PFG 2012 / 6, 0711 – 0726 Article Stuttgart, December 2012

Lithological Mapping of Dahab Basin, South Sinai, , using ASTER Data

Adel OmrAn, Tübingen, michAel hAhn, Stuttgart,VOlker hOchschild, Tübingen, Ahmed el-rAyes & mOhAmed Geriesh, Ismailia, Egypt

Keywords: lithological mapping, band ratio stacking, classiication, ASTER, Dahab basin, Sinai, accuracy assessment

Summary: The Advanced Spaceborne Thermal Zusammenfassung: Lithologische Kartierung des Emission and Relection Radiometer (ASTER) has Dahab Beckens im Süd-Sinai (Ägypten) mit AS- gained importance for lithological mapping over TER-Daten. Mit dem Sensorsystem ASTER (Ad- the last decade. Its suitability for creating a litho- vanced Spaceborne Thermal Emission and Relec- logical map of the Dahab basin in south eastern Si- tion Radiometer) werden seit Februar 2000 Daten nai in Egypt is studied and presented in this paper. aufgezeichnet, die einen Beitrag zur detaillierten For classiication and discrimination of different lithologischen Kartierung leisten können. In die- rock types spectral features, in particular band ra- sem Artikel wird die Eignung von ASTER für die tios, are used. A new band ratio stacking with a Herstellung einer lithologischen Karte des Dahab false colour composite created by the ratio stack Beckens im Südosten des Sinai in Ägypten unter- image (8/5, 4/8, 11/14) is proposed for differentia- sucht. Zur Klassiizierung und Diskriminierung tion between younger granitoids and older grani- verschiedener Gesteinsarten werden aus den 14 toids in the southern and central part of the study ASTER-Bändern ausgewählte spektrale Merkmale area. Band ratios 7/6 and 6/4 have turned out to be (Indexbilder) verwendet, wofür einzelne Bänder very suitable for discriminating between cambrian zueinander ins Verhältnis gesetzt werden. Die In- rocks and upper cretaceous rocks in the northern dexbilder der Bandverhältnisse 6/7 und 4/6 haben part of study area. Field investigations at different sich als sehr geeignet für die Unterscheidung zwi- locations of the study area have been carried out to schen kambrischen Felsen und oberen Kreidefelsen aid in the interpretation and analysis of the ratio im nördlichen Teil des Untersuchungsgebietes er- stack images. Together with available geological wiesen. Mit den zu einem Falschfarbenbild zusam- maps the ground truth data is taken into account for mengesetzten Bandverhältnissen (8/5, 4/8, 11/14) selecting training areas and for creating a new map wird eine bislang zur lithologischen Kartierung using maximum likelihood classiication. An accu- nicht verwendete Kombination vorgeschlagen, die racy assessment of the classiication result with re- sich zur Differenzierung zwischen den jüngeren spect to the regional geological map created by und älteren Granitoiden im südlichen und zentralen EGSMA (1994) and the local area maps of Hassen Teil des Untersuchungsgebietes besonders eignet. et al. (2007) and el Masry et al. (2003) indicates Feldbegehungen des Untersuchungsgebietes dien- overall accuracies between 83% and 94%. An ten sowohl dem besseren Verständnis der zu inter- achievement of this study is a lithological map pretierenden Bilder als auch der stichprobenhaften which extends the EGSMA map by adding some Erfassung von Ground-Truth-Daten für die Klassi- rock units such as ring-dykes at Wadi Ferani and izierung. In einem für die im Dahab Becken vor- metasediment, acidic metavolcanics and basic kommenden Gesteinsarten optimierten Prozess metavolcanics in Wadi Saal and Wadi Ramthy. wird eine neue lithologische Karte durch Maxi- mum-Likelihood-Klassiizierung derIndexbilder und eine entsprechende Nachbearbeitung erstellt. Vorhandene geologische Karten (EGSMA 1994, Hassen et al. 2007, el Masry et al. 2003) wurden für die Auswahl von Trainingsgebieten, insbeson- dere auch zur Validierung der erstellten Karte her- angezogen. Die Beurteilung der neuen Karte er- folgte anhand von Konfusionsmatrizen, die auf

© 2012 E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart, Germanywww.schweizerbart.de DOI: 10.1127/1432-8364/2012/01511432-8364/12/0151$4.00 712 Photogrammetrie • Fernerkundung • Geoinformation 6/2012 eine Übereinstimmung mit den vorhandenen Kar- EGSMA-Karte um Ring-Dykes im Wadi Ferani ten zwischen 83% und 94% hindeuten. Die neue und um Metasedimente sowie um saure und basi- Karte erweitert zudem die amtliche geologische sche Metavulkaniteinden WadisSaalund Ramthy.

