models for
lournal of
5. Wildlife
ation. |n,
Systems,
Butterfield,
, D'Erchia,
ight, R. G.
proach to
Wildlife
a deer on
nmalogical
anese with
a deer on
forest on
Ecological
. A mass-
in Island,
), pp. 215
Il, M. J. &
lysis using
GIS with
ce of old
595-600.
PRINCIPAL COMPONENT ANALYSES FOR LITHOLOGIC AND ALTERATION MAPPINGS:
Examples From The Red Sea Hills, Sudan
Nasir Hasen Kenea and Harald Haenisch
Freie Universitat Berlin, Institute of Geology, Geophysics and Geoinformatics
Malteser str. 74-100, 12249 Berlin, Germany
Commission VII, Working Group 4
KEY WORDS: Mapping, Processing, Imagery, Landsat, GIS, Principal Component Analyses, Signal-to-Noise Ratio
ABSTRACT
Prinicipal Component Analyses (PCA) of Landsat-TM data covering parts of the Red Sea Hills, Sudan were conducted.
Improved lithologic discrimination has been achieved, comapred to commonly used band composites like 7 4 1 and principal
component images obtained from covariance matrix, through standardized transformation. The obtained image enhancement
is corroborated by improved signal-to-noise ratio. Using feature-oriented PCA intended to enhance iron-rich and hydroxylated
minerals, followed by low-pass filtering, it has been possible to successfully map alteration bodies related to sulphide and gold
mineralisations. By computing band ratios and applying a GIS matrix-overlay technique substantial improvement has been
achieved in mapping the alteration bodies enhanced through both processings. The latter appear less useful for lithologic
mapping due to lack of morophological features and incorporation of noise.
1. INTRODUCTION
The Red Sea Hills (RSH) of the Sudan is an arid to semi-arid
and rugged terrane with little or no vegetation cover.
Geologically the region makes part of the Nubian Shield and
is mainly covered by basement rocks that comprise of the
older gneissic complexes, calc-alkaline metavolcano-
sedimentary sequence of arc assembleges, migmatites and
calc-alkaline granitoids, and mafic-ultramafic complexes
related to ophiolitic suites (Vail, 1985). The tectono-thermal
events that prevailed in the region are thought to be
dominated by accretion tectonics of the Pan-African orogeny
that gave rise to the development of prominant shear zones,
complex fold patterns and related collisional structures.
These geodynamical processes are thought to have also
dictated the associated mineralizations (eg. Wipfler, 1994).
In addition to the basement rocks there are isolated
occurrances of younger sediments and Cenozoic volcanic
rocks in a few localities over the RSH (Hassan, 1991).
Study area
— Road
77 Railway \
7- River s
m ur ETHIOPIA
ed x S
F12900'N
\
30000" E 38»oQ'
Figure 1. Location of the study areas; "D" for lithologic
and "A" for alteration mappings.
Lake Tapa
& _,/ 12°00N7
M
34200°
271
AS part of an on-going research on Landsat-data applications
for geological studies, TM-imageries covering parts of the
RSH (fig. 1) have been processed and interpreted. In this
paper we present examples for which improved lithologic
discrimination has been obtained and detection of alteration
Zones made possible using PC transformations.
2. LITHOLOGIC MAPPING
Part of a TM scene, path/row 171/48, captured on
06/10/1985 covering an area with diverse lithologies has
been selected (area D, fig. 1) and geometrically corrected.
Fast atmospheric correction was also applied following the
method given by Chavez (1975). Commonly used band
composites like 7 4 1 in R G B, respectively (Crippen, 1989)
gave very impressive color, however, distinction among
similar looking rocks such as rhyolite and dacite,
metavolcano-sedimentary rocks with and without amphibole,
graphitic schist and migmaitzed graphitic gneiss etc. appear
difficult. PC transformation of the six reflective bands were
conducted for the area of interest using covariance matrix
(table 1), and individually analysed. As expected PC1, with
94.8396 variance and positive loadings from all the TM
bands, contains significant albedo and tpographic
information. PCs 2, 3, and 4 mapped both morphological
and spectral information pertaining to mineral composition in
decreasing order, wheras PCs 5 & 6 are more of noise.
Composites 1 2 3, and 1 2 4, in R G B, respectively, have
been found to provide a better contrast. These composites,
however, display no noticeable improvement over the row
data composite in context of lithologic discrimination.
Standardized PC transformation (Singh & Harrison, 1985)
was conducted for the study area using correlation coefficent
(table 2) of the same input bands. Here the input bands are
more or less uniformly weighted in the first PC, although the
least correlated bands 1, 4 and 7 contribute relatively low.
Bands 5 & 7 of the TM, that are dominant in the 1st PC in
case of the unstandardized transformation (table 1) because
of their larger variance, appear to contribute the highest in the
2nd PC for the standardized transformation. Examination of
the variance distribution shows that in the latter case
significant increament has been achieved for components 2,
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
