wavelengths with large emissivity contrast. For example, the garnet rich training site
(green training site and spectrum) has much weaker atmospheric-line absorptions in the
wavelength region of strong emissivity contrast between 10.25 and 12.25 µm. This
indicates a pixel-dependent relationship between the surface emissivity contrast and the
intensity of atmospheric-line features.

The effects of down-welling sky radiance (Section 7.4.1) can explain this apparent
dependence of the measured intensity of the atmospheric line absorptions on the surface
emissivity (Figure 26). That is, down-welling sky radiance, especially gas-line emissions,
has its greatest effect at those wavelengths (and those pixels) that have deep surface
emissivity (reflectance) features, because of greater reflectance (Kirchoff’s Law) of this
incident atmospheric radiation. As a consequence, the significant contribution of the
reflected down-welling atmospheric-line emission features are partially, and in some
cases completely (such as at 11.0 µm in Figure 26), filling in the atmospheric-line
absorption features. This shows that a correction for down-welling radiance would
improve the mineralogy-related spectral integrity.

7.1.7 Scene-Independent Mineral Identification

The suitability of the SEBASS instrument calibration and ISAC atmospheric correction
for scene-independent mineral mapping was assessed through comparison of the reduced
SEBASS spectra with ground spectra for a range of geological materials (Figure 27). The
ground validation data included microFTIR spectra of related field samples and
laboratory DHR spectra of pure minerals from the JHU mineral spectra library. As
recommended in Section 7.4.3, both the airborne and ground spectral data were processed
to remove the “hull” using ENVI’s continuum-removal routine.

The results of the airborne-ground comparison (Figure 27) show that the SEBASS data
show the same diagnostic mineral spectral features as the ground data, though the relative
intensity of the signatures are much less. This can be explained (at least in part) by the
dampening effect of uncorrected down-welling radiance which effectively fills in those
wavelengths that show emissivity/reflectance contrast (Figure 15). As explained above,
this effect is not corrected by ISAC. More importantly, the diagnostic mineral signatures
are clearly evident in the SEBASS spectra. For example, andradite (Figure 27a) shows a
narrow emissivity peak at 11.58 µm, which is positioned between deep reststrahlen
minima. In contrast, calcite shows a sharp emissivity minimum at 11.2 µm which is 100
nm shorter than a similar dolomite reststrahlen minimum (Figure 27b). Diopside (Figure
27c) has emissivity minima at 8.8, 9.3, 10.4 and 10.9 µm. These are better observed in
the plot without the DHR spectra (Figure 27d). The albite SEBASS spectrum (Figure
27e) shows relatively little emissivity contrast compared with the other SEBASS spectra,
though in the plot without the related DHR spectra (Figure 27f), the diagnostic albite
features at 8.7, 9.6, 9.7 and 9.8 µm can be recognised. The epidote spectra show
different wavelengths to the reststrahlen bands, though the diagnostic emissivity minima
are located at approximately 9.2, 10.4 and 11.3 µm with peaks at 9.9 and 10.8 µm.

These excellent results demonstrate that the reduced SEBASS spectra can be used to
accurately map diagnostic mineral signatures, without problems of scene-dependent
artifacts.

CSIRO Exploration and Mining Report 734R 35


Figure 27: Composite plots of SEBASS (blue and/or magenta) and field microFTIR
emissivity (mauve, green and/or brown) spectra and laboratory directional
hemispherical reflectance (red) spectra of (a) andradite, (b) dolimite and calcite, (c)
pyroxene (diopside); (d) pyroxene with enhanced y-scale; (e) albite (f) albite with
enhanced y-scale, (g) epidote, and (h) epidote with the y-scale enhanced.

36 CSIRO Exploration and Mining Report 734R


7.2 Instrument Noise and Spectral Subsampling
The signal to noise ratio (SNR) of the SEBASS system at most (all) TIR wavelengths is
typically considerably higher than 1000:1 (Hackwell, personal communication 1999). It
is difficult to measure precisely the effective SNR outside the laboratory (in effect, the
NE∆T is measured in the laboratory) however inspection of the derived ground spectra
(Figure 27) indicates the spectral data are of very high quality. Often the calculation of
scene statistics, like scene/column means and standard deviations (STD), provides
information about the relative nature of the noise, including whether it is systematic or
random, as well as whether it is multiplicative or additive.

The mean and the STD of one SEBASS line are shown in Figure 28. The mean spectrum
shows the characteristic Planck Blackbody Radiation Function shape. The STD spectral
shape is difficult to interpret mainly because of the complicating and dominating effects
that surface temperature has on passive TIR signal. That is, the largest temperature
variations theoretically occurs at wavelengths <10µm for surface temperatures higher
than 293K (Figure 22). At these wavelengths, there occurs the greatest variation in
surface silicate emissivities such that the high STDs at these wavelengths (and for the
other wavelengths) cannot be uniquely related to instrument noise. The apparently
greater band-to-band variation at shorter (<9 µm) and longer (>11.5 µm) wavelengths is
more likely a function of the intensity of atmospheric-line absorptions at these
wavelengths (Figure 18).
1400
1300
1200
1100
radiance units

1000
7.832
900
7.574
800
700
600 Mean 12.264
STD*10 12.502
500 12.736

400
7.5 8.5 9.5 10.5 11.5 12.5 13.5
wavelength (µm)

Figure 28: Mean and standard deviation (*10) spectra for the Douglas Hill SEBASS run.

As SEBASS is a pushbroom imaging system, then any systematic instrument noise is


expected to be column-dependent. This type of noise if highlighted in the MNF
transformed raw SEBASS data in Figure 29 and ranges from broad, slowly changing
instrument noise (e.g., MNF bands 1-4 and 6) to sharp, column-specific noise (e.g., MNF
bands 16, 22 and 28). The individual column noise is suggestive of miscalibrated or “bad”
detector element/s. There is also random noise (e.g. MNF band 119) and a slightly
oblique line noise (e.g. MNF bands 43, 49, 59). This last form of noise is unexpected for
a pushbroom system. More is discussed on this issue below. There are also MNF bands
free of apparent noise including MNF bands 5, 7, 8, 9 and 10, with band 9 containing the
bulk of the surface temperature information. (If Jack H needs convincing “Such effects
were also observed by Hewson et al. (2000) within the Oatman SEBASS data. They
excluded those MNF bands containing column effects prior to mineralogical analysis.

CSIRO Exploration and Mining Report 734R 37


Figure 29: Selected images from a segment of the MNF transformed SEBASS
Ludwig line with all 124 available spectral surface radiance bands used as input.

38 CSIRO Exploration and Mining Report 734R


Figure 30: Selected images from a segment of the MNF transformed SEBASS Ludwig
line with only 72 of the available spectral surface radiance bands used as input.

CSIRO Exploration and Mining Report 734R 39


The slightly oblique noise in MNF bands 43, 49 and 59 in Figure 29 is possibly related to
problems in the alignment of the prism and the related dispersed light projected across the
128 by 128 pixel detector array. This possible misalignment is considered by Aerospace
engineers to be the most likely explanation to account for much of the apparent spatial-
spectral noise in SEBASS data, especially for those spectral bands deep within
atmospheric line-absorptions. To test this hypothesis, all SEBASS TIR spectral from
within any apparent atmospheric-line absorptions, as well as any noisy spectral bands
(observed from the image data and from its statistics), were omitted and reprocessed
using MNF. The noisy bands were typically located at shorter and longer wavelengths.
This culling resulted in 72 bands remaining from the original 1242 bands (Appendix 1).

The MNF transformation of the culled SEBASS data (Figure 30) show more noise-free
MNF bands [e.g. MNF bands: 4, 5, 6, 7 (temperature), 8, 12, 14] though the total number
of MNF bands with apparent surface information, but also with some instrument noise,
remains the same (approximately 45 bands in both) as the MNF transformed raw
SEBASS data (Figure 29). The culled data show a lesser number of MNF bands with
instrument noise, though broad column banding (e.g. MNF bands 1, 2 and 3), sharp
column striping (e.g. MNF band 18) and slightly oblique striping are still present,
indicating that these are not dependent on isolated “noisy” channels but affect much of
the data. In summary, the removal of 52 spectral bands did not appear to result in a
significant loss of surface compositional spectral information but did succeed in removing
some but not all of the of the instrument noise.

Finally, Figures 29 and 30, show no spectral-spatial variations along the length of the run
that may suggest the thermal calibration of the data has been compromised. That is,
measurement of the hot and cold blackbody calibrations at the beginning and end of the
line only (and not regularly through the run), assuming linear drift between these two
measurements, appears to be sufficient for characterising the instrument temperature drift.
This is an important result with implications for TIR sensor design and operation
(expand).

As an aside, an MNF transformation of continuum-removed SEBASS “emissivity”


spectra, to simulate removal of the dominating though arguably redundant surface
temperature information, showed that approximately 45 MNF bands contained surface
emissivity information. Furthermore, the partial unmixing (MTMF) products of these
continuum-removed SEBASS image data appear to be the same as those generated
without continuum-removal. That is, there is no apparent gain in implementing this
additional processing step of continuum-removal, especially if the products of partial
unmixing are essentially temperature-independent (Section 7.3.1).

