Discrimination of surface alteration based on digital
processing of airborne multispectral data
Kazuya Okada, Takeo Yokoyama
Sumi tomo ~1etal Mining Co., Ltd.
5-11 3 Shimbashi Minato-ku Tokyo 105
JAPAN
and
Katsuo Arai
Earth Resources Satellite Data Analysis Center
2 4-5 Azabudai Minato ku Tokyo 106
JAPAN
Comission VII
ABSTRACT
This
study
concerns
the
analysis
of
spectra
from
a
high-spectral resolution airborne spectroradiometer.
Airborne
data are severely affected by two factors, topographic effects
and atmospheric effects.
These effects must be removed to
compare
airborne
data
with
laboratory
measurements
of
reflectance
spectra.
Three
techniques
were
applied
in
analyzing the Yerington and Virginia City
(Nevada,
U.S.A.)
data;
Logarithmic Residual,
Flat
Field Correction and Band
Ratio.
Interpretation
based
upon
a
combination
of
the
Logarithmic Residual analysis and field measurements provided
successful delineation of the alteration mineralogy.
INTRODUCTION
High spectral
resolution airborne and field surveys were
carried out over the Yerington and Virginia City mining areas
in Nevada.
The Yerington district was picked for research
because it is a major, well exposed porphyry copper deposit,
that
is
geologically
(Dilles,
1984,
Carten,
1986)
and
spectrally (Green and Lyon, 1984) well-studied.
On the other
hand,
the Virginia City district
is
a
typical
example of
vein-type silver and gold deposits, that is also geologically
(Thompson,
1956)
and
spectrally
(Elvidge
and
Lyon,
1984)
,~ell-studied .
Recent studies with high spectral resolution remote sensing
data
have
shown
that
the
identification
of
individual
alteration
minerals
is
possible
(Marsh
and McKeon,
1983).
However,
researchers
have
been
beset
with many
perplexing
difficulties
during
the
analysis
of
spectral
curve-shape.
Radiance spectra acquired in aircraft based spectrometers bear
little resemblance to laboratory measurements of reflectance
spectra.
Though the cause or mechanism of these phenomena has
yet to be sufficiently clarified, topographic and atmospheric
effects can be considered as the main factors.
In
point
of
mineral
exploration,
airborne
data must
be
converted into a form similar to reflectance spectra.
Several
practical methods for extracting mineral-related features have
been developed (Green and Craig,
1985).
In this study the
authors appl ied three techniques ; Logarithmic Residual,
Flat
Field Correction and Band Ratio.
1.
2.
INSTRUMENTATION AND DATA ACQUISITION
The airborne
survey was flown with the GER
(Geophysical
Environmental
Reserch
Corp.)
visible
and
infrared
spectroradiometer
system.
This
system
has
512
channels
covering the visible to near infrared (VNIR; 400nm to 1000nm)
region, where the iron minerals have characteristic spectral
features.
The short wave infrared (SWIR; 2000nm to 2500nm)
region,
where the
clay and other alteration minerals have
distinctive absorption features, is covered with 64 channels.
Each channel has an individual detector and the measurements
are
m de
simultaneously.
The
airc aft
was
flown
at
the
barometric altitude of 8500 feet for the major part of the
survey.
Three traverses were flown at a hi
er al tit ude of
20000 feet in order to study the effect of the pixel-size.
The flying speed was about 110 miles per hour.
The 8500 feet
ASL
is rou
ly translated to 2000 feet
above the average
ground
surface
of
6500
feet
elevation,
so
that
an
instantaneous field of view (IFOV) was 60 feet square.
At a
hi
er altitude, the IFOV was 60 feet along the flight path
and 400 feet perpendicular to it.
The airborne survey was made between 9: 00 AM and 2: 00 PM
fixing upon clear days; May 29 and 30, 1986.
The field survey was carried out with the GER IRIS Mk-II and
Mk-III system, which cover the 350nm to 3000nm region with the
spectral bandwidth of 2nm (350nm to 1000nm) to 4nm (1000nm to
3000nm).
IRIS is a dual beam system that looks at a target
and a reference material (barium sulfate plate) simultaneously
so that reflectance spectra can be measured with immunity to
variations
in
the
source
illumination.
