Mapping of volcanic series rock units using landsat thematic mapper imagery, troodos...

Mapping of volcanic series rock units using landsat thematic mapper imagery, troodos ophiolite complex, cyprus


K.T.U.S.De Silva
Geological Survey And Mines Bureau
No;4, Galle Road, Dehiwala Sri Lanka

The island of Cyprus hosts one of the best-preserved ophiolite complexes in the world, known as the Troodos Ophiolite Complex. Litho logically, it comprises a displaced slab of altered ultra basic and basic plutonic igneous rocks capped by intermediate and basic lava flows. Its unique dormal structure permits one to study the complete ophiolite stratigraphy without different. Out of the complete ophiolite sequence, the volcanic series rocks units, namely, Upper Pillow Lava (UPL), Lower Pillow lava (LPL) and basal Group (BG) are difficult to map in the field due to their close similarities.

In the research, several advanced image processing techniques were applied and tested their effectiveness to the three sub scene which were selected from a TM image in the North and eastern part of the Troodos massif, comprising the volcanic stratigraphy. Out of these, images produced by Saturation Enhancement, Decor relation Stretching and Selective data ad well as with the field observations. It is important to mention that in the previously published geological maps (maps published up to 1979), most of the UPL unit in the Delikipo area in the estern part of the Troodos massif had been mapped erroneously as LPL rocks. Later in the early eighties, during the map updating programme. This has been corrected. With the use of the above mentioned image processing techniques, this can easily be confirmed.

The selective PCA using three bands is a new method which was created for enhancing geological formations in this study. The method uses only there of the possible six Landsat TM bands for mapping iron (bands1,3&5) and hydroxyl (bands 1,5 & 7 )alteration, readily apparent in TM imagery. Careful examination of the eigen vector loadings allow the identification of the correct Principal Component needed to map these alteration types. Final demarcation of the volcanic stratigraphy was based on the difference in hydroxyl and iron oxide contents in each rock unit.

The Troodos massif of the Island of Cyprus hosts one of the best preserved and most thoroughly studied ophiolite complexes in the world. Troodos has received a special attention among geological community throughout the world due to its complete and non disruptive sequence tougher with a displaced slab of altered ultra basic and basic platonic complex, stratigrephically overlain successively by a sheeted dike complex, extrusive sequence and pelagic sediments (Moore & Vine1971; Gase 1980). Due to subsequent erosion, the original upward succession of ophiolite stratigraphy was arranged in an outward succession from centrally exposed platonic rocks to sheeted dike complex and peripheral pillow lavas.

Out of the complete ophiolite sequence, the volcanic series rock unit, especially the UPL and LPL are very difficult to differentiate in the field due to their close similarities. Complete discrimination of the two lava rock units is necessary as massive sulphide ore bodies that have long been exploited in Troodos massif, occur at their contact zone. In this research, applying remote sensing techniques together with field verfication studies carried out a study of the pillow lava series rock units.

Geology of Troodos Pillow Lava Series
The volcanic sequence, also known as the pillow lava series, is extending as an irregular incomplete ring around the periphery of the massif in a belt of metamorphosed pillow lavas and dikes, that from the uppermost 1.5-2 km of the ophiolite complex. On the basis of colour, mineralogy and abundance of dikes, the volcanic sequence is divided into three namely, Upper Pillow Lavas (UPL), Lower Pillow Lavas (LPL) and Basal Group (BG) (Wilson, 1959).

Upper Pillow Lavas (UPL)
Rocks of the UPL occupy a relatively narrow strip in the outer periphery of the ophiolite sequence, having hummocky topography. Overstepped in places by the sedimentary rocks, the outcrops of this division averages nearly one kilometer in width in the western part and gradually widens toward the east. Pathologically the UPL succession consist of prophyritic olivine basalts, limburgite and olivine free basalts.

Throughout the study area, UPL are well exposed as no vegetation cover is noted on these rocks, except for occasional pine trees and thorny bushes. However, although the pillowed nature of the lavas can be seen easily on majority of surfaces, fresh specimens are hard to obtain as the rocks have been extensively weathered, decomposed and shattered into small fragments when struck with a hammer.

The UPL are characterized by well formed in abundance and intrusive such as dikes and sills are almost absent. The pillows vary in size and shape. The size of the pillow range from a few centimeters across by around 20 cm in diameter to 50 centimeters across by nearly 2 meters in diameter. Their shapes vary from spheroidal to ovoidal, the latter being the prominent.