1Introduction wavelength region permits to distinguish sili- cates and carbonates”. Although Remote Sensing techniques have The study area (Dahab basin, Sinai) is lo- opened new ways for mapping lithology over cated within an arid climatic belt that crosses the past three decades, the geological maps northern Africa. Rocks dominate the appear- of the were created based on ance of the landscape which is sparsely cov- conventional ground surveys with suitable ered by desert vegetation. The goal of this ield observations. A standard procedure was study is to work out details of a classiication to probe along traverse lines at regular inter- approach for improving the existing geologic vals and to plot this point information on a map in the Dahab basin. It further intends to topographic map. The inal geological maps point a way forward for updating the geolog- then have been created by interpolating the ical maps of the southern Sinai Peninsula as point information and cartographic post pro- well as for the geologically similar regions in cessing. By interpolating the sparse point data Egypt’s eastern desert and in the western part in a mapping area certain errors are unavoid- of . A second reason for generat- able and lead to inaccuracies in the map. With ing the geological map in this study is to pro- remote sensing data the mapping procedures vide an updated basis for hydrogeological in- have undergone signiicant changes. The vestigations of the Dahab basin. availability of high resolution multispectral and hyperspectral data has further increased Related work the potential of remote sensing in delineating the lithological contacts and geological struc- ASTER data have been successfully used in tures into greater detail and with better accu- geological mapping since early 2000. In com- racy (drury 1987). parison to Landsat TM data, ASTER data has A main purpose of the Advanced Space- the advantage of combining wide spectral borne Thermal Emission and Relection Ra- coverage and high spatial resolution in the vis- diometer (ASTER) mission is to extend the ible and infrared regions which makes it at- understanding of local and regional phenom- tractive for geological mapping (e.g. Hewson ena on the earth surface and its atmosphere. et al. 2001, Bedell 2001). yaMagucHi & naito Goals of geologic research using ASTER are (2003) studied spectral index images for litho- summarized by goMez et al. (2005). They put logical mapping. Index images are found by the focus on the “study the geologic phenom- a linear transformation of relectance values ena of tectonic surfaces and geologic histo- of the ive ASTER SWIR bands. The idea of ry through detailed mapping of the Earth to- this transformation is to direct the transforma- pography and geological formation”. For the tion axes to the spectral pattern of the target discussion of the importance of the different minerals. The calculated indices are named ASTER wave bands the before mentioned au- according to the minerals: alunite, kaolinite, thors refer to early work of Knipling (1970), calcite and montmorillonite. An advantage of Hunt (1977, 1979), and salisBury et al. (1987): this approach is that the transformation coef- “The visible and near infrared (VNIR) wave- icients are not scene dependent. A simulated length region provides some information for ASTER dataset was used to prove the useful- the presence of vegetation, iron oxides (hema- ness of the spectral index images. tite, goethite, jarosite) and rare earth elements. Some studies used band ratios and the spec- The shortwave infrared (SWIR) wavelength tral unmixing technique applied to SWIR, region assesses bearing minerals (clays, phyl- TIR and emissivity data of the TIR bands losilicates, …) and the thermal infrared (TIR) to improve the mapping of the sedimenta- Adel Omran et al., Lithological Mapping of Dahab Basin 713 ry, metasedimentary and volcanic areas. For age spectra of various lithological groups into example, cudaHy & Hewson (2002) distin- classiication produces. Hewson et al. (2001) guished some minerals in epithermal, porphy- studied a regolith and alteration area in Aus- ry and skarn groups by using the band ratio tralia and showed how to improve an existing technique. aBdeen et al. (2001) used ASTER 1:100,000 geological map with ASTER data. band ratios (4/7, 4/1, 2/3*4/3) and (4/7, 3/4, 2/1) For their investigations they relied on previ- for mapping ophiolites, metasediment, vol- ous surveys using airborne HyMap recordings caniclastics, and granitoid lithologic units of of visible, shortwave and thermal IR bands the Allaqi suture in the southeastern part of and spectral measurements collected in ield eastern desert of Egypt. campaigns. They concluded that their exper- gad & KusKy (2007) recommended to use imental results indicate “that ASTER could band ratios (4/7, 4/6, 4/10) for mapping the discriminate mineral groups not achievable granite and metamorphic belt of the Wadi Kid from Landsat TM, though more precise min- area of Sinai and for mapping metamorphic eral species mapping is not possible”. rocks in the Arabian Nubian Shield and other In the next section background information arid areas. reda et al. (2010) used band ratios about the location of the study area and the (2+4)/3, (5+7)/6, (7+9)/8 to discriminate be- available ASTER data is given. The geology tween different ophiolitic and granitic rocks in of the study area is outlined in section 3. In the the central eastern desert of Egypt. Madani & following two sections 4 and 5 the approach eMaM (2009) investigated the band ratio com- pursued in this research for creating a litho- posite (8/5, 5/4, 7/8) to differentiate between logical map of the Dahab basin is presented alkali granites (younger granitoids), granodi- and processing of the ASTER data using band orites and quartz diorite (old granitoids) in the ratios andsupervisedclassiication is dis- Wadi el-Hudi area, which is located in south- cussed in the context of the achieved results. eastern desert of Egypt. Qari et al. (2008) uti- Section 6 summarizes an accuracy assessment lized the (6/8, 4/8, and 11/14) band ratio image of the generated map and presents conclusions to discriminate thebasementrocks in theAra- of the achievements. fat area of the Western Arabian Shield, Sau- di Arabia, and created a 1:100,000 geological map. Inspired by the work of Madani & eMaM 2Location of the Study Area and (2009) as well as Qari et al. (2008) a new band ASTER Data ratio combination (8/5, 4/8, 11/14) is proposed in this study. Details for this new band ratio The Dahab basin is considered as one of the stacking are given in sections 4 and 5. major hydrographic basins along the Gulf of waldHoff et al. (2008) used Hyperion data Aqaba. The basin is located in the southeast- and hyperspectral analysis techniques like ern part of the Sinai Peninsula and is bounded the minimum noise fraction transformation by Latitude 28° 22’ 43.4” and 28° 52’ 18.5” (MNF), the pixel purity index technique (PPI) N and Longitude 33° 55’ 46.9” and 34° 31’ and the spectral angle mapper classiication 28.8” E. It occupies an area of about 2080 km2 (SAM) for geological mapping and compared (Fig.1,left). It is bordered by the Gulf of Aqa- it with ASTER data classiication results. ba to the east, Gebel Gidid, Gebel Sheikh El- They concluded that the ASTER data allowed Arab and Gebel Ferani to the south, Gebel Um a more detailed classiication of the surface Alawi, Gebel Um Loz and Gebel Hamami to composition of the study area. This is particu- the west, and Gebel Bradi and Gebel Gunah to larly noteworthy because Hyperion excels by the north (Fig.1,right). It includes some main its 220 contiguous spectral bands which cover wadis in South Sinai as Wadi Nasab, Wadi the spectral range from 400nm to 2.5 µm at Ramthy, Wadi Saal, Wadi Zaghraa, and Wadi a ground resolution of 30 m. rowan & Mars El Ghaieb. (2003) used in situ measurements of spectral ASTER data are offered at various process- relectance curves for calibrating the VNIR ing levels. Level-1A data are reconstructed, bands of ASTER. Lithological mapping was unprocessed instrument data at full resolution carried out by selecting and introducing im- which consist of the image data, radiometric 714 Photogrammetrie • Fernerkundung • Geoinformation 6/2012