7.3 Temperature-Emissivity Separation

7.3.1 Background
The atmospherically corrected SEBASS surface radiance data comprises two
components, namely: (a) Planck Blackbody Radiation as a function of the surface kinetic
temperature; and (b) attenuation of this ideal Planck Blackbody Radiation because of

2
The last four of the 128 SEBASS bands do not contain image data.

40 CSIRO Exploration and Mining Report 734R


composition-related emissivity contrast. This emissivity is a function the material’s
absorption coefficient (related to selective absorption by molecular vibrations like Si-O)
and the real part of the refractive index (relative to the transmitting medium - Fresnel
reflectance)3. To retrieve this desired emissivity it is necessary to separate it from the
temperature component in the measured surface radiance. This is non-trivial because it
results in an under-determined set of algebraic expression based on the Planck Radiation
Function (N-channels and N+1 unknowns). As a consequence, numerous TES methods
have been developed (Kahle et al. 1980, Price 1984, Kealy and Gabell 1990a and 1990b,
Realmuto 1990, Warner and Levandowski 1990, Hook et al. 1992, Kahle and Alley 1992,
Watson 1992a, Watson 1992b, Matsunaga 1994, Richter 1994, Barducci and Pippi 1996),
though the accuracy of these has been difficult to establish (Cudahy et al., 2000).

These methods were developed for passive “broadband” multispectral TIR data, which
have too few spectral bands to accurately define the background hull spectrum for each
pixel. High spectral resolution TIR data, like that from SEBASS, possibly have sufficient
bands to characterise both the emissivity lows as well as the “background”, making TES
theoretically simpler. That is, a simple correction for the background on a per-pixel basis
(continuum-removal) could be sufficient for TES. It is also possible, that wavelength-
dependent temperature could be separated crudely as endmembers and unmixed
(Boardman, 1993) as part of the spectral “background”.

Some preliminary testing by one of the authors (K. Okada) found that the effects of
temperature could be separated (in part) from the desired emissivity information in raw
SEBASS data using a simple normalisation procedure like log residuals. He also found
that the derived log residual pixel spectra could be interpreted for their mineralogy.
Appendix 3 explores this possibility a little further and confirms these preliminary results,
and describes how scene-mean-dependent spectral artefacts complicate the pixel
signatures. There also remain residual temperature effects because the log residual
normalisations do not properly account for the wavelength-dependent and pixel-
dependent surface temperature.

7.3.1 Is there a need for TES?

To test the need for TES on the SEBASS data, a segment of the Douglas Hill run over
pronounced topographic relief was examined for the effect of residual temperature effects
on partial unmixing (MTMF) products with and without TES. In this case, TES was
implemented by removing each pixel spectrum’s background hull through “continuum-
removal” available in ENVI.

The results of this comparison for three mineral-spectral endmembers (calcite, dolomite
and andradite), together with a brightness temperature image, are presented Figure 31.
The striking similarity of the unmixed products derived with TES (Figure 31b) and
without TES (Figure 31c) and the lack of any correlation with the surface kinetic
temperature (Figure 31a) indicates that TES is not required to derive temperature-
independent mineral distribution maps using partial unmixing (although I also found this
result for one of my SEBASS runs, I guess we should qualify this for our data and not
assume its always the case-requires further testing maybe for many different conditions

3
Scattering also modifies spectral behaviour.

CSIRO Exploration and Mining Report 734R 41


scene topo etc). This was found to be the case for all mineralogical endmembers when
thresholding for “high” abundances. However, at “low” abundances the effects of
residual temperature in the non-TES data sometimes became apparent (Section 10.1). As
part of this experiment, it was noted that both “hot” and “cold” blackbody-like
endmembers, with different spectral shapes and intensities, were apparent in the non-TES
corrected image data. It appears that the pronounced contribution made by these
wavelength-dependent hot and cold endmembers to each pixel, can be well modelled as
the “background” for most (if not all) mineralogical endmembers in partial unmixing.

Figure 31: (a)


SEBASS brightness
temperature image;
(b) Partial unmixing
of SEBASS data with
TES (TES calculated
using pixel spectrum
continuum-removal);
and (c) Partial
unmixing of SEBASS
data without TES.
Red: calcite; Green:
dolomite; Blue:
andradite.

8. EXTRACTION OF MINERAL INFORMATION

8.1 Introduction
The extraction of SEBASS mineral information products concentrated on two methods,
namely:

1. Spectral indices using bands ratios and derivatives of fitted polynomials; and

2. Partial unmixing (Boardman, 1993).

8.2 Spectral Indices


A ratio of bands is a simple and fast method for exploring a scene for “candidate”
signatures. However, it does require a priori knowledge of the diagnostic wavelengths
and does not capitalise on the full spectral information content of the hyperspectral data
but does target diagnostic features. Unexpected materials with overlapping spectral

42 CSIRO Exploration and Mining Report 734R


signatures can compromise the resultant information though and so checking derived
product by perusing the pixel spectra is important as part of the validation procedure.

A ratio of SEBASS bands located at 11.4 and 11.0 µm were used to locate pixels either
rich in garnet or poor in carbonate (Figure 32c and 32e, respectively). The second
spectral index used was the estimate of the wavelength position of the garnet and
carbonate emissivity peak near 11.4 µm. This was achieved by calculating the 1st
derivative of a fitted 3rd order polynomial to the wavelength segment between 11.1 and
11.6 µm. An estimate of the RMS (root mean square) error of the fit helps in the
assessment of the suitability of the measurement. Thus the same spectral index was used
to map both the garnet chemistry and carbonate mineralogy. They are easily separated
from each other because they have an opposing relationship. That is, a peak for garnet
contrasts with a trough for carbonate.

8.3 Image Endmember Analysis and Partial Unmixing


Image endmember analysis and partial unmixing has the objective of generating new
variables capable of accounting for the image data in terms of a mixing model. The
strategy is that the geometry of the n-dimensional image data cloud can be used to both
isolate the spectrally “pure” (and hopefully mineralogically pure) endmembers, as well as
to define the image “background”, in order to derive the necessary mixing projection
(Boardman, 1993). Importantly, the mixing vector for a given spectral endmember is
calculated without direct reference to any other scene endmembers. It uses the
endmember of interest and the data cloud mean. The routine is available in ENVI (BSC,
1998) with the partial unmixing procedure called “Mixture Tuned Matched Filtering” or
MTMF. MTMF has the advantage of being able to address problems with commission,
which arises when there are two or more partial unmixing vectors sharing the same
mixing space. The “solution” is achieved through an infeasibility score, which is a
measure of the distance between a given pixel and the selected partial unmixing vector,
with large distances equating to high infeasibility.

8.4 Comparison of Results


The mineral maps generated using the methods described above show a high degree of
similarity for calcite and dolomite (Figure 32d and 32e) and reasonable correlation for
garnet (Figures 32b and 32c). Note that the threshold cut-off for low abundance for both
methods is subjective and inevitably leads to some misclassification.

The spectral indices map a continuous variation in garnet wavelengths whereas the
MTMF yielded four (only three are presented in Figure 32b) separate garnet endmembers.
As described above, garnets comprise a complete range of solid solution chemistries (and
related TIR wavelengths - Section 9.3). Clearly, MTMF is capturing some of this garnet
wavelength variation, though it is also sensitive to the relative depths of the spectral
diagnostic features, which is important for abundance mapping, but not necessarily for
mapping mineral chemistry. This helps explain the differences between these products
(Figures 32b and 32c).

CSIRO Exploration and Mining Report 734R 43


Figure 32: Comparison of mineral information products generated using partial
unmixing and spectral indices: (a) raw band 8 (8.03 um); (b) Partial unmixing image of
garnet composition - red is long wavelength, green is medium wavelength and blue is
short wavelength; (c) Spectral index map of garnet composition – blue is no garnet (ratio
of 11.4 over 11.0 µm), black is short wavelength, grey is medium wavelength and white is
long wavelength (1st derivative of polynomial fit); (d) Partial unmixing image (MTMF) of
calcite (red), dolomite (green) and quartz (blue); (e) similar to (c) except it is mapping
carbonate calcite as white and dolomite as black with blue as no carbonate (ratio of 11.3
over 11.1 µm).

44 CSIRO Exploration and Mining Report 734R


In contrast with the garnet maps, the two carbonate products are very similar (Figures 32d
and 32e). This is probably a function of the lack of solid-solution chemical (and
wavelength) variation that exists between the calcite and dolomite “endmembers”. Under
such circumstances, MTMF is as an effective and accurate method.

Thus, the spectral index method is arguably better suited to mapping solid solution
chemical (wavelength) information (not abundance) compared with partial unmixing,
especially if the wavelength variations related to solid solution chemistry are small.
Another advantage of the spectral index method is that it is simple to apply, unlike partial
unmixing.

9. SEBASS MINERAL MAPS OF SKARN ALTERATION

Mineral maps were generated for the skarn alteration found along the Douglas Hill and
Ludwig SEBASS lines (Figures 1 and 5) using MTMF (Section 8.3). Note that the
SEBASS images or their derived products were not georeferenced to a map-base. The
results are described in detail below.

9.1 Douglas Hill Line


The MTMF-derived SEBASS mineral map of the skarn alteration near the Casting
Copper and Douglas Hill Copper mines (Figures 2a, 2b, 5 and 6) is presented together
with supporting information in Figures 33 and 34. The supporting information includes:
the SEBASS endmember spectra; surface temperature image with field localities
annotated; the published alteration map (Figure 6); and field microFTIR emissivity
spectra. Note that the colours of the endmember spectra relate directly to the colours of
the MTMF SEBASS mineral map.

Minerals that can be identified from the SEBASS data (Figure 33a) include: four types of
garnets; pyroxene+garnet; pyroxene (diopside-hedenbergite); epidote; dolomite; calcite;
quartz+clay+feldspar; albite; and two unidentified spectra, one of which may relate to
idocrase. The associated microFTIR spectra (Figure 34) are consistent with both the
shapes of these SEBASS endmembers as well as their location on the mineral map
(Figure 33b and 33c). For example: sample sites 5 and 6 show calcite; localities 28 and
31 show dolomite; localities 3, 27 and 29 show andradite garnet; locality 30 shows
pyroxene (diopside); locality 17 shows pyroxene-garnet; and locality 10 shows epidote in
both the field microFTIR and SEBASS data.