This
immunity has
merits in the fields because a measurement takes not less than
10
minutes,
which
depends
on
the
i lumination
level,
and
variations in this period in the solar illumination cannot be
disregarded.
Field
spectra
were
converted
to
percent
reflectance
by
dividing the target radiance by the reference spectra.
3.
GEOLOGY
The
Yerington
district
is
composed
of
Cretaceous
metamorphosed
volcanics
and
marine
sediments
which
were
intruded
during
the
Jurassic
by
Yerington
batholith
and
Shamrock batholith,
and of early Mesozoic pluton.
Regional
metamorphism converted these Mesozoic rocks into greenschist
facies;
chlorite,
epidote,
zoisite, albite, quartz and clay
minerals.
The Mesozoic system is unconformably overlain by
Oligocene
ignimbrite
and
Miocene
andesite,
basalt
and
sedimentary rocks.
Figure 1 shows a
geologic map of the
Yerington disrict (Oilles, 1984).
The
Virginia
City
district
is
composed
mainly
of
post-Cretaceous
volcanics.
The
most
extensively
exposed
formations are the Miocene Kate Peak formation and the Miocene
Alta
formation
composed
of
andesitic
lava
and pyroclastic
rocks.
The Alta formation and the lower Kate Peak formation
were
subjected to
hydrothermal
alteration;
propylitization,
a gillization
and
al nitization.
Major
silver
and
gold
deposits were mined from the Comstock Lode district.
Figure 2
shows a geologic map of the Virginia City district (Thompson,
1956).
11930
15
EXPLAIIA TI ON
~KE
-r'" F.ult
ASIA
Qu.tO AlluvllJlll
TertE2]
V~!~~~~l!ry
LUDWIG
Ij
Rocks
Cr.t~ Gronlte
o
PlutonIc Rocks
t~::;:)Sh4..,.od 8athollth
c=::J GranodIorite
c::::::::::J Porphyry
"
~ ~ v~~~:~~~~ ~!ring
~ f~TIP~;~:~:~\~:e
~ Verlngton 8othollth
J-l1D Granodlori ta
J-li
Sed lmentary Rocks
IT:rfl
1i
~\I'5Suk Olorite
}
~ A~d~! ~~;F~! ~:~:
o
10
t
15 KM
,
Figure 1 Geologic map of the Yerington district
(after Dilles)
119..... 5·
EXPlANATtON
39"30' ,-----~
,ndlJ
HOIOC<,..
}
AlJuyjum
a
YOYAf;er 10clu
lnclud_z Sfumb{)<llt Hllu
Plelstocene
QUATERNARY
PI,iSlo" ... ) QUATERNARY
and
ANt>
TERTIAKY('!J
AAyoUl~
PI~ne{'~)
(P1Jo«,,~? aM l'teuloC4~ J.
14"",·
.. ItS A,.tlttsiu(P1,lsllX'itl&4?j.• Ad
JrkOltlkut ~ok 0I1t4~ &u.olr
(RriI'OC"nI~J
lOTI:]
fI.iu"",,,,, }QUATERNAR1f
Ro_
ANU
TERTIA RY
loci
lAUt.eIOWA FormaltoD
PI",«",
~
"'"",sWN Df bU/1Jric
lIII'UIu"~ .ru:t haulr
CEJ
E:J
}
Pho«nc
WashinJ;lon Hill Rhyoblt
} fI"'~.n'
and
MIOcene
COAl Valley Fornu lion
E]
Kate ruk Formalion
'Rhyodoclu
'0
.nd~,j't
flows.
b""CCIIU.IIMut'~siw,tXJt.r
B
o" ... io)()n vanodlOriu:.
TERTIARY
MK>crne
~
All" form,lIKln
Chit!ly IJ'IO('UI.(' (JOWl IJfld
flowbrtt"cil&J
~
H.n(o,d )hll RhyolHe Tuff
I
CRETACEOUS
AND
TRIASSIC(?)
~
~
hkaoche:d IQCA.sof the J.:alt f'c,ik
and
Aha
FQrm;Jl~n~
Comloct
Normal
r... ult
IJ,ashf'd "",h~rr 'PP'OHI'I'IQllly
Ivc.tI~;J
&r Q,u;l b",U UII
d()w'*'hTC ....1t ~idC!
eDDH-4
o
1
2 MILES
f~-'-,-----'
- " - - - ' - - - - ' - - - - -_ _ _ _ _- '
l19~
GtolOICY mu·.Jdu,,J .. llu 1 tlf>lllpwn(l '1S6. pi
3gel~'
30
..