Fig. 1: Left: Location map of Wadi Dahab; right: Basin catchment area. and geometric coeficients and other auxil- types which can be subdivided according to iary data without applying the coeficients to their age into basement rocks and phanerozoic the image data to maintain the original data rocks. In this section the published knowledge values. The Level-1B data are generated ap- on the geology of the basin is summarized plying these coeficients for radiometric cali- with respect to the occurring rock types and bration and geometric resampling. The scene its forming minerals. used in this research is AST3A1 – 15 bands – The relectance spectra of minerals are well 2006 which is a Level-3A data product. This known and catalogued, e.g. in the USGS Digi- so-called Terrain Correction Image includes tal Spectral Library (clarK et al. 2007). The Level-1B image data projected in UTM zone fact that rocks area complex mixture of mate- 36 (WGS 84) which are orthorectiied using rials limits the direct utilization of those spec- a DEM. The scene covers an area of 60 by tra for remote sensing. The use of the spectra is 60 km2 which apart from a small area in south- further lowered by the fairly broad bandwidth east encloses the Dahab basin completely. and the small number of spectral bands of AS- The proposed procedure for lithological TER. The challenge for the remote sensing mapping of the study area using ASTER data approach is to analyse the relectance of the (sections 4 and 5) mainly relies on the AS- mineral mix recorded by the ASTER bands. TER bands 4, 5, 6, 7, 8, 11 and 14. The spec- For example, the spectral characteristics of tral ranges of the 30 m resolution SWIR bands different rocks in the thermal infrared shows are: 1.600–1.700μm(band 4), 2.145–2.185μm a direct dependency on the silica (quartz) con- (band 5), 2.185–2.225μm (band 6), 2.235– tents (KoBayasHi et al. 2010), so that the AS- 2.285μm(band 7), and 2.295–2.365μm(band TER TIR bands will reveal quartz-related in- 8). The TIR bands 11 and 14 record emitted formation about the rocks. How to reveal the radiation in the wavelength ranges 8.475– rock speciic information and how to exploit it 8.825μm and 10.95–11.65μm, respectively. within a classiication scheme for lithological The spatial resolution of the TIR bands is at mapping will be discussed in sections 4 and 5. 90 m ground pixel size.

3.1 Basement rocks 3Geology of the Dahab Basin The EGSMA map (Fig.2) shows that base- The geology of the Dahab basin is discussed ment rocks are the prevalent rocks in the basin by many authors such as HuMe (1906), said area, in particular in the southern and the cen- (1962), soliMan (1986), el sHafei et al. (1992), tral part of the basin. Basement rocks can be Kora & genedi (1995), zalata et al. (1997), differentiated into igneous rocks, which cover el Masry et al. (2003), and Hassen et al. more than 70% of the study area, and meta- (2007). The study area includes many rock morphic rocks which mainly can be found at Adel Omran et al., Lithological Mapping of Dahab Basin 715