Although, the lavas themselves are pale gary in colour, their inter pillow material gives a pink cast to their outcrops, in many places: The inter pillow spaces are occupied by iron stained calcite, altered glass, jasper and zeolites and their pinkish colour is due to sea floor oxidation. Some of its pillows are as cracked and the fractures are filled with pinkish colour inter pillow material and are seen as veins. In some areas, the pillow exhibit a dark appearance which resembles the Lower Pillow Lavas.

The UPL are resting on an undulating surface of LPL and they are immediately overlaid by the umbers. However, when the umbers are absent, they are sandwiched between underlying LPL and overlying rocks of Lefkara Formation.

Lower Pillow Lavas (LPL)
The lower pillow lavas are characterized by the presence of both extrusive and intrusive rocks. Extrusive rocks mainly consist of pillow lavas, non-pillowed lava flows and hyaloclastites whereas intrusive rocks which may constitute upto 50% of the outcrop, include dikes, sills and irregular intrusive masses. This lower lava unit composes of aphyric rocks with rare microphenocrysts of plagioclase, clinopyroxene, orthopyroxene and iron-titanium oxides. The rocks of this divesion consist of andesine, andesitic dacites, dacites and rhyodacites.

Similar to UPL, the rocks belonging to LPL are much decomposed at th esurface. They exhibit a greenish gary tint and the soils are more mature. However, fresh to slightly weathered outcrops are exposed in soma parts of the major river sections (e.g.: – Akalki river section). Due to the differences in composition structure and degree of chilling, an uneven but not rugged relief is observed in the LPL rocks. Numerous narrow ridges and deep gulleys, corresponding to the effects of erosion on the harder intrusives and softer pillow lava, characterize it

Individual pillow vary in size from few centimeters to nearly 40 cm across by approximately 50 cm to 3 meter in diameter. However, the diameter of many pillow fall between 1 and 2 meter range with the average width of 0.5 m. as in UPL, their shape is varied from spheroidal to oval. Most of the pillow are strongly vesicular, some vesicles are empty whereas the others are generally infilled with zeolites, calcites, quartz and green chlorite.

The most characteristic mineral identified in the lower lava series is the green colored celadonite. Celadonite occurs as stains and open space filling in lavas and in some of the intrusions. The LL unit transitionally overlies the basal group, which consists of over 60% dikes with screens of pillow lava and underlies the UPL in which intrusive are almost absent.

Basal Group (BG)
The transition zone from the extrusive series to the sheeted dike complex is known as the Basal Group. The lower part of the BG merges into the diabase and the contact between the two is arbitrarily placed where pillow lavas are absent (Wilson, 1959). The BG consists of over 60% dikes (usually upto 90%) with screens of pillow lava. The presences of the lavas are screens results in the separation of the group from the diabase in which pillow lavas are absent. Although the whole basal Group unit is marked by a characteristic high relief, fresh exposures are hard to be seen at the surface which is mostly covered by the brownish colored weathered products.

The pillow lavas of the Basal Group have heavily intruded by dikes and the host rock recognizable only as narrow intervening screens of lavas which form a monor proportion of the unit. The fresh pillow lavas are greenish-gary in colour, possessing considerable amount of small vesicles. Vesicles are mainly filled with quartz, epidote and calcite. Generally, the harder intrusives are grayish brown in colour and are similar in character to the diabase dikes of the Sheted Dike Complex. In intrusives, vesicles are not as common as in the host pillow lava unit.

The boundary of the BG with overlying lava units is faulted in many places. A well-marked change in topography, which reflects the difference in hardness of the two units, indicates the approximate boundary between the BG and the peripheral pillow lavas. In addition to the relief, increasing number of pillow lava screens and the appearance of celadonite are excellent indicators to mark the proximity to the LPL unit. The diabase intrusive and the associated pillow lava screens of the BG form a high, rugged ground, which is oftenly, covered with pine forest, whilst the pillow lava has a low relief. The nature is responsible for the occurrence of some of the BG inliers in the study area. However, majority of BG inliers observed in the area appear to be formed due to subsequent faulting, where highly brecciaed rocks are located along the contact zone of the two units.