Sampk locality
J)
Figure 2 Geologic map of the Virginia City district
Thompson)
VII
(after
4.FLAT FIELD CORRECTION
Flat field correction is a method that uses a pixel or a
number
of
pixels which are spectrally flat
as
references.
Every pixel is devided by the reference spectra.
Basalt Hill
located in the southwestern part of the Virginia City area is
underlain
literally
by
basaltic
rocks.
Figure
3
shows
a
reflectance spectrum of a basalt sample which was collected at
Basalt Hill.
The reflectance curve is flat in the interesting
region (500 to 2500nm).
Radiance spectra acquired over Basalt
Hill were employed as references.
Figure 4 and 5 show an
unprocessed radiance spectrum (SWIR) over an altered area and
the result of Flat Field Correction.
The resulting spectrum
has a distinctive absorption feature of 2210nm clay mineral.
R
E
F
L
E
80
C
T
A
N
C
E
60
40
20
25=00
WAVE LENGTH
Figure 3
3.
R
oq
(* 10-
3
nm
Reflectance spectrum
Basalt
)
2.50
A
D
I
2. 00
A
N
1. 50
C
E
1. 00
R
A
T
I
0
0.50
WAVE LENGTH nm
Figure 4
WAVE LENGTH nm
Unprocessed radiance
spectrum
5.LOGARITHMIC RESIDUAL
Logarithmic
Residual
following relationship.
T
(Log
Figure 5
Residual)
Flat field
spectrum
is
corrected
based
upon
the
(1)
j
where
X iA
Ti
PiA
IA
Radiance In channel A at point i
Topographic factor at point i
Reflectance of the terain in channel
Solar illumination in channel A
The Log Residual
Y
iA
is defined by the
Xi.
X.
A
following
+-
X ..
A
at
point
i
formula.
(2)
where
X. A
M- 1
Xi.
N
X ..
1
Li
og
X iA
LA
1 og
Xi,\.
(MN)
2:.
1
LA
log
Xi,\.
Figure 6 shows the result of Log Residual for the same
The Log Residual spectrum shows a
location as Figure 4 and 5.
feature
of
2210nm
clay.
The
weak
absorption
distinctive
2350nm is cha acteristic of sericite.
centered nea
l
0
G
R
E
S
I
D
U
O. 06
O. 04
O. 02
O. 00
-0. 02
L
-0. 04
WAVE LENGTH nm
Figure 6
Log Residual spectrum
Figure
7
and
8
show
another
comparison
of
Flat
Field
Correction and Log Residual.
Log Residual
spectrum has a
doublet
near
2200nm which
indicates the
characteristic of
kaolinite.
This doublet can be hardly detected even in the
labolatry
measurement
of
rock
samples.
The
Flat
Field
Correction could not extract the doublet.
1. 1
l
0
G
1.0
R
A
O. 04
O. 02
R
E
O.S
O. 00
S
I
D
U -0. 02
T
I
0
A
O. 7
L
WAVE LENGTH nm
Figure 7 Flat field c rrected
spectrum
-0.04
WAVE LENGTH nm
Figure 8 Log Residual
spectrum
Comparing figure 7 to figure 8, so far as extracting the
mineral-related feature,
it is evident that Log Residual is
superior to Flat Field Correctio
Flat Field Corrected spectrum is formularized as follows on
referring to equation (1).
Z
iA
(3 )
The topographic effect can be considered as a constant with
the wavelength, so that the ratio of topographic factors could
not have influence on the curve-shape but shifts the average
level.
The assumption used in equation (1) holds valid when
the
solar
illumination
can
be
regarded
as
a
function
of
wavelength.
The fluctuations with time in the solar illumination would
of
Flat
Field
Correction.
The
reduce
the
efficiency
difference of the data acquisition time between the target
The fluctuations
point and the reference is about 30 minutes.
a
fatal
effect
on
the
Flat
Field
in
this
period
had
Correction.
6.
BAND RATIO
Band Ratio is used widely because it is very convenient to
extract spectral features from a large quantity of data such
as satellite data.
Figure 9 shows a ratio profile of channel
365
(BOOnm)
over channel 220 (600nm).