Wadi Feirani, Wadi Saal, and some parts of tite, sercite (kaolinite) and opaque minerals. Wadi Zaghraa (Fig.2). The igneous rock com- The older granitoids are mainly composed prises a variety of granitic rocks and young- of plagioclase feldspar, K-feldspar, micro- er gabbros. These rocks are intersected with perthite, hornblende, biotite, and quartz. Zir- acidic and basic dykes. el Masry et al. (2003) con, sphene, apatite and opaque minerals are indicate the presence of ring-dyke at the Gebel among the accessory components. Weathering Laig area with younger granitoid rocks. of hornblende, biotite and plagioclase leads to form clay mineral (el Masry et al. 2003). Granitic rocks Metamorphic rocks Granitic rocks have wide areal extension at Wadi Nasab, Wadi Ramthy, Wadi Dahab and Metamorphic rocks in the study area are dis- Wadi El Ghaieb. Hassen et al. (2004) and el tributed along Wadi Saal, Wadi Ramthy and Masry et al. (2003) divided the granitic rocks Wadi Zaghraa. soliMan (1986) and Hassen in the study area into late kinematic and past et al. (2004, 2007) mapped the metamorphic kinematic rocks which is substantially equiva- rocks at Wadi Saal and Wadi Zaghraa-Ramthy. lent to older granitoid and younger granitoid They differentiated these rocks into metasedi- rocks. The two groups of older and younger ment, basic metavolcanic, acidic metavolcan- granitoids (Tab.1) include the EGSMA dif- ic and metagabbro. The metamorphic belts ferentiation of the granitic rocks into monzo- are intruded by syn (older) and late (younger) granite, alkali granite, granodiorite and quartz granitoids and gabbroic rocks. Metasedimen- diorite. tary rocks consist mainly of phyllite, metasilt- The mineral composition of younger gra- stone, metaconglomerate and volcanogenic nitic rocks is quartz and K-feldspar, plagio- sediment, whereas the metavolcanics include clase, hornblende and biotite, zircon, apa- a wide variety of metamorphosed rock types such as andesite, dacite and rhyolite associat- ed with minor basaltic bodies (Hassen et al. Tab. 1: Granitic rock types in the Dahab basin 2007). area.