A comparison between the published alteration map (Figure 33d) and the SEBASS-
derived mineral map (Figure 33b) shows similarities as well as differences. Near the
Casting Copper and Douglas Hill Mines, there is excellent correspondence, especially
with regard to the distribution of the unaltered limestone (brown), dolomotised limestone
(orange), andradite skarn (red) and pyroxene skarn (dark blue). Even the detail of a
“halo” of andradite skarn surrounding a thin band of pyroxene skarn near the Casting
Copper mine (A) is accurately depicted by the unmixed SEBASS data. Towards the
northeastern end of the line (Figure 33), the contact between the altered skarn and the
granodiorite is well expressed as a change from garnet-rich to quartz-albite-rich pixels in
the SEBASS data (B). A wide albite-rich felsic dyke is also mapped in both (C).

CSIRO Exploration and Mining Report 734R 45


Figure 33: Douglas Hill SEBASS composite Mineral Map: (a) SEBASS
endmember emissivity spectra (continuum-removed surface radiance); (b) MTMF
SEBASS mineral theme map with colours related to the endmeber spectra in (a);
(c) SEBASS temperature image with field sample localities annotated; and (d) the
published alteration (after Harris and Einaudi, 1982) – see Figure 6 for details.

The dissimilarities between the published mapping and the SEBASS mineral map (Figure
33b and 33d) are related to the skarnoid alteraton (Al-rich garnet and/or pyroxene). The
published mapping shows a strong parent rock control on this mineralogy (Figure 5)
though this is not expressed in the processed SEBASS data. Harris and Einaudi (1982)
commented that mapping in some areas associated with complete alteration to skarnoid is
difficult. There is also the possible problem of the accuracy of the MTMF garnet
chemistry maps as discussed in Section 8.4.

46 CSIRO Exploration and Mining Report 734R


Figure 34: Douglas Hill composite diagram: (a) field microFTIR emissivity
spectra; (b) SEBASS temperature imag e with field sample localities annotated;
and (c) MTMF SEBASS mineral theme map as shown in Figure 33b.

9.2 Ludwig Line


The MTMF-derived SEBASS mineral map for the skarn alteration near the McConnell
Cu Mine and the Ludwig Gypsum mine (Figures 2c, 2f, 5 and 6) along Ludwig line is
presented in Figure 35. The associated endmember spectra and surface temperature
image with field localities annotated are also shown in Figure 35 while the field
microFTIR spectra are presented in Figure 36. No detailed published mapping was
available for this area. The colours of the endmember spectra directly relate to the
colours of the MTMF SEBASS mineral map, except for garnet, which is derived using
the polyfit method and colour coded as a grey-scale.

CSIRO Exploration and Mining Report 734R 47


Figure 35: Ludwig SEBASS composite diagram: (a) and (c) SEBASS temperature image
with field sample localities annotated; (b) and (d) SEBASS mineral maps derived using
both MTMF for all minerals (colours relate to the endmember spectra (e)), except for
garnet, which was derived using the polyfit method; (e) SEBASS endmember emissivity
spectra (continuum-removed surface radiance).

Minerals that can be identified in Figure 35 include: garnet of variable solid chemistry;
pyroxene (diopside-hedenbergite); epidote; dolomite; calcite; quartz; kaolinite; and
gypsum. The associated microFTIR spectra (Figure 35) of samples from the field
localities are consistent both with the shapes of the SEBASS endmembers and with their
location on the mineral map (Figure 34). For example: sample sites 51, 52 and 53 show
quartz; localities 39A and 39C show kaolinite; localities 34 and 38 show garnet,
localities 33 and 36 show dolomite; and locality 35 shows calcite in both the field
microFTIR and SEBASS data.

48 CSIRO Exploration and Mining Report 734R


Figure 36: Ludwig composite diagram: (a) and (c) SEBASS temperature image
with field sample localities annotated; (b) and (d) SEBASS mineral maps derived
using both MTMF for all minerals (colours relate to endmember spectra (e)),
except garnet, which was derived using the polyfit method; (e) field microFTIR
emissivity spectra with their field location numbers annotated.

9.3 Garnet Solid-Solution Chemistry


The results of the SEBASS garnet mineral processing described above show that there
exist systematic variations in spectral behaviour that appear to correspond with the
published distribution of Fe-garnet (andradite) and Al-garnet (grossular). The ability of
the TIR to provide quantitative measurement of the garnet solid solution chemistry
(Cudahy et al., 1999) is first examined below through a laboratory study of field samples.
This is then followed by an examination of the SEBASS garnet chemistry maps derived
through spectral indices.

CSIRO Exploration and Mining Report 734R 49


9.4.1 Laboratory Study
Of the forty samples analysed by XRD (Appendix 2), seventeen were found to contain
garnets. MicroFTIR emissivity spectra of a selection of these garnet-bearing samples
(Figure 37) show changes in the position of the narrow peaks and troughs consistent with
variations in garnet solid solution chemistry. These laboratory spectra show similarities
with the five SEBASS MTMF “garnet” endmember spectra collected from the Douglas
Hill line (Figure 38), especially the wavelengths of the emissivity lows near 11.2, 11.9
and 12.3 µm and the emissivity high near 11.5 µm.

Laboratory emissivity - field samples


1.1

0.9
emissivity

0.8

0.7

0.6

0.5

0.4
7.5 8.5 9.5 10.5 11.5 12.5
wavelength (µm)

Figure 37: Laboratory emissivity spectra of selected garnet-bearing field samples.

1.02

1
apparent emissivity

0.98

0.96

0.94
andradite skarn
0.92
skarnoid (west)
0.9 skarnoid (northeast)
skarnoid (southeast)
0.88
pyroxene skarn
0.86
7 8 9 10 11 12 13 14
wavelength (µm)

Figure 38: SEBASS apparent emissivity (continuum-removed) spectra of garnet-rich end


members.

The garnet XRD 420 hkl reflection peak is the strongest of the garnet XRD features and is
sensitive to the cation chemistry (Fe-Al). The d-spacing of this XRD peak was used for
comparison with the wavelength of the 11.5 µm feature. The results (Figure 39) show
significant linear correlation, confirming this relationship between spectral properties and
XRD mineral crystal lattice dimensions (cation chemistry).

50 CSIRO Exploration and Mining Report 734R


2.7
Fe-garnet - andradite
2.695 (2-d spacing of 2.696<0.01% Al, Cr, Mg)

XRD 2-d spacing (420 hkl peak)


2.69

2.685

2.68

2.675

2.67
y = 0.1148x + 1.3591
2
2.665 R = 0.8348

2.66
11.3 11.35 11.4 11.45 11.5 11.55 11.6 11.65
"emissivity maximum" wavelength (µm)

Figure 39: Scattergram of the XRD 2-d spacing for the 420 hkl garnet reflection peak
versus the wavelength position of the emissivity peak for 17 garnet bearing Yerington
samples. These data show linear correlation for these two parameters related to
changes in the cation chemistry (amount of Fe relative to Al) of the garnet.

Electron microprobe chemical analyses of garnet grains from fourteen of the 17 garnet-
bearing samples in Figure 39 were measured by Sumitomo Metal Mining in Japan. This
analysis found that the garnet grains were generally unzoned. A consideration in
comparing these analyses with the XRD and spectral data is all are effectively from
different sample media (crystal versus bulk versus whole rock surface, respectively).

A comparison of the electron microprobe analyses for Fe and Al against the


corresponding XRD (Figure 40) and microFTIR (Figure 41 data) show significant linear
correlations.

EPMA %andradite
1
EPMA %grossular
EPMA cation (Fe, Al) content

0.8

y = 1.5x - 15.9
0.6 R2 = 0.67

y = -1.5x + 16.9
0.4 R2 = 0.67

0.2

0
11.3 11.35 11.4 11.45 11.5 11.55 11.6 11.65
emissivity maximum wavelength (µm)

Figure 40: Scattergram of the wavelength of the microFTIR emissivity peak near 11.5 µm
versus the electron microprobe analyses of Fe and Al for a selection of Ludwig field
samples (enlarged).

CSIRO Exploration and Mining Report 734R 51


EPMA %andradite
1 EPMA %grossular

EPMA cation (Fe, Al) content


0.8

y = 11.9x - 31.0
R2 = 0.70
0.6

y = -11.9 + 32.0
0.4 R2 = 0.70

0.2

0
2.66 2.665 2.67 2.675 2.68 2.685 2.69 2.695 2.7
XRD 2-d spacing (420 hkl peak)

Figure 41: Scattergram of the wavelength of the XRD d-spacing of the 420 hkl
garnet reflection versus the electron microprobe analyses of Fe and Al for a
selection of Ludwig field samples.

These excellent results show that the wavelength of the garnet TIR feature near 11.5 µm
can be used to directly measure the grandite Fe-Al solid soution chemistry.

9.4.2 SEBASS Garnet Chemistry Mapping

The wavelength position of the garnet emissivity peak near 11.5 µm was calculated using
the spectral indices (described in Section 8.2) for the Douglas Hill and Ludwig SEBASS
lines. These data were then used to generate a predictive map of the grandite Fe-Al solid
solution chemistry (Figure 42).

The first point to note about Figure 42 is how well the two lines overlap both with respect
to geometry and colour. That is, this method of tracking the wavelength position of the
emissivity peak for extracting the garnet information appears to be relatively scene-
independent, that is, generates seamless maps.