Chlorophyll pigments
cause strong absorption in the visible region so that
the
vegetation has a
reflectance plateau in the near
infrared
region.
The high values shown in Figure 9 represent vegetated
area.
minerals, all the channels must be
In order to identify the
taken into consideration.
However, Band Ratio is very useful
as an expedient method to throw light on the remote sensing
data at the first stage.
8
0
0
n
m
}''' 'I
I
I
I
I
I
I
I
I
6
-
4
-
2
-
/
6
0
0
n
m
III
0
~A
100
200
300
400
500
POINT
Figure 9
7.
Ratio profile
channel 365 / channel 220
IDENTIFICATION OF ALTERATION MINERALS
Of the three techniques investigated, the Log Residual was
the best.
The resul ting spectra could be directly compared
wi th the ref lectance spectra.
However, visual examinat ion of
the spectral curve-shape takes much time.
There were 94,928
spectra
to
investigate,
so
that
the
identification
of
individual minerals was done automatically by a computer to
some degree.
The algorithm was based upon the combination of
the absorption detection by averaging differentiation and the
correlation with the simulated standard reflectance spectra.
The following are the selected typical samples of Log Residual
spectra and the standard reflectance simulated from the field
measurements by resampling the digital readings into the same
wavelength centers as the airborne spectroradiometer in order
to calculate correlation coefficients.
(a) Alunite
Alunite has
and 2335nm.
absorption
bands
centered
L
E
C
O. 00
T
S
A
N
C
I
D
U
A
L
2210nm
E
F
O. 02
R
E
2165nm,
R
L
0
G
near
E
Yo
-0,04
WAVE LENGTH nm
WAVE LENGTH nm
Figure 10
Log Residual
spectrum, Alunite
Figure 11
Simulated standard
reflectance, Alunite
(b) Calcite
Calcite has an absorption band centered near 2340nm.
R
L
0
O. 02
R
O. 00
G
E
F
L
E
C
E
T
A
N
S
I
D
C
E
U -0. 04
A
L
Yo
-0. 06
WAVE LENGTH nm
Figure 12
Log Residual
spectrum, Calcite
WAVE LENGTH nm
Figure 13
Simulated standard
reflectance, Calcite
(c) Chlorite
Chlorite has an absorption band centered near 2340nm.
L
0
G
O. 02
R
E
0.01
E
F
L
R
E
S
O. 00
I
0
U
A -0.0'1
Figure 14
C
T
A
N
C
E
WAVE LENGTH nm
WAVE LENGTH nm
Log Residual
Figure 15
spectrum, Chlorite
Simulated standard
reflectance, Chlorite
(d) Epidote
Epidote
has
absorption
bands
curve-shape bears resemblance to
16 and 17.
l
0
G
near
2340nm.
The
as shown In Figure
R
E
F
O. 02
L
E
R
E
S
I
D
U
centered
chlorite
C
T
A
O. 00
N
C
E
A -0. 02
L
Yo
WAVE LENGTH nm
WAVE LENGTH nm
Figure 16
Log Residual
spectrum, Epidote
Figure 17
Simulated standard
reflectance, Epidote
(e) Gypsum
Gypsum has a wide and shallow absorption band from near
2140nm to 2310nm which is formed by several absorption bands.
L
0.04~mmmrnmmmrnm=~~n=~~~~
G
0.02
R
E
O. 00
o
S
I -0.02
D
U -0. 04
A
l -0.06
-0.
R
E
F
L
E
C
T
A
,
N
C
E
I
:
:
1
:
~~t'I!!l!i!!lI~g!b!OI!!!I!j."!2~loo
WAVE LENGTH nm
WAVE LENGTH nm
Figure 18
Log Residual
spectrum, Gypsum
Figure 19
Simulated standard
reflectance, Gypsum
(f) Montmorillonite
Montmorillonite has an absorption band centered near 2200nm.
Montmorillonite can be distinguished from sericite by lack of
2350nm absorption.