Types of Occurring Forming granites rocks minerals 3.2 Phanerozoic Rocks older granodiorite, quartz, plagio- granitoid quartz diorite, clase, hornblende, Phanerozoic rocks cover the northern part of diorite pyroxene study area. Their outcrops expose mainly at the scarp of Gebel El Gounah and the north- younger alkali granite, quartz, K-feldspar, granitoid monzogranite plagioclase, biotite ern Part of Wadi El Ghaieb, as well as mi- nor exposures at Wadi Saal and Wadi Genah (Fig.2). They are represented by a number of geological formations arranged from the oldest to the youngest as: 1) Cambrian rocks consist of laminated sandstone with interca- lations of clay and ferruginous bands. They unconformable overlay the basement rocks, are coarse-to-medium-grained, weakly indu- rated to friable, and include kaolin matrix. 2) Lower cretaceous rocks consisting of grey and violet coloured pebbly and granular sandstone intercalated with kaolin in the upper part and impregnated with iron oxides at the top parts of the sequences. 3) Upper cretaceous rocks Fig. 2: Geological map of the Dahab basin have yellow beds of fossiliferous sandstones, (adopted from EGSMA 1994). dolostones, limestones, marls, and glauconitic 716 Photogrammetrie • Fernerkundung • Geoinformation 6/2012 shales with pelecypod moulds, echinoids, and is separated from the analysis of the southern trace fossils of horizontal burrowings. The part where the basement rocks are forming the top of this formation constitutes a thick car- main rock component. bonate sequence of limestone, marl and dolo- The proposed overall process low for cre- mitic limestone, with thin inter-beds of silty atinga lithological map of the Dahab basin is claystone and yellowish-orange, ine-grained shown in Fig. 3. The basic idea of this process sandstone (Kora & genedi 1995). Upper cre- low is to use the prior knowledge of the exist- taceous rocks are mainly exposed at Wadi El ing EGSMA map to guide classiication of the Ghaib and Gebel El Gounah Scarp. rocks. With this knowledge supervised classi- Mineralogically, the lower cretaceous rocks ication will be speciically applied to the in- are mainly composed of quartz and kaolinite put image data for a certain area. According to with a minor amount of calcite. This helps to the EGSMA map the phanerozoic rocks main- distinguish them from upper cretaceous rocks ly cover the northern part and basement rocks in which the amount of calcite is high. A high dominate in the central and southern part of degree of similarity exists in the mineral com- the Dahab basin. For the primary differentia- position of cambrian rocks and the clastic part tion of phanerozoic rocks and basement rocks of lower cretaceous rocks especially with re- our process follows the proposal of aBdeen et spect to quartz and kaolinite minerals. But al. (2001). Spatial separation between north- the cambrian rocks are more ferriginated than ern part and central part is done by manual lower cretaceous rocks, which allows differ- digitization using ASTER band combination entiation between them. 7-3-1. In each of the two regions band ratio im- ages are used as input for supervised classii- 4Methodology cation of different rock types. For this purpose some band ratios which have been successful- Lithological mapping may be carried out on ly used by other researcher are used. In addi- the computer screen by human interpretation tion, a new ASTER band ratio stacking (8/5, of the images. This is a promising way in par- 4/8, 11/14) is introduced. The training areas ticular if the human operator is very experi- for the maximum likelihood classiication are enced. To increase the degree of automation selected on the basis of existing geologic maps within the mapping process the tools of image together with supporting ield visits. As a part classiication can be employed. The human op- of this study, ield visits at 23 locations of the erator is stilla key factor for the mapping suc- study area have been undertaken. However, cess as he will be involved in selecting prop- the idea to use the ield mapped data as refer- er training areas for supervised classiication ence was rejected because of the small sample by taking advantage of existing maps or ield size. Therefore, the geologic maps are further visits. The digitized training regions are used used as references for evaluating the accuracy to determine statistical parameters for classi- of the classiication result. ication. For the well-known maximum like- Band ratioing has been widely used for lihood classiication these are the mean vec- lithological mapping due to its proven ability tors and covariance matrices for each training to produce distinct grey tones of imaged ma- class. In maximum likelihood classiication, terials in certain ratios. A band ratio is cre- all pixels are evaluated and assigned to the ated by dividing the digital number (DN) of class of highest probability. Maximum likeli- one band by the corresponding DN of another hood classiication of the entire study area ina band for each pixel (drury 1987). The major- one step process has been found to be not op- ity of fractional values are between zero and timal. The mineral compositions the sedimen- two or three. Thus, for visibility reasons the tary rocks with sandstone of cambrian rocks ratios are often rescaled to produce ratio imag- and of lower cretaceous rocks are similar to es with higher contrast. Another well-known the mineral compositions of granite rocks. effect of ratioing is the reduction of the impact Therefore, the analysis of the northern part of of shadow in the ratio images. Which band study area, which includes sedimentary rocks, ratio is particularly suitable for enhancing a Adel Omran et al., Lithological Mapping of Dahab Basin 717 certain rock type or mineral depends on the ence in the areas of Wadi Saal and the Wadis dominance of the mineral in the relected data. Zaghraa and Ramthy. The map of el Masry Spectral signatures give useful hints to decide (2003) is taken as reference in the areas of about the bands used for ratioing. Combina- Wadi Nasab, Wadi El Ghaieb und Wadi Fei- tions of three band ratio images can be visual- rani. The limitations of this accuracy analysis ized as colour composites. Features or miner- are obvious; the error matrix provides infor- als show up in distinct colours in these stacked mation about how well the classiication map ratio images. The question of which band ra- and the existing maps coincide. tios or band ratio stackings enhances the vis- ibility of a particular rock type is analysed and discussed extensively in the next section. 5ASTER Data Analysis, The processing low (Fig.3) points out the Classiication Results and different band ratio images used for maximum likelihood classiication. In the northern part Discussion of the basin the mapped classes are cambrian rocks, upper cretaceous and lower cretaceous The usefulness of Landsat ETM band com- rocks, granodiorite and wadi deposits. In the bination 7-4-2 for geological mapping in arid central and southern part of the basin there are regions and the far-reaching consistency of the metamorphic rocks with metavolcanic, ba- this band combination to ASTER band com- sic metavolcanic,acidicmetavolcanic, meta- bination 7-3-1 are pointed out by aBdeen et al. sediment, phyllite and metagabbro. Wadi de- (2001). Fig. 4 left shows ASTER band combi- posits are also taken into account in particu- nation 7-3-1 in which the metamorphic rocks lar for comparison reasons with the reference appear in greenish and reddish colour, the gra- map. For the igneous rocks as the other major nitic rocks as light yellow to light brown, the group in the central and southern part of the sedimentary rocks as white colour, and the study area the rock type classes granodiorite, wadi deposits as light grey colour. A manual alkali granite, monzogranite are mapped by mapping result of the wadi deposits is shown image classiication. Ring dykes are visually in Fig. 4 right. This vector layer of the wadi recognizable in the image. Supported by ield deposits will be used as overlay in other ig- visits they have been digitized interactively. ures to simplify visual orientation. In classifying this group of igneous rocks the metamorphic Feirani metavolcaniclastic rocks are added for comparison reasons with the ref- 5.1 Central and Southern Part of the erence map. Dahab Basin: Igneous Rocks A post classiication smoothing of the clas- siication results is carried out by majority il- A new band ratio stacking with the ratio im- tering and with the suppression of very small ages (8/5, 4/8, 11/14) is used for differentiation areas. The results of the lithological mapping between younger granitoids and older grani- process are the data found by vectorisation of toids in the central and southern part of the the post processed classiication maps. study area. For a better understanding of this The classiication accuracy assessment is stacking the band ratio images are discussed the last step in the overall processing low in the context of the spectral signature of the (Fig.3). For assessing the derived map quanti- dominant minerals. tatively the error matrix method is used which Band ratio 4/8: Alkali granite appears dark, compares the classiication map against a ref- monzogranite grey, older granitoids show erence. Ideally a representative sample of ield light grey to bright colour (Fig.5A). The light mapped ground truth data are used as refer- colour of older granitoids is due to alteration ence data. Due to the lack of suitable ground products of hornblende and plagioclase into truth data the existing geologic maps are used. chlorite and clay minerals, whereas the pres- The EGSMA map is used as reference in the ence of biotite and K-feldspar minerals in al- northern part of the basin. The more detailed kali granites (el Masry et al. 2003) produces map of Hassen et al. (2007) is used as refer- a dark colour in band ratio 4/8. The dark col- 718 Photogrammetrie • Fernerkundung • Geoinformation 6/2012

Fig. 3: Overall process low for lithological mapping of the Dahab basin area (* is a metamorphic rock, ** is an igneous rock, + are not introduced into ML classiication).