The accuracy of the estimated wavelength of the SEBASS 11.5 µm garnet peak was
checked using a selection of pixels covering field sites where microFTIR spectra were
extracted. Field sites Y027 and Y029 near the Douglas Hill mine (Figure 34) have a peak
at 11.58 µm for the field data and a peak at 11.56 µm for the SEBASS data. Field sites at
Y025 and Y026 in mapped Al-rich skarnoid alteration (Figure 34) have peaks at 11.37 to
11.38 µm for the field data and 11.37 µm for the SEBASS data. These favourable results
confirm the accuracy of the measured Fe-Al solid solution chemistry as derived from the
SEBASS data.

The map of the garnet Fe-Al solid solution chemistry (Figure 42) clearly delineates those
areas of skarn-alteration and shows a progressive enrichment in Fe towards the southwest
(away from the northern granodiorite contact). This is consistent with the skarnoid
alteration zonation mapped by Harris and Einaudi (1982).

52 CSIRO Exploration and Mining Report 734R


A

Figure 42: Mosaic of SEBASS-derived garnet solid solution chemistry images for
the overlapping Douglas Hill and Ludwig lines. 100% Fe garnet is depicted as
white and 50:50 Fe:Al in black. Grey tones represent intermediate solid solution
chemistries. The blue tones are those areas poor in garnet and was determined
using a ratio of spectra bands. The underlying brightness temperature mosaic
provides spatial context. The positions of the three main skarn-related copper
mines covered by these lines area are also annotated.

The late andradite skarn alteration associated with Cu mineralisation is immediately


adjacent to carbonate at the Casting Copper and McConnell Mines (Figure 42). Note that
there is no apparent zonation in the garnet chemistry peripheral to these areas, which is
consistent with the mapping Harris and Einaudi (1982). However, the Douglas Hill
Mines are clearly separated from the closest exposure of dolomite and there exists a
pronounced zonation in the garnet chemistry around this area. This observation questions
the validity of a late andradite skarn alteration phase overprinting an earlier skarnoid
alteration phase.

Arguably the most important feature evident in Figure 42 is that all the skarn-style Cu
mines intersected by the SEBASS lines are clearly delineated as near pure andradite.
That is, the TIR-derived garnet Fe-Al chemistry provides a potential vector towards Cu
mineralisation (at least in this area).

Figure 42 also shows that garnet and its chemical composition can be mapped in alluvial-
colluvial materials that have moved at least 2 km from their source (A). This suggests
that the provenance of transported materials can be targeted remotely with recognition of
Fe-rich garnets indicating “local” Cu mineralisation.

CSIRO Exploration and Mining Report 734R 53


10. SEBASS MINERAL MAPS OF PORPHYRY ALTERATION

The results presented here are only a preliminary analysis of the SEBASS data for
mapping porphyry style alteration at Yerington. A later study of the SEBASS data will
look in detail at the alteration associated with the well exposed Ann-Mason porphyry
system which was intersected by four SEBASS runs and has available large-scale
published maps of the geology and alteration (Dilles and Einaudi, 1992).

10.1 Douglas Hill Line


The Douglas Hill SEBASS line extends north-eastward from the skarn-alteration around
Ludwig across the Yerington Batholith and to the Tertiary volcanics (Figures 1, 5 and
43). SEBASS MTMF mineral maps were generated of garnet mineralogy (Figure 43c
and 43i), carbonate mineralogy (Figure 43d and 43j), epidote-quartz-muscovite (Figure
43e and 43k) and feldspar mineralogy (Figure 43f 43l). These were selected because they
provide information about the extent of the carbonate metasediments and related exoskarn
alteration and how it changes to endoskarn and porphyry style alteration in the adjacent
garnodiorite (grandite, quartz, epidote, feldspar, muscovite).

Figure 43: SEBASS Mineral maps associated information for the Douglas Line.
Note the line has been cut in two and placed side-by-side for convenience. Strips
(a) and (g) are the published 1:24000 scale geology (Figure 5 - Proffett and
Dilles, 1984); (b) and (h) are the surface “brightness” temperature with field
sample localities annotated. Strips (c) and (i) are Fe-garnet (red), Fe-Al garnet
(green) and Al garnet (blue). Strips (d) and (k) are calcite (red), dolmite (green)
and quartz (blue). Strips (e) and (k) are epidote (red), white mica (green) and
quartz (blue), but with a lower threshold compared with (d) and (j) though is
sensitive to residual temperature variations. The green elipse on Figure (a) shows
the approximate position of the oligoclase-andesine alteration and the blue elispse
(g) shows the location of oligoclase-albite alteration mapped by Dilles and
Einaudi (1992).

54 CSIRO Exploration and Mining Report 734R


The MTMF distribution of garnet (Figure 43c) is restricted to the skarn-altered limestones
(light blue in Figure 43a) and within a wedge of carbonate-rich sediments (A in Figure
43b). No carbonate is found using SEBASS beyond those areas mapped as carbonate-rich
by Proffett and Dilles (1984) (Figure 43d and 43j). Only the Al-garnet is developed in the
carbonate-rich rocks close to the granodiorite. There is no conclusive evidence for garnet
being mapped by SEBASS in the granodiorite, though Dilles and Einaudi (1992) had
noted this type of alteration close to the contact with the limestones.

The MTMF distribution of epidote and quartz (Figure 43e and 43k) shows a pronounced
relationship with respect to the position of a wedge of carbonate-rich rocks in the
granodiorite (A – Figure 43b). This wedge is characterised by a high abundance of
epidote and quartz in the SEBASS data (Figure 43e). To the southwest of this wedge,
including within the adjacent granodiorite, there is also epidote and quartz. However,
immediately to the NW of this wedge in the granodiorite, epidote and quartz are not
developed. Instead Ca-albite is dominant (Figure 43f - discussed below). Elsewhere
within the granodiorite, a weak quartz signature is developed, though this is typically
associated with residual temperature effects (topographic shading), indicating that its
level of MTMF abundance is at the “noise” limits.

Muscovite is well developed in the Tertiary volcanics to the NW of the batholith (B –


Figure 43l), with only minor, localised, occurrences within the granodiorite and possibly
in felsic dykes in the limestones (Figure 43e and 43l). This is consistent with the
granodiorite being biotite-hornblende rich but muscovite-poor (Dilles and Einaudi, 1992).

Three feldspar endmembers were found in the SEBASS data (Figure 44) with all showing
narrow emissivity lows at 9.6 and 10.0 µm indicating Na-plagioclase. Two of these
endmembers also share a peak at 9.2 µm (red and blue spectra – Figure 44), consistent
with the albite signature in Figure 11, while the third has no feature at this wavelength as
well as a relatively large 10.0 µm feature (green spectrum – Figure 44). This green
spectrum is consistent with greater Ca content, as the lack of 9.2 µm feature and a strong
10.0 µm feature is similar the andesine feldspar measured by Cudahy et al. (2000). The
red and blue spectra also differ in the 8-9 µm region with the red spectrum not showing as
well developed emissivity lows at 8.4 µm and 8.7 µm. The reason for this is not yet clear
as all fresh feldspars show features at these wavelengths (Figure 11) though it may be
related to a weathering effect or a function of the continuum-removal. Thus, the SEBASS
endmember spectra appear to be of two types: (i) “Na-rich” albite (red and blue spectra in
Figure 44); and (ii) “Ca-rich” andesine (green spectrum in Figure 44).

The SEBASS image of Na-Ca feldspar composition (Figure 43f and 43l) shows a pattern
of andesine (green and red) in the SW and the albite in the centre and NW parts of the
batholith. This distribution does not correlate well with the 1:24,000 scale published
geology (Figures 43a and 43g) though does correlate with the more detailed alteration
mapping of Dilles and Einaudi (1992). Specifically, Dilles and Einaudi (1992) mapped
oligoclase-andesine alteration in the SW (blue ellipse on Figure 43a) and oligoclase-albite
in the central-northeastern part (green ellipse in Figure 43g).

The SEBASS data show that the andesine abruptly terminates against a “stoped” wedge
of sediments (A – Figure 43b), with the other side free of feldspar but rich in epidote and
quartz. This suggests some structural/stratigraphic control to the alteration.

CSIRO Exploration and Mining Report 734R 55


1.005

8.4 8.7 9.2 9.6 10.0

apparent emissivity
0.995

0.99

0.985
albite-oligoclase
oligoclase-andesine
albite oligoclase
0.98
8 8.5 9 9.5 10 10.5 11 11.5
wavelength (µm)

Figure 44: Douglas Hill SEBASS line continuum-removed endmember spectra.


The three spectra labelled as albite, as well as their colour, are used in Figure 34f
and 34l. The white mica spectrum was used in Figure 34e and 34k.(enlarged)

The associated microFTIR emissivity spectra of field samples (Figures 45 and 46) are
consistent with the SEBASS spectra and the distribution of Ca versus Na alteration. For
example, the microFTIR spectra of samples from the “Ca-rich” area (Figure 45) show
relatively pronounced development of the 10.0 µm feature and relatively weaker
development of the 9.2 µm feature compared with the microFTIR spectra of samples
collected from the Na-rich area (Figure 46). Type examples of these include the Ca-
albite/andesine (XRD - Appendix 2) of Y021_A (Figure 45) and the Ca-albite (XRD –
Appendix 2) of Y042_B.