Digital Image processing
The main objective of applying image processing techniques in this study is to differentiate pillow lava series rock units. Since the entire pillow lava rock series is too large for a detailed study of this nature, there sub above mentioned sub scenes, a full thematic mapper scene (path 176, row 36), acquired on 3rd August 1994 was used. This image is ideal for applying image processing techniques, as it does not posses any cloud cover and also the Troodos massif is well centered. Digital image processing techniques were applied to these three sub scenes using the Integrated Land and Water Information System (ILWIS) software version 1.4, developed by ITC in 1993

Form the statistical methods, it was found that band combinations532 and 731 are the best suited sets for lithgolical interpretations in the area. Several image processing techniques were applied to Landsat Tm sub scenes using the selected band sets, for discrimination of volcanic series rock units. Out of these techniques, Saturation Enhancement, Decorrelation Stretching and Selective PCA using 3 bands are the best suited methods for litho logical discrimination.

Saturation Enhancement
By applying Saturation Enhancement technique, the visual interpretability of the sim[le flase colour composites (SFCC) can be improved. Instead of the RGB space, MMI space is applied in this technique. Although, saturation and hue can be used to describe every point within the colour triangle, a simple mathematical method which uses two orthogonal axes (M1 &M2) is usein this method. The origin of the two orthogonal axea is placed at point where Intensity axis crosses the triangle plane. M1 is arbitrarily chosen in line between origin and position of red in the triangle while M2 is orthogonal to M1, thus parallel to the line from green to blue. In this way, the data space is reduced to two dimensions and the maximum variation is limited to triangle. Within the two dimensional m1m2 space, the data cluster can be transformed in various ways (shifting, rotating and scaling)in order to spread it cover the complete triangle. In the resultant image colour is balanced and thus the image is easy to interpret.

Decorrelation Stretching
Decorrelation Stretching is a from of multi-image manipulation that is particularly useful when displaying multispectral data that are highly correlated. Traditional contrast stretching of highly correlated data, as R,G and B displays, only expands the range of intensities. It does little to expand the image of colour displayed and the stretched image still contains only pastel (saturated) huies. For example, no areas in a highly correlated image are likely to have dogotal numbers in the red display channel but low values in the green and blue (which would produce a pure red). Instead, the areas are merely a reddish-gray. To circumvent this problem, decorrelation Stretching involves exaggeration of the least correlated information in an image primarily in terms of saturation, with minimal change in image intensity and hue.

As with IHS transformation, decorrelation Stretching I applied in a transformed image space and the results are then transformed back to the RGB system for final display. The major difference in decorrelation Stretching is that the original image’s principal components are used in the transformed image space. The successive principal components of the original image are stretched independently along the respective principal component axes. By definition is to emphasize the poorly correlated components of the original data. When the stretched data are transformed back to the RGB System, increased colour saturation is observed pastel enables a direct interpretation image in which previously observed pastel hues much saturated.

As decorrelation Stretching is based on principal components analysis, it is readily extended to any number of image channels. This is the main advantage of Decorrelation Stretching over the saturation Enhancement procedure (IHStrasformation), which is, applies only to 3 channels at a time.

Selective PCA using three bands
This in a new method, which was created for enhancing geological formations in this study. The method is based on the modification of Laughlin’s (1993) technique. The method uses only three of the possible six Landsat TM bands for mapping iron and hydroxyl alternation zones, readily apparent in TM imagery. Hydroxyl alteration is highlighted by the presence of limonitic bearing phyllosilicates which cases strong absorption in TM band 7 and limonitic iron oxide alteration, which cases absorption in bands 1 and higher reflectance in band 3. In this method, band 4 was omitted from the input band combination which helps to mapping of vegetation from the resultant images.

Mapping of Hydroxyl Minerals
In order to map hydroxyl, input bands are restricted to TM 1,TM5 and TM7. The eigen vector matrix generated by selecting the above mentioned band combination to map hydroxyl minerals in one the 3 selected sub scenes are shown in table 1. By carefully examining the eigen vector loadings, the PC involved in mapping hydroxyls can be identified. The resultant eigen vector loadings for other sum scenes are similar to the values of table 1

Table 1: eigen vector of the covariance matrix of 3 band combination to map hydroxyl minerals in sub scene1.

PC2 TM1 TM5 TM7 Var%
PC1 0.394 0.807 0.441 88.13
PC2 0.899 -0.439 0 10.55
PC3 -0.193 -0.396 0.898 1.32

Eigenvector values of the table show that both PC2 an dPC3 map the hydroxyl minerals. However in both PC2 and PC3, the hydroxyl are highlighted as dark pixels. Therefore both images have to be negated to map hydroxyl as bright pixels. Then they have to be stretched to get similar means and standard deviations, before they are subjected to pariwise PCA. One of the resultant PC images from this transformation will have two positive eigen vector loadings (PC2 of the Table 2) and that will be the final hydroxyl image.