Fig. 4: Left: band combination 7-3-1 of Dahab basin area; right: map of the wadi deposits. Adel Omran et al., Lithological Mapping of Dahab Basin 719 our is a consequence of the lower relectance Band ratio 11/14: Younger granitoids appear in band 4 (Figs. 5 B and C). as dark colour, older granitoids show light Band ratio 8/5: Younger granitoids appear grey to bright colour (Fig.7). The dark colour as light grey and older granitoids as grey col- of younger granitoids may be interpreted by our (Fig.6A). In band ratio image 8/5, Feirani their high ability to relect the sun radiation on metavolcaniclastic rocks show dark grey col- their light coloured surfaces, and hence they our. This is due to the presence of biotite and represent cooler surfaces which appear dark in K-feldspar in addition to the alteration prod- the band image. In contrast, older granitoids ucts of hornblende and plagioclase into clay minerals. Rocks rich in feldspar commonly weather to kaolinite. Fig. 6B shows an absorp- tion feature of kaolinite near band 5 thus the high 8/5 band ratio values indicate younger granitoids. On the other hand, dark grey and grey colours of old granitoids and metamor- phic rocks are due to the absorption proper- ty of chlorite (Fig.5C), which leads to low 8/5 band ratio values.

Fig. 5: A: band ratio 4/8 image; B: spectral signature of biotite; C: spectral signature of chlorite (clArk et al. 2007).

Fig. 6: A: band ratio 8/5 image; B: spectral signature of kaolinite (clArk et al. 2007). 720 Photogrammetrie • Fernerkundung • Geoinformation 6/2012

The colour composite of the single band ra- tios is used for deining the training areas of the four rock classes alkali granites, monzo- granites, Ferani metavolcaniclastic, and old granitoids. For each class several (minimum 4) training areas are selected to get representa- tive samples. In Fig. 10B the maximum like- lihood classiication result of the four rock classes is shown together with the wadi depos- it layer. Alkali granites are colourized in red, monzogranites in pink, Ferani metavolcani- clastic in light green, old granitoids in dark green and wadi deposits in yellow. Fig. 7: Band ratio 11/14 image. For the accuracy investigation of the classi- absorb more sun radiation and get warmer ication result the geologic map of el Masry et than other rock types and hence appear bright- al. (2003) is used as reference. This local area er in the band image. map was generated based on detailed ield In experiments with different band ratio work and covers the Gebel Feirani area. Fig. 2 stackings, we found that the colour compos- shows the location of the Gebel Feirani area ite (8/5, 4/8, 11/14) reveals subtle differences at the downstream part of the Dahab basin. between the younger and older granitoids. In addition to alkali granites, monzogranites, Fig. 8A shows alkali granites that appear in and Feirani metavolcaniclastic, the el Masry red colour, monzogranites and acidic volcan- map includes granodiorite, Qz-diorite, and to- ic which show up in pink colour. The Feira- nalite. The latter two cannot be distinguished ni group appears as light green whereas older from granodiorite by image classiication. granitoid looks green and purple. The visual Therefore, the prevalent granodiorite is intro- comparison of this new stacking with the ratio duced as a class on its own. Ring-dyke is in- stack image (8/5, 5/4, 7/8) used by Medani & teractively mapped because it is visually rec- eMaM (2009) in the El Hudi area of southeast- ognizable by its dyke-like body (Fig.8). It has ern desert in Egypt shows its strength with re- the same composition as alkali granites thus spect to the discrimination of younger grani- belongs to this class in the classiication map. toids. Alkali granite and monzogranite rocks The quantitative comparison between the are much better discriminated in the new results of the classiication map and the refer- stacking. But with respect to older granitoids ence map is summarized by the error matrix (granodiorite and Qz diorite) the Medani & in Tab. 2. Reference data are selected from the eMaM (2009) stacking seems to be a bit more reference map for the raster locations deined favourable. by the classiication map. The procedure for

Fig. 8: A: stacked band ratio image (8/5, 4/8, 11/14); B: ML Classiication using the stacked band ratio image (8/5, 4/8, 11/14). Adel Omran et al., Lithological Mapping of Dahab Basin 721

Tab. 2: Accuracy evaluation of the classiication map for igneous rock types.