L
0
G
R
E
F
l
O. 02
E
C
R
E
S
T
O. 00
A
N
C
I
D
U
A
L
E
Yo
WAVE LENGTH nm
Figure 20
Log Residual
spectrum
Montmorillonite
WAVE LENGTH nm
Figure 21
Simulated standard
reflectance
IVlontmori lloni te
Figure 22 Alteration map of the Yerington district
1,19040'"
Background
II
ALunite
+
*
Gypsum
0
Montmori Honi tEl
Ca1.cite
Kao Li nite
39°20'N
x
Serici fa
6
Ch1.orite
0
Epidote
Kao Linite or ALunite
o
39°15'/1
Figure 23 Alteration map of the Virginia City district
5
10km
(g) Sericite and Kaolinite
The typical Log Residual spectra
are shown in Figure 6 and 8.
of
sericite
and
kaolinite
Figure 22 shows the result of the automatic identification
of alteration minerals in the Yerington district.
The notable
alteration pattern composed mainly of epidote and chlorite was
detected
in
the
Yerington
batholith
that
has
undergone
intrusion and mineralization several times.
In the Tertiary
facies,
montmorillonite
is common
in Mickey Pass tuff and
Singatse tuff.
Typical gypsum spectra were detected in the
southwestern part of Singatse Range where a sedimentary gypsum
deposit was once mined.
Figure 23 shows the spectral-alteration map of Virginia City
district.
This region is characterized by kaolinite, alunite,
sericite and montmorillonite.
These minerals were detected in
an
order
that
reflects
the
alteration
intensity.
e. g.,
transitional pattern from sericite to alunite.
This result
has proven that
the high resolution spectral
survey is
an
effective means for zoning of the alteration area.
Problems of mixed layer minerals and of mixed pixel, i. e. ,
the ambiguity introduced by the heterogeneous composition of
each pixel are left pending. However, the fact that the low
and high altitude surveys have detected the same mineral zones
leads to the conclusion that the IFOV was not critical as the
spectral resolution in the investigated areas.
8.
CONCLUSIONS
The
airborne
high-resolution
spectral
surveys
over
the
Yerington and Virginia City mining areas have proven to be an
effective means of identifying alteration minerals which are
important for mineral exploration.
The Log Residual spectra
have mapped the surface mineral pattern in fine detail along
the traverses.
In accord with the field spectral survey, the
result
provided
successful
delineation
of
the
alteration
mineralogy.
The
computerized mineral
identification system
that
has
still
been under development will
provide useful
means for future research into remote sensing data of larger
area with a large number of spectral channels.
9.
REFERENCES
Carten,
R.B ..
1986, Sodium--Calcium Metasomatism: Chemical, Temporal.
and
Spatial Relationships at the Yerington, Nevada. POL'phyry Copper Deposit;
Economic Geology. Vol.8l. pp.1495-I519
D ill e s ,
J . H .,
1 983 .
The
I' e t I' 010 g y
and
G e 0 c hem i s try
0 f
the
Ye r i n g ton
bat hoI .i t h a n d
the
Ann ~1 a son
Po r p h y r y
Cop per
De DOS it.
We s t ern
Ne v a d a ;
Stanford Ph.D. Thesis
Elvidge. C.D., Lyon, R.J .P., 1984, Mapping Clay Alteration in the Virginia
Range - Comstock Lode, Nevada with Airborne Thematic Mapper Imagery; Int.
Symp.
on Rem.
Sen.
of Env.,
Third Thematic Conference,
Hem.
Sen.
for
Exploration Geology, p.16J 170
Green, A.A., Craig. M.D .. 1985. Analysis of Aircraft Spectrometer Data with
Logarithmic Residuals;
Proceedings of the /\irboene Imaging Spectrometer
Data Analysis Workshop. pp.111-1I9
Green, I~.O., Lyon, H.J.P .. 1981. Mapping Mineral Alteration with Airborne
Thematic
Mapper
Imagery
in
the
Ann-Mason
Region,
Yerington [jistl'ict.
Nevada; Int. Symp. on Rem. Sen. of Env.,· Third Thematic Conference, Rem.
Sen. for Exploration Geology, pp.775-784
Mar s h ,
S . E.,
~j eKe 0 Jl ,
J. B.,
1983 .
I n t e g rat c dAn a 1 y sis 0 f
II i g h - He sol uti 0 n
Field and Airborne Spectroradiometer Data for Alteration Mapping; Economic
Geology, Vo1.78, DP.Gl8 632
Thompson,
G.A .•
1956,
Geology of the Virginia City (Juadrangle,
Nevada;
U.S.G.S. Bulletin 1042-C
1