56 CSIRO Exploration and Mining Report 734R


Figure 45: Stack profile of field
microFTIR emissivity spectra of selected
samples collected from felsic intrusive
rocks near the contact with the Ludwig
skarn (See Figure 43 for sample
locations). (enlarged) Figure 46: Stack profile of field
microFTIR emissivity spectra of
selected samples collected from felsic
intrusive rocks of the Northern
Batholith (See Figure 43 for sample
locations). (enlarged)

10.2 Macarthur Line


A segment of the Macarthur SEBASS line that covered the Macarthur porphyry copper
open pit mine was processed using MTMF (Figure 47). Field sampling within the open
pit mine (Figures 2e, 4c and 4d) found the mine floor was well graded with the surface
strewn with “disturbed” material. Nevertheless, the processed SEBASS imagery yielded
coherent maps that clearly delineated narrow albite-rich, quartz-poor dykes (green –
Figure 47) crosscutting a quartz-albite rich granodiorite (red). This is an important result
as it indicates that the mine floor is not disturbed such that coherent map of the mine floor
geology can be obtained though remote sensing. The quartz-rich granodiorite appears to
continue laterally beyond the confines of the open pit mine (red - Figure 47), whereas the
albite-rich dykes appear to abut against a unit (cyan – Figure 47) that has a signature
comprising emissivity peaks near 9.2 µm and 10 µm.

CSIRO Exploration and Mining Report 734R 57


Figure 47: SEBASS mine map of the Macarthur mine area including the open pit
floor. The red spectrum shows quartz and possibly K feldspar (microcline) + Na
feldspar (whereas the green is Na felspar and quartz. This mapping of two
feldspars is shown by a small change in the wavelength position of the 9.6 um
feldspar emissivity low, which is slightly shorter for the red spectrum. This may
reflect albitisation of the K feldspar during alteration. Note how the areas
mapped are located areas preferentially mined especially in the NW end of the
open pit.

The associated field microFTIR spectra (Figure 48) are consistent with the SEBASS
endmember spectra (Figure 47) as they also show albite and quartz signatures but also a
strong emissivity low at 9.2 µm (for example, Y050_D). This 9.2 µm feature is possibly a
function of a clay-weathering product, though comparable PIMA spectra also indicate the
development of epidote. Further work is required to resolve this issue.

58 CSIRO Exploration and Mining Report 734R


Figure 48: Stack profile of field microFTIR emissivity spectra of selected samples
collected from the Macarthur open pit porphyry copper mine (See Figure 37 for
sample location).

11. AVIRIS MINERAL MAPS

The high-altitude AVIRIS radiance at sensor data were reduced to surface radiance using
a CSIRO developed version of ATREM, called HYCORR. HYCORR uses the radiative
transfer model 6S for deriving corrections for all the gases except water vapour, as well as
aerosol scattering. These corrections are calculated on a per-scene basis with
consideration given to scan angle and path length. Input parameters were based on a
standard climate model. The water vapour is calculated on a per-pixel basis using the
measured depth of the 3rd and 4th water vapour combination tones, similar to ATREM. As
a final step, the reduced pixel spectra are processed using EFFORT, which effectively
removes (normalises) any residual systematic high frequency atmospheric signature.

The geological information extraction from the derived surface radiance was implemented
using MTMF. The endmembers were selected from those pixels AVIRIS data that
essentially overlapped with SEBASS runs (Figure 49).

CSIRO Exploration and Mining Report 734R 59


Figure 49: (a) AVIRIS-derived partial unmixing mineral distribution map of the
Yerington area. (b) AVIRIS endmember spectra. (c) Same as (b) but scaled
between 2.0 and 2.5 µm. The assigned mineralogy of those endmember spectra
with a question mark are inferred not from any diagnostic spectral shape but their
relation to the mapped geology.

60 CSIRO Exploration and Mining Report 734R


The results of the MTMF unmixing of the AVIRIS data (Figure 49) show discrimination
of most of the primary rock types and superimposed skarn alteration zones mapped by
Harris and Einaudi (1982). The spectral signatures of the related image endmembers
provide diagnostic information about the mineralogical character of many of these
discriminated units. For example, the yellow spectrum shows left-asymmetric absorption
at 2.340 µm diagnostic of calcite while the cyan spectrum shows a similar feature but
positioned at shorter wavelengths 2.320 µm indicating dolomite (Figure 49c). The
corresponding MTMF “classified” images of these two spectral endmembers accurately
defines the exposure of limestone and dolomitised limestones (Harris and Einaudi, 1982).
Similar spectral identification and mapping of mineralogy is observed for gypsum, two
types of white mica, epidote, amphibole and chlorite. These OH-bearing minerals are
therefore well mapped at the SWIR wavelengths.

There are other image endmembers that generate coherent image units consistent with the
published alteration mapping (Figure 6) though their spectral signatures do not contain
mineralogically diagnostic behaviour. For example, a dark green spectrum and a mauve
spectrum generate unmixed results consistent with the spatial distribution of andradite and
grossular garnet respectively though they contain no diagnostic absorption, especially at
SWIR wavelengths (Figure 49c). That is, both of these spectra, together with that of the
brown (quartz?), ochre (idocrase?) and dark blue (pyroxene?) spectra all have a broad but
shallow minimum centred near 2.3 µm. The mineralogical interpretation of these spectra
is based on their spatial context referenced to the published mapping (Figure 6). All of
these minerals are non-OH-bearing silicates, which are SWIR inactive. The
discriminatory information that separates these silicates is essentially from the VNIR
where there exist “electronic” spectral differences related to the types and proportions of
transition elements (primarily Fe). However, this is subtle and not diagnostic. Windeler
(1993) found a similar result with Geoscan Mk II scanner from the same test site where it
was possible to map garnet-rich rocks based on the VNIR response.

Therefore, the hyperspectral VNIR-SWIR AVIRIS provides discriminatory information of


most (if not all) of the main primary rock type and skarn alteration zones but does allow
for easy identification of the non-OH bearing silicate mineralogy, such as garnets,
pyroxenes, feldspars and quartz.

12. DISCUSSION

12.1 Mineral Alteration Mapping and Exploration

The mineral mapping results described above show that SEBASS data provide spectral
identification and mapping of:

‰ Quartz;
‰ Grandite including their complete Fe-Al solid solution chemistry;
‰ Clinopyroxene (diopside-hedenbergite) though their Fe-Mg solid solution chemical is
not available;

CSIRO Exploration and Mining Report 734R 61


‰ Na-plagioclase feldspars including subtle Na-Ca solid solution variations from albite-
oligoclase to oligoclase-andesine;
‰ Carbonates including calcite and dolomite;
‰ Epidote;
‰ Amphibole;
‰ Muscovite;
‰ Kaolinite;
‰ Gypsum; and
‰ Clay.

No K-feldspars have yet been identified even though these account for up to 25% of some
of the felsic rocks.

AVIRIS data were found to be useful in mapping the OH-bearing minerals (epidote,
amphibole, white micas of different chemistry, kaolinite, gypsum and clays) and
carbonates (calcite and dolomite), though it was not suitable for the spectral identification
of the non-OH-bearing silicates (quartz, garnet, pyroxene, feldspar). In the context of
mapping the skarn-style and the porphyry-style alteration at Yerington, these non-OH-
bearing minerals are critical, especially for the Fe-Al chemistry of the grandites and the
Na-Ca chemistry of the sodic plagioclase feldspars.

The skarn-style alteration mineralogy mapped using SEBASS in the Ludwig area is
consistent with the detailed mapping by Harris and Einaudi (1982). Einaudi and Burt
(1982) propose that Cu skarns alteration zonation is characterised by the development of
grandite and diopside. The SEBASS data were used to map the distribution of grandites,
though the identification of diopside rather than hedenbergite (characteristic of Zn skarns)
was not available. Mapping grandite distribution effectively delineates both the skarnoid
and late Fe-skarn alteration in the metasdiments (exoskarn). Furthermore, the progressive
enrichment in Fe away from the contact with the Northern Batholith is consistent with
metasomatic fluids emanating from the granodiorite. The areas of most Fe-richment
coincide with Cu mineralisation.

The porphyry-related alteration mineralogy mapped in the Northern Batholith using


SEBASS is consistent with the detailed mapping by Dilles and Einaudi (1992). In
particular, SEBASS was used to map the distribution of albite-oligoclase and oligoclase-
andesine feldspar, which appear to be important in defining the alteration zonation. This
variation in feldspar mineralogy is related to changes in the Na-Ca composition of the
alteration fluids. These fluids are proposed by Dilles and Einaudi (1992) to have involved
(early) Ca-rich fluids from the carbonate-rich wall rocks and deeper (later) saline fluids
from a magma source. This is consistent with the general model proposed by Einaudi and
Burt (1982) for endoskarn formation with progressive addition of calcium in the felsic
rocks towards the limestone contact. In oxidising conditions, epidote-quartz is
preferentially developed which may explain the SEBASS results (Figure 43 - A). That
is, there is an abrupt change from epidote-quartz in the granodiorite positioned between
the limestones and a block of carbonate-rich metasediments within the granodiorite to
oligoclase-andesine and the absence of epidote and quartz on the other side of this block.
More work is planned later and will examine other aspects of the porphyry systems at
Ann-Mason (Yerington) using SEBASS and AVIRIS data.

62 CSIRO Exploration and Mining Report 734R


13. CONCLUSIONS

The salient results of this study of SEBASS data collected from areas of skarn and
porphyry style alteration in then Yerington district, Nevada, include:

1. The SEBASS data are of very high quality (signal to noise and spectral calibration).
Instrument noise is essentially not apparent in the raw data but is enhanced through
methods like MNF. Most of this noise comprises column banding/striping related to
the 128 by 128 detector “pushbroom” array nature of the instrument. A slightly
oblique line striping is also apparent. Over 40 MNF bands of the 124 available
spectral bands contain geologic information. Removing 52 of the noisiest bands as
well as those at the base of atmospheric-line absorptions did not reduce the number of
good MNF bands. Nor did it remove all the column or oblique line striping;

2. The limitation of conducting blackbody calibration (hot and cold) of the instrument at
the beginning and end of each flight-line only (scanning systems like TIMS conduct it
for each scan-line) did not result in any scene artefacts;

3. The InScene Atmospheric Correction (ISAC) method for reducing the radiance at
sensor SEBASS data to surface radiance is effective in correcting for the additive
(primarily up-welling) and multiplicative (transmission) atmospheric effects. The
ISAC model assumptions of a homogeneous atmospheric profile and no changes in
path-length were not a problem even for variations in topographic relief of >200 m.
However, the pixel-spectra were compromised by a lack of proper correction for
atmospheric continuum-absorption and reflected down-welling radiance from pixels
with a large emissivity (reflectance) contrast. This meant that a Planck Blackbody
Radiation Function could not be fitted accurately to the pixel spectra and both residual
atmospheric-line absorptions and less overall spectral contrast were apparent in many
pixel spectra;

4. Separation of the temperature component from the emissivity component of the


measured pixel surface radiance (temperature-emissivity separation or TES) can be
approximated using a hull normalisation method, like continuum-removal in ENVI.
However, TES was found not to be necessary in the generation of mineral and mineral
chemistry maps for dominating minerals like garnets and carbonates.