Class types from the reference map

User’s e Pixels FeraniGranoAlkaliMonzo Total

th Accuracy map of s on Feirani 393 47 05445 88% ti pe

ca Granodiorite 81 392 83484 81% ty ii

ass Alkali Granite 057751879897% ass Cl cl Monzogranite 113177 413604 68% Total 475457 960439 2330 Producer’s Accuracy 83%86% 81%94% 85% selecting samples in each category follows the reddish brown colour. The acidic metavolcan- stratiied random sampling strategy. For each ic rocks show up by light violet colours, the class in the classiication map 4 to 5 samples metasediment phyllite rocks in light green, the are taken as reference data. younger granitoid rocks have dark blue colour The results indicate an overall accuracy of and the light blue colour refers to old granitoid 85%. Apart from the user’s accuracy of 68% rocks (Figs. 9 A and B). for monzogranite the user’s and producer’s ac- Training areas have been deined in the col- curacies of the other classes are all above 80%. our composite (4/6, 4/7, 4/10) by following the The interfering contacts between monzogran- same procedure described already in section ite and alkali granite lead to a fairly high mis- 5.1. The classiication is carried out for seven classiication of monzogranite with the conse- classes of rock units. Fig. 10 shows the clas- quence of a low user’s accuracy for this class. siication results for Wadi Saal (Fig.10A) and Similar is the situation for the Ferani metavol- Wadi Ramthy (Fig.10B). Metavolcanics is col- canics and the granodiorite but with less sig- ourized in blue, basic metavolcanics in purple, niicant consequences for the user’s accuracy. acidic metavolcanics in light green, metasedi- ments (phyillite) in dark green, younger grani- toid in red, older granitoid in pink and meta- 5.2 Metamorphic Rocks sedimentary rocks (in Wadi Ramthy only) in dark green. Wadi deposits have been also The band ratio stacking (4/6, 4/7, 4/10) shows introduced into classiication and show up in the metavolcanic (metatuffa) rocks in green- yellow in the classiication map. ish yellow and the basic metavolcanic rocks in

Fig. 9: Stacked band ratio (4/6, 4/7, 4/10) images; A: Wadi Saal area; B: Wadi Ramthy area. 722 Photogrammetrie • Fernerkundung • Geoinformation 6/2012

For the accuracy investigation the classii- ication of the igneous rocks (Tab.2). The pro- cation maps are compared to the geologic map ducer’s accuracy is fairly high between 79% created by Hassen et al. (2007) based on ield and 96% for all classes. The user’s accuracy of observations. This geologic map covers the ar- 66% for metasedimentary rocks suffers from eas of the Wadis Saal, Zaghraa and Ramthy. misclassiication with metavolcanics andba- The error matrix found by stratiied random sic metavolcanics rocks. The small outcrops sampling is listed in Tab. 3. of acidic metavolcanics, metasediments, wadi The overall accuracy of 83% conirms the deposits, and older granitoids produce a lot of good matching between the classiication map uncertainty with respect to the other classes and Hassen’s reference map for both wadis. which results in user’s accuracies between The error matrix (Tab.3)has quite some simi- 66% and 76% for these classes. larity to the error matrix found for the classi-

Fig. 10: ML classiication of stacked band ratio (4/7, 4/6, 4/10) images; A: Wadi Saal; B: Wadi Ramthy.

Tab. 3: Accuracy evaluation of the classiication map for metamorphic rock types.

Class types from reference source

User’s Pixels met_volmet_bamet_acmet_sed wd youngold Total Accuracy Metavolcanic 281231 000130692% map

on Basic

ti 49 1810 230023577% metavolcanic ca

ii Acidic 31156013074 76% ass metavolcanic cl Metasediment 811141 10062 66% the

of Wadi 5213281 04070% deposits pes

ty Younger 0040073 07795% ass granitoid Cl Older 002005222976% granitoid Total 346228 65 46 33 82 23 823 Producer’s 81%79% 86%89% 85%89% 96%83% Accuracy Adel Omran et al., Lithological Mapping of Dahab Basin 723

5.3 Northern Part of the Dahab sorption feature of kaolinite near band 6 and Basin: Sedimentary Rocks the high relectance in band 4. Upper creta- ceous rocks show up in grey and granites ap- Band ratio 7/6 can be used to differentiate pear in bright tones in band ratio 6/4 (Fig.12). sedimentary rocks with respect to carbonate A visual comparison of the 7/6 versus the 6/4 minerals, in particular calcite and aragonite band ratioindicates that thediscriminative ef- which are the main components of limestone iciency of 7/6 ratio is higher than that of the (upper cretaceous) rocks. The absorption fea- 6/4 ratio. Therefore, the analysis focused in tures of calcite near band 7 (Fig.11B) together analysing band ratio 7/6. A joint use of both with its high relectance in band 7 produces bands would have been possible but was not the dark appearance of the cretaceous rocks in pursued in this analysis. Fig. 11A. The cambrian rocks appear as bright Supervised classiication including post colour, whereas the granites appear as darker classiication smoothing and vectorization is grey tone. carried out by taking band ratio 7/6 as input Alternatively band ratio 6/4 can be used to image. In the classiication map, upper creta- discriminate between different classes of sed- ceous rocks is colourized in red and lower cre- imentary rocks. Quartz with kaolin (altered taceous rocks in light green. For granites vio- to clay) together with iron oxide are the main let and for cambrian rocks blue is used. Wadi minerals of cambrian rocks which lead to dark deposits appear yellow in the classiication colours in this band ratio because of the ab- map (Fig.13).

Fig.11: A: band ratio 7/6 image of the northern part of Dahab basin area; B: spectral signature of calcite (clArk et al. 2007).