Table2: Eigen vectors of the covariance matrix of pariwise PCA using negated PC2 and PC3 images.

PCs Negated PC2 Negated PC3 Var %
PC1 -0.224 0.975 57.04
PC2 0.975 0.224 42.96

Mapping of Iron Oxide Minerals: –
To map iron oxide minerals, the band combination was limited to TM1, TM2 and TM5. The resultant eigen vector matrix in sub scene 1 are shown in table 3. The other 3 sub scenes also shoe eigen vector value to the value display in Table 3

Table 3: Eigen vectors of the covariance matrix of 3 band combination to map iron oxide minerals in sub scene 1

PCs TM1 TM3 TM5 Var%
PC1 0.414 0.516 0.750 85.58
PC2 -0.597 -0.468 0.652 13.23
PC3 -0.687 0.718 -0.144 1.19

By studying the above eigen vector value. It is clear that only PC3 maps the iron oxides. The stretched PC3 is the final iron oxide image.

Final colour composite was made by using the following band combination and the resultant is good enough for demarcating geological formations.

Red – Hydroxyl image
Green – Iron oxide image
Blue – Stretched TM5 image
The images produced by all the above three techniques are fairly matched with the available published geological data as well as with the field observations.

It is important to mention that in the previously published geological maps (maps published upto 1979), most of the UPL unit in the Delikipo area in the eastern part Troodos massif had been mapped erroneously as LPL rocks. Later in the early eighties, during the map-updating programme this has been corrected. With the use of remote sensing techniques, this can easily be confirmed.

Out of the remote sensing techniques applied to the sub scenes selected from the TM images, saturation Enhancement, Decorrelation Stretching and Selective PCA using three bands can be considered as the best suited methods for the demarcation of volcanic series rock units. The resultant images obtained from the above techniques have reasonably matched with the available geological data as well as with the field observations. In some parts, improvements to the already published geological boundaries and units can be made by interpreting the resultant images together with field checkings.

The material presented in this paper the writers MSc thesis presented at the International Institute of Aerospace Survey and Earth Sciences (ITC), The Netherlands. The study was supported by the Netherlands Fellowship Programme (NFP). The write is indebted to Drs. P.M. Vandijk and Vander Meer for their valuable guidance and advices during the period of study. The author also wishes to thank Miss Uddhika Perera for providing assistance in preparing the manuscript.


  • Clentuar, M.R>,Knox, G.J., Ealey, P.J., 1997 the geology of Cyprus and its place in the East- Mediterranian frame work; Gelogie & Mijnbouw, v.56(1),p.66-82.
  • Gass, IG 1987,Ophiolites and Oceanic Lithosphere, in : Malpas, J., Moores, E.M., Panaayiotus, a., and Xenophontos, C.,(Editors), Ophiolites and Oeanic Crust analogues, Proc. Int. Symposium on Troodos 87, Cyprus geo. Surv. Dept., p.1-10.
  • Gillespie, a.r., kahle, a.b., and walker, r.e.,1986, colour enhancement of highly correlated images. 1. Decorrelation and HSI contrast stretches; Remote sensing of environment, v.20, and p.209-235.
  • Loughlin, w.p., 1991, principal component analysis for alteration mapping: proceedings of eight thematic conference on geologic remote sensing, Denver, Colorado, USA, April 29 May 02v.2,p.293-306.
  • Panayiotou, A.,1987, Editor Field excursion guid book; International symposium on Troodos ’87, Cyprus Geological Survey Department publication, p. 1-285.
  • Rothey, D.A.,1987, Improved discrimintion: of rock units using landsat Thematic Mapper imagery of the Oman Ophiolite: Journal of the Geological Socity of London, v., 144,p. 587-597
  • Tores,M.V.,1996, Regional geological interpretation and spectral analysis of Troodos massif ophiolits (Cyprus), Unpublished M.Sc. thesis, ITC, p,1-112.
  • Van der Meer, F., 1996, Geologic prospecting for mtneral exploration in the Troodos ophiolite massif of Cyprus, using Landsat TM data: Int. Archives of Photogrammetry and Remote Sensing, v.xxxi, part B7, p.716-721.
  • Van der Meer. F., Torres, M.V., and Vandijk,P.M.,1997,Special characterization of ophiolite lithologies in the Troodos ophiolite complex of Cyprus and its potential in prospecting for massive sulphide deposits; International Journal of Rempote Sensing, v.18, no.6,p.1245-1257.