5. The continuum-removed surface radiance spectra closely match field emissivity


spectra of related field samples.

6. The minerals identified included:


• Quartz,
• Grandite (andradite and grossular);
• Clinopyroxene (diopside/hedenbergite);
• Plagioclase feldspar (albite-oligoclase and oligoclase-andesine);
• Carbonate (calcite and dolomite);
• Epidote;
• Amphibole (actinolite/hornblende);
• Gypsum;
• Kaolinite;

CSIRO Exploration and Mining Report 734R 63


• Muscovite; and
• Clay.

7. The solid solution chemistry of minerals like grandite Fe-Al chemistry is measured by
calculating the first derivative of fitted a 3rd order polynomial to the wavelength
segment between 11.0 and 11.6 µm.

8. The Cu skarn alteration is clearly identified and delineated by the presence of grandite
(Ca-rich garnet) and diopside. Early skarnoid shows progressive increase in grandite
Fe content away from the associated granodiorite (source of hydrothermal fluids).

9. Cu skarn mineralisation is characterised by near pure andradite. Mapping grandite Fe-


Al chemistry provides a method for targeting potential Cu mineralisation.

10. Porphyry alteration in the granodiorite shows a progressive increase in Ca content of


the plagioclase feldspars (oligoclase-andesine) towards the contact with the
limestones, though nearest the contact is developed epidote and quartz. Deeper in the
granodiorite, the plagioclase is more Na-rich (albite-oligoclase). Muscovite is poorly
developed throughout as K is taken up largely in biotite.

11. A comparison of the mineral information content of VNIR-SWIR AVIRIS data and
TIR SEBASS data for the skarn and porphyry styles of alteration showed that AVIRIS
could be used to discriminate most of the skarn alteration zones though could not be
used to identify the non-OH-bearing silicate mineralogy.

14. ACKNOWLEDGMENTS

The 1999 SEBASS airborne “group shoot” campaign to Nevada was supported by
Aerospace Corporation, CSIRO EOC, CSIRO Exploration and Mining, Sumitomo Metal
Mining Co. Ltd., ERSDAC, Perry and Associates, Noranda Inc, University of Nevada and
Stanford University. Sumitomo Metal Mining, ERSDAC and CSIRO Exploration and
Mining supported the field October 1999 campaign. Masashi Endo (Sumitomo) assisted
in the fieldwork. Flight line and geological planning for the Yerington test site was
assisted by Emeritus Professor Ron Lyon (Stanford University). Professor Jim Taranik
and Amber from University of Nevada provided useful information about the regional
geology and field logistics. Stephen Young (Aerospace) provided useful information
about the ISAC algorithm while Patricia Lew (Aerospace) organised the preliminary
processing and shipment of the SEBASS data to Australia. Mark Favas (CSIRO E&M)
read the raw SEBASS data from DLT4000 tape to disk. Mike Caccetta (CSIRO E&M)
ensured that the Aerospace software worked on the CSIRO NT computers and then
processed all the SEBASS data to surface radiance. Jenny Wilson organised the copying
of all the SEBASS data to CD and distribution to other group shoot members. Mike Hart
conducted the XRD measurements. Useful discussions about the SEBASS results were
conducted with Rob Hewson (CSIRO E&M) and Ron Lyon (Stanford). Rob Hewson and
Peter Hick (CSIRO E&M) reviewed the final manuscript. To all of these people and
organisations, we extend our sincere appreciation and thanks.

64 CSIRO Exploration and Mining Report 734R


15. FURTHER WORK

Areas for future work include:

1. Detailed study of the four SEBASS lines that intersect the Ann-Mason porphyry
system;
2. Collect low-altitude hyperspectral VNIR-SWIR data from the Yerington test site for
comparison with the SEBASS data at comparable pixel resolution; and
3. TIR spectral characterisation of feldspar minerals including adularia and the complete
spectrum of Na to Ca plagioclase.

16. REFERENCES

Agar, R.A., 1994. Geoscan airborne multispectral scanners as applied to exploration for
Western Australian diamond and gold deposits, Proceedings 10th Thematic
Conference on Geological Remote Sensing, San Antonio, pp. 651.
Aronson, J.R., Emslie, A.G., Allen, R.V., and McLinden, H.G., 1967. Studies of the
middle- and far-infrared spectra of mineral surfaces for application in remote
compositional mapping of the moon and planets. Journal of Geophysical
Research, Vol. 72, pp. 687-703.
ASTER. 2000. ASTER web page: http://asterweb.jpl.nasa.gov/
Barducci, A., and Pippi, I., 1996. Temperature and emissivity retrieval from remotely
sensed images using the “Grey Body Emissivity” method. IEEE Transactions on
Geoscience and Remote Sensing, Vol. 34, No. 3, pp. 681-695.
Bianchi, R., and Marino, C.M., 1994. CNR LARA Project Italy: MIVIS/MIDAS
environmental airborne hyperspectral remote sensing, Proceedings of the 1st
Airborne Remote Sensing Conference, Strasbourg, pp., 613.
Boardman, J.W. 1993. Automatic spectral unmixing of AVIRIS data using convex
geometry concepts. Proceedings 4th Annual Airborne Geoscience Workshop,
October, 25-29 1993, JPL Publication 93-26, page 11.
Chang, S.H., Westfield, M.J., Lehmann, F., Oertel, D., and Richter, R., 1993. 79-channel
airborne imaging spectrometer, SPIE, Vol. 1937, pp.387.
Conel, J.E., 1969. Infrared emissivities of silicates: Experimental results and a cloudy
atmosphere model of spectra emission from condensed particulate mediums.
Journal of Geophysical Research, Vol. 74, B6, pp. 1614-1634.
Cudahy, T.J., L.B. Whitbourn, P. Connor, P. Mason, and R.N. Phillips 1999a. Mapping
surface mineralogy and scattering behaviour uses backscattered reflectance from a
hyperspectral midinfrared airborne CO2 laser system (MIRACO2LAS). IEEE
Transactions on Geoscience and Remote Sensing, Vol. 37 No. (4), pp. 2019-2034.
Cudahy, T.J., Okada, K., Matsunaga, T., Whitbourn, L.B., and Oikawa, N. 2000. Testing
the accuracy of temperature-emissivity separation methods using airborne bi-
directional 9-11 µm reflectance. International Journal of Remote Sensing
(accepted).
Cudahy, T.J., Okada, K., Ninomiya, Y., and Whitbourn, L.B. 2000a. Mapping feldspar
mineralogy in igneous rocks using airborne hyperspectral 9-11 µm reflectance.
International Journal of Remote Sensing (accepted).

CSIRO Exploration and Mining Report 734R 65


Crowley, J.K., and Hook, S.J., 1996. Mapping playa evaporite minerals and associated
sediments in Death Valley, California, with multispectral thermal infrared images.
Journal of Geophysical Research, Vol. 101, No. B1, pp. 643-660.
Dilles, J.H., and Einaudi, M.T., 1992. Wall-rock alteration and hydrothermal flow paths
about the Ann-Mason porphyry copper deposit, Nevada – A 6-km vertical
reconstruction. Economic Geology, Vol. 87, pp. 1963-2001.
Einaudi, M.T., 1977. Petrogenesis of the copper-bearing skarn at the Mason Valley Mine,
Yerington district, Nevada. Economic Geology, Vol. 72, pp. 769-795.
Einaudi, M.T., and Burt, D.M., 1982. Terminology, classification and composition of
skarn deposits. Economic Geology Special Issue, Vol. 77, No.4, pp. 745-753.
Green, A.A., and Craig, M.A., 1985. Analysis of aircraft spectrometer data with
logarithmic residuals. Proceedings of the Airborne Imaging Spectrometer Data
Analysis Workshop, April 8-10, JPL Publications 85-41, pp. 111-119.
Green, A.A., Berman, M., Switzer, P. and Craig, M.D., 1988. A transformation for
ordering multispectral data in terms of image quality with implications for noise
removal. IEEE Transactions on Geoscience and Remote Sensing, Vol. 26, No. 1,
pp. 65-74.
Hackwell, J.A., Warren, D.W., Bongiovi, R.P., Hansel, S.J., Hayhurst, T.L., Mabry, D.J.,
Sivjee, M.G., and Skinner, J.W., 1996. LWIR/MWIR Imaging Hyperspectral
sensor for airborne and ground-based remote sensing. SPIE Vol. 2819, pp. 102-
107.
Haynes, D.W., Cross, K.C., Bills, R.T., Reed, M. 1995. Olympic Dam ore genesis: A
fluid mixing model. Economic Geology, Vol. 90, pp. 281-307.
Harris, N.B., and Einaudi, M.T., 1982. Skarn deposits in the Yerington district:
Metasomatic skarn evolution near Ludwig. Economic Geology, Vol. 77, pp. 877-
898.
Henry, R.L., 1948. The transmission of powdered films in the infrared. Journal Optical
Society of America, Vol. 38, pp. 775-789.
Hook, S.J., Gabell, A.R., Green, A.A., and Kealy, P.S., 1992. A comparison of
techniques for extracting emissivity information from thermal infrared data for
geologic studies. Remote Sensing of Environment, Vol. 42, pp. 123-135.
Hook, S.J., and Kahle, A.B., 1996. The micro fourier transform interferometer (mFTIR) –
A new field spectrometer for acquisition of infrared data of natural surfaces.
Remote Sensing of Environment, Vol. 56, pp. 172-181.
Hunt, G.R., and Logan, L.M., 1972. Variation of single particle mid-infrared emission
spectrum with particle size, Applied Optics, Vol. 11, No.1, pp. 142-147.
Hunt, G.R., and Vincent, R.K., 1968. The behaviour of spectral features in the infrared
emission from particulate surfaces of various grain sizes. Journal of Geophysical
Research, Vol. 73, No. 18, pp. 6039-6046.
Ishihara, S. 1981. The granitoid series and mineralisation. Economic Geology 75th
Anniversary Edition, pp. 458-484
Johnson, B.R., 1998. Inscene Atmospheric Compensation: Application to SEBASS data
collected at the ARM site. Aerospace Report No. ATR-99(8407)-1 Part I.
Kahle, A.B., and Alley, R.E., 1992. Separation of temperature from emittance in
remotely sensed radiance measurements. Remote Sensing of Environment, Vol. 42,
pp. 107-111.
Kahle, A.B., Madura, D.P., and Soha, J.M., 1980. Middle infrared multispectral aircraft
scanner data: Analysis for geological applications. Applied Optics, Vol. 19, No.
14, pp. 2279-2290.
Kealy, P.S., and Gabell, A.R., 1990a. Calculation of emissivities for lithological mapping
using thermal infrared multispectral scanner (TIMS) data in the Kimberley region

66 CSIRO Exploration and Mining Report 734R


of W.A.. Proceedings of the 5th Australasian Conference Perth, Western Australia,
8-12 October 1990, pp. 362-370.
Kealy, P.S., and Gabell, A.R., 1990b. Estimation of temperature and emissivity using
alpha coefficients. In Proceedings of the 2nd TIMS Workshop, JPL Publication,
90-55, pp. 11-16.
Kealy, P.S., and Hook, S.J., 1993. Separating temperature and emissivity in Thermal
Infrared Multispectral Scanner Data: Implications for recovering land surface
temperatures. IEEE Transactions on Geoscience and Remote Sensing, Vol. 30,
No. 6, pp. 1155-1164.
Korb, A.R., Dybwad, P., Wadsworth, W., and Salisbury, J.W., 1996. Portable fourier
transform infrared spectroradiometer for field measurements of radiance and
emissivity. Applied Optics, Vol. 35, No. 10, pp. 1679-1692.
Lang, H.R., Adams, S.L., Conel, J.E., McGuffie, B.A., Paylor, E.,D., and Walker, R.E.,
1987. Multispectral remote sensing as stratigraphic and structural tool, Wind
River Basin and Big Horn Basin areas, Wyoming. The American Society of
Petroleum Geologists Bulletin, Vol. 71, No. 4, April, 1987.
Launer, P.J., 1952. Regularities in the infrared absorption spectra of silicate minerals.
American Mineralogist, Vol. 37, pp. 764-784.
Logan, L.M., Hunt, G.R., Salisbury, J.W., and Balsamo, S.R., 1973. Compositional
implications of Christiansen frequencies for infrared remote sensing. Journal of
Geophysical Research, Vol. 78, No. 23, pp. 4983-5003.
Lyon, R.J.P., 1964. Evaluation of infrared spectrophotometry for compositional analysis
of lunar and planetary soils, Part 2: Rough and powdered surfaces, Final Report,
NASA Rep. CR-100, 115 pp.
Lyon, R.J.P., 1965. Analysis of rocks and minerals by reflected infrared radiation.
Economic Geology, Vol. 60, pp. 715-736.
Lyon, R.J.P., and Burns, E.A., 1963. Analysis of rocks and minerals by reflected infrared
radiation. Economic Geology, Vol. 58, pp. 274-284.
Matsunaga, T., 1994. A temperature-emissivity separation method using an empirical
relationships between the mean, the maximum and the minimum of the thermal
infrared spectrum. Journal of the Remote Sensing Society of Japan, Vol. 28, No.
40, pp. 28-39.
Moore, J.G., 1969. Geology and mineral deposits of the Lyon, Douglas and Ormsby
counties, Nevada. Nevada Bureau of Mines and Geology, Bulletin 75, 45 pages.
Nakano, T., Yoshino, T., Shimazaki, H., and Shimizu, M., 1994. Pyroxene composition
as an indicator in the classification of skarn deposits. Economic Geology, Vol.
89, pp. 1567-1580.
Nash, D. B. and Salisbury, J. W., 1991, Infrared reflectance spectra of plagioclase
feldspars: Geophysical Research Letters, V 18, pp. 1151-1154.
NBMG, 1996. Nevada 1:100,000 scale digital raster geology and cadastral information
(CD), Nevada Bureau of Mines Open File Report 96-5
Ninomiya, Y. 1995. Quantitative estimation of SiO2 content in igneous rocks using
thermal infrared spectra with a neural network approach. IEEE Transactions on
Geoscience and Remote Sensing, Vol. 33, pp.684-691.
Palluconi, F., D., and Meeks, G.R., 1985. Thermal Infrared Multispectral Scanner
(TIMS): An investigators guide to TIMS data, JPL-Publication 85-32, 32 pages.
Pollard, P.J. 1995. Geology of rare metal deposits: An introduction and overview.
Economic Geology, Vol. 90, pp. 489-494.
Price, J.C., 1984. Land surface temperature measurements from the split window
channels of the NOAA 7 advanced very high-resolution radiometer, Journal of
Geophysical Research, Vol. 89, No. 33, pp. 5039-5044.

CSIRO Exploration and Mining Report 734R 67


Proffett, J.M., 1977, Cenozoic geology of the Yerington district, Nevada, and implications
for the nature and origin of Basin and Range faulting. Geological Society of
America Bulletin, Vol. 88, pp. 247-266.
Proffett, J.M. and Dilles, J.H., 1984. 1:24,000 scale geologic map of the Yerington
district, Nevada. Nevada Bureau of Mines and Geology, Map 77, 89557-0088.
Realmuto, V.J., 1990. Separating the effects of temperature and emissivity: emissivity
spectrum normalisation. Proceedings of the Second Thermal Infrared
Multispectral Scanner (TIMS) Workshop. June 6, 1990, JPL Publication 90-55,
pp. 31-35.
Richter, R., 1994. Derivation of temperature and emittance from airborne multispectral
thermal infrared scanner data. Infrared Physics Technology, Vol. 35, No. 6,
pp.817-826.
Sabine, C., Realmuto, V.J., and Taranik, J.V., 1994. Quantitative estimation of granitoid
composition from thermal infrared multispectral scanner (TIMS) data, Desolation
Wilderness, northern Sierra Nevada, California. Journal of Geophysical Research,
Vol. 99, No. B3, pp. 4261-4270.
Salisbury, J.W. and Eastes, J.W. 1985. The effect of particle size and porosity on spectral
contrast in the mid-infrared. Icarus, Vol. 64, pp. 586-588.
Salisbury, J.W., Hapke, B., and Eastes, J.W., 1987. Usefulness of weak bands in
midinfrared remote sensing of particulate planetary surfaces. Journal of
Geophysical Research, Vol. 92, No. B1, pp. 702-710.
Salisbury, J.W., and Walter, L.S., 1989. Thermal infrared (2.5-13.5 µm) spectroscopic
remote sensing of igneous rock types on particulate planetary surfaces. Journal of
Geophysical Research, Vol. 94, No. B7, pp. 9192-9202.
Spry, P.G. and Wonder, D. 1989. Manganese-rich garnet rocks associated with the
Broken Hill lead-zinc-silver deposit, New South Wales, Australia. Canadian
Mineralogist, Vol. 27, pp. 297-327.
Takashima, T. and Masuda, K., 1987. Emissivities of quartz and Sahara dust powders in
the infrared region (7-17 µm). Remote Sensing of Environment, Vol. 23, pp. 51-
63.
Vincent, R.K., and Thomson, F., 1972. Spectral composition imaging of silicate rocks.
Journal of Geophysical Research, Vol. 77, No. 14, pp. 2465-2472.
Wald, A.E., and Salisbury, J.W., 1995. Thermal infrared directional emissivity of
powdered quartz. Journal of Geophysical Research, Vol. 100, No. B12, pp.
24,665-24,675.
Warner, T.A., and Levandowski, D.W., 1989. Optimum band selections for estimating
emittance using TIMS data. Proceedings of the Second Thermal Infrared
Multispectral Scanner (TIMS) Workshop. June 6, 1990, JPL Publication 90-55,
pp. 26-30.
Watson, K., 1992a. Spectral ratio method for measuring emissivity. Remote Sensing of
Environment, Vol. 42, pp. 113-116.
Watson, K., 1992b. Two-temperature method for measuring emissivity. Remote Sensing
of Environment, Vol. 42, pp. 117-121.
Whitbourn, L.B., Phillips, R.N., James, G., and O’Brien, M.T. 1990. An airborne
multiline CO2 laser system for remote sensing of minerals. Journal of Modern
Optics, Vol. 37, No. 11, pp. 1865-1872.
Windeler, D.S., 1993. Garnet-pyroxene alteration mapping in the Ludwig skarn
(Yerington, Nevada) with Geoscan airborne multispectral data/. Photogrammetric
Engineering and Remote Sensing, Vol. 59, No. 8, pp. 1277-1286.

68 CSIRO Exploration and Mining Report 734R


Windeler, D.S., and Lyon, R.J.P., 1991. Discriminating dolomitization of marble in the
Ludwig skarn near Yerington, Nevada using high-resolution airborne infrared
imagery. Photogrammetric Engineering and Remote Sensing, Vol. 57, No. 9, pp.
1171-1177.
Wyman, D and Kerrich, R. 1988. Alkaline magmatism, major structures and gold
deposits: Implications for greenstone belt gold metallogeny. Economic Geology,
Vol. 83, pp. 454-461
Young, S.J.,1998. Inscene Atmospheric Compensation: Application to SEBASS data
collected at the ARM site. Aerospace Report No. ATR-99(8407)-1 Part II.

CSIRO Exploration and Mining Report 734R 69


17. APPENDICES

A1. SEBASS Spectral Band Subsetting

Table A1: SEBASS band centres and selected band used (outside atmospheric line
absorptions). * is deselected; 1 is selected.
bands available selected/deselected bands available selected/deselected bands available selected/deselected
7.57447 * 10.0038 1 11.9387 *
7.63993 * 10.0532 1 11.9799 1
7.70483 * 10.1023 1 12.0209 1
7.76917 1 10.1512 1 12.0618 *
7.83298 * 10.1998 1 12.1025 *
7.89626 * 10.2482 1 12.1431 *
7.95903 1 10.2964 1 12.1835 1
8.0213 * 10.3443 1 12.2238 1
8.08308 1 10.392 1 12.264 *
8.14439 * 10.4395 1 12.304 *
8.20523 1 10.4868 1 12.3439 *
8.26561 1 10.5338 * 12.3837 *
8.32555 * 10.5806 1 12.4233 *
8.38505 * 10.6272 1 12.4628 *
8.44413 1 10.6736 1 12.5021 *
8.50278 * 10.7198 1 12.5414 *
8.56103 * 10.7658 1 12.5805 *
8.61888 1 10.8116 1 12.6195 *
8.67633 * 10.8572 1 12.6583 *
8.7334 1 10.9026 1 12.697 1
8.79008 * 10.9478 * 12.7357 *
8.8464 1 10.9928 * 12.7741 *
8.90236 * 11.0376 1 12.8125 *
8.95795 1 11.0823 * 12.8508 *
9.0132 1 11.1267 * 12.8889 *
9.0681 1 11.171 * 12.9269 *
9.12267 1 11.2151 1 12.9648 *
9.1769 1 11.259 * 13.0026 *
9.23081 1 11.3027 1 13.0402 *
9.28439 * 11.3463 1 13.0778 1
9.33767 1 11.3896 1 13.1152 1
9.39063 1 11.4328 1 13.1525 1
9.44329 1 11.4759 * 13.1897 1
9.49565 1 11.5188 * 13.2268 *
9.54772 1 11.5615 1 13.2638 *
9.59949 1 11.604 1 13.3007 1
9.65098 1 11.6464 1 13.3375 *
9.7022 1 11.6886 * 13.3742 1
9.75313 1 11.7307 1 13.4108 *
9.8038 1 11.7726 1 13.4472 *
9.85419 1 11.8144 1 13.4836 *
9.90433 1 11.856 1 13.5198 *
9.9542 1 11.8974 *

70 CSIRO Exploration and Mining Report 734R


A2. XRD Mineralogy

Table A2: Automatic Mineral identification of the XRD Traces of Selected Samples.

sample mineral 1 mineral 2 mineral 3 mineral 4 mineral 5 mineral 6


number
Y001 quartz augite/diopside albite, calcian, magnesiohornblende
ordered
Y002 calcite quartz
Y003 andradite
Y004 tremolite-actinolite calcite iron (III) andradite unidentified
chamosite
(chlorite)
Y005 calcite
Y006 calcite unidentified
Y007 diopside garnet
(uvarovite)
Y008 andradite quartz
Y009 garnet (uvarovite, montmorillonite
goldmanite)
Y010 epidote calcite quartz unidentified
Y011 albite-andesine augite-aegerine epidote clinochlore (ferroan),
(acmite)/diopsid chamosite
e (magnesian)
Y012 calcite hornblende (Fe- quartz elyite?
pargasite) (CuPb4(SO4)(OH)8
Y013 albite quartz calcite chlorite unidentified
Y014_5 grossular
0m
Y015 grossular
Y016A albite quartz unidentified
Y017A grossular hedenbergite?
Y017B grossular epidote
Y018 uvarovite/andradit chalcosine?
e/grossular (Cu2S)
Y019 goldmanite/uvaro diopside/augite
vite/andradite
Y021 albite (calcian, epidote augite ferropargasite
ordered) -
andesine
Y022 uvarovite/goldma quartz diopside/augit
nite e
Y023B quartz anorthite? epidote
(sodian,
ordered)
Y026 garnet (Ca-Fe) quartz epidote potassium pargasite
(hornblende)
Y027 andradite unidentified
Y028 dolomite (75% of calcite (25% of
sample), ferroan sample)
Y029 andradite unidentified
Y030 hedenbergite, ferropargasite grossular unidentified
magnesian
Y030_1 andradite quartz
00
Y031 dolomite (95% of calcite (5% of
sample), ferroan sample)
Y032 hedenbergite, hornblende, unidentified
magnesian magnesian,
ferroan
Y033 bytownite hematite nontronite

CSIRO Exploration and Mining Report 734R 71


Y033_1 albite, calcian - quartz magnesiohorn dolomite
00 ordered blende,
ferroan
Y034B bytownite unidentified
Y036 magnesiohornble uvarovite
nde, ferroan
Y037 quartz magnesiohornbl andesine- microcline unidentified
ende, ferroan albite
Y038A ferromagnesian- uvarovite/goldm clinochlore
hornblende/tremol anite (1MIIB,
ite- ferroan)
actinolite/ferropar
gasite
Y038E albite, calcian - uvarovite augite cordierite unidentified
ordered
Y039A albite, calcian - magnesiohornbl clinochlore quartz unidentified
ordered ende, ferroan
Y039B albite, calcian - magnesiohornbl quartz clinochlore dolomite? unidentified
ordered ende, ferroan
Y039C albite, calcian - quartz magnesiohorn clinochlore dolomite?
ordered blende,
ferroan
Y042A albite, calcian - quartz magnesiohorn clinochlore
ordered blende,
ferroan

72 CSIRO Exploration and Mining Report 734R


A3. Log Residuals

Some preliminary testing by one of the authors (K. Okada) found that log residuals (Green
and Craig, 1985) applied to raw SEBASS data yielded mineralogically interpretable
spectral signatures as well as some temperature-independent imagery.

In theory, log residuals is not well suited to correcting TIR “radiance at sensor” data to
surface emissivity. This is because log residuals implements a series of normalisations,
including: (i) pixel-spectrum-mean normalisation; and (ii) a scene-wavelength-mean
normalisation. However, TIR radiance at sensor data has at least two components that
cannot be simply corrected through these normalisations, namely: (i) an additive up-
welling radiance component; (ii) a wavelength-pixel-dependent blackbody temperature
curve. For such reasons, the “thermal log residuals” method was developed by CSIRO
(Hook and others, 1992). However, both thermal log residuals and log residuals are based
on scene statistics, resulting in pixel spectra that depend on the scene-mean spectral shape.

Figure 45f shows log residual pixel spectra of selected pixels and Figure 45g shows ISAC
corrected, continuum-removed surface radiance spectra of the same pixels from a part of
the SEBASS Ludwig line. Even though, the log residual spectra show diagnostic features
that permit their identification as quartz-rich (blue) gypsum-rich (coral) and garnet-rich
(red), all spectra show garnet and to a lesser degree quartz spectral signatures. In contrast,
the ISAC corrected spectra do not show this pervasive development garnet and quartz
features. Furthermore, the green log residual spectrum from a dolomite-rich pixel (Figure
45f), does not clearly show the diagnostic sharp, carbonate emissivity low at 11.2 µm,
which is shown by the the ISAC corrected spectrum (Figure 45g). These results emphasis
the scene-mean dependence on the log residual spectra on the strong development of
garnet and quartz in this scene which appears to have affected all pixel spectra, even
masking some diagnostic minerals features such as carbonate.

The log residual images of the Ludwig show varying success in separating the surface
temperature from the surface emissivity information. Generally, scenes without apparent
topographic effects demonstrate effective TES. This appears to be the case for SEBASS
bands between 10.5 and 12 µm (Figures 45d and 45d) which contrasts with the albedo
(brightness temperature) image (Figure 45a), and which highlights the cold and hot sides
of topographic relief. Poor TES has occurred at other wavelengths, with “positive” and
“negative” residual temperature variations being typical at shorter and longer
wavelengths, respectively (Figures 45b and 45e). These results highlight the limitations
of using log residuals for TES mainly because the wavelength-dependent nature of the
blackbody temperature curve is poorly modelled through simple normalisations.

CSIRO Exploration and Mining Report 734R 73


Figure 50: Various SEBASS products generated using Log Residuals: (a) albedo
image; (b) log residual band at 9.0132 µm showing residual “positive”surface
temperature variation; (c) log residual band at 10.8572 µm showing little or no
residual surface temperature variation; (d) log residual band at 11.3896 µm
showing no residual temperature variation (emissivity only); (e) log residual band
at 12.4233 µm showing residual “negative”surface temperature variation; (f) log
(KWIK) residual spectra of selected pixels shown on (a); (g) ISAC “emissivity”
spectra (surface radiance spectrum with continuum-removal) of selected pixels
shown on (a).

74 CSIRO Exploration and Mining Report 734R

S-ar putea să vă placă și