Fig. 12: Band ratio 6/4 image of the northern Fig. 13: Supervised classiication of band ratio part of the Dahab basin. 7/6 image of the northern part of Dahab basin. 724 Photogrammetrie • Fernerkundung • Geoinformation 6/2012

Tab. 4: Accuracy evaluation of the classiication map for phanerozoic rock types.

Class types from reference source Up. Low. User’s Pixels Camb. wd Gran. Total Cret. Cret. Accuracy map

of Cambrian rocks191 00023214 89% s on ti pe Upper Cretaceous 0 97 060103 94% ca ty ii

ass Wadi deposits 00994 2131009 99% ass Cl cl LowerCretaceous 01211313147 89% the Granites 1036 2111 150 74% Total 192109 1031 141150 1623 Producer’s Accuracy 99% 89%96% 93%74% 94%

For the accuracy investigation the geologi- ring dykes. The study shows the suitability cal map provided by EGSMA (1994) is used as of maximum likelihood classiication taking a reference. The overall accuracy is quite high various band ratios and band ratio stackings (94%, Tab. 4). User’s and producer’s accura- into account. Through classiication different cies of 89% to 99% indicate a high agreement types of granitoid rocks (monzogranites, alka- of the classiication map and the reference. li granites, granodiorites), metamorphic rocks The only exception are the granitic rocks with (metasediments and metavolcanics) and phan- a user’s and a producer’s accuracy of 74%. erozoic rocks (cambrian, lower cretaceous, Granitic rocks and alluvial wadi deposits are upper cretaceous rocks and loose wadi depos- in contact with each other which is probably its) have been differentiated. For checking the the main reason for this lower accuracy. quality of the classiication map an accuracy The inal geological map of the Dahab ba- assessment is carried out. For this purpose an sin created according to the proposed process error matrix is determined which compares low (Fig.3) is shown in Fig. 14. It comprises the classiication map with respect to existing 19 classes of phanerozoic, metamorphic and geologic maps. igneous rocks. Apart from the ring dykes all As a part of the development a new band other classes have been created by supervised ratio stacking (8/5R, 4/8G and 11/14B) is pro- image classiication followed by post classii- posed for differentiation of younger granitoids cation smoothing and vectorization of the ras- (monzogranites and alkali granites) and older ter data. granitoids (granodiorite) in the Dahab area. Alkali granite and monzogranite rocks are very well discriminated by the new stacking. 6Conclusion With respect to older granitoids (granodior- ite, Qz diorite and tonalite) Medani & eMaM The investigations on the suitability of AS- (2009) stacking seems to be still more favour- TER data for lithological mapping have led able. For the discrimination of different met- to the development of an overall processing amorphic rock types our procedure follows strategy for mapping granitic rocks, metamor- the proposal of gad & KusKy (2007) by ap- phic rocks and sedimentary rock in Wadi Da- plying the band ratio stacking (4/6R, 4/7G and hab basin. The goal to develop a classiication 4/10B). The phanerozoic rocks in the northern based approach in which only minor interac- part of study are diffentiated using the (7/6) tive digitization is included is fully achieved. band ratio. Interactive digitization was used to separate The accuracy investigations of the classii- the northern and the central and southern part cation results are carried out with respect to of the basin as well as for the mapping of the the geological reference maps published by Adel Omran et al., Lithological Mapping of Dahab Basin 725

Hassen et al. (2004), el Masry et al. (2003) 2007: USGS Digital Spectral Library splib06a. and EGSMA (1994). The calculated error ma- – U.S. Geological Survey, Data Series 231. trices indicate overallaccuracies between 83% cudaHy,t.& Hewson, r., 2002: ASTER Geologi- and 94%. Altogether this underlines that the cal Case Histories: Porphyry-Skarn-Epithermal, created lithological map its reasonably well Iron Oxide Cu-Au and Broken Hill Pb-Zn-Ag. – Communication in the Workshop “Mapping the to the existing maps which have been mainly Earth with ASTER”, London. created by ield work. The created lithological drury, s.a., 1987: Remote sensing of geologic map adds some rock units to the general geo- structures in temperate agriculture terrains. – logical map of EGSMA. Added are ring-dykes Geological Magazine 123: 113–121. at Wadi El Ghaieb, acidic metavolcanics, ba- EGSMA, 1994: Egyptian Geological Survey and sic metavolcanics and metasediments at the Mining Authority. – Geologic map of Sinai, Wadis Saal, Zaghraa and Ramthy and meta- Arab Republic of Egypt. Sheet No. 1, scale sediments at Wadi Ramthy. Alluvial wadi de- 1:250,000. posits are also included in the inal lithological el Masry, n.n., Hassen, i.s. & Hegazi, a.M., map as shown in Fig. 14. 2003: A Newly recognized example of a Late Precambrian subsurface cauldron subsidence in- trusion in Southern Sinai: Jabal Laiq ring dyke. – 5th International Conference on the Geology References of the Middle East: 549–558. el sHafei, M.& KusKy, t.M., 2003: Structural and aBdeen,M.M., allison,t.K., aBdelsalaM,M.g.& tectonic evolution of the Neoproterozoic Feiran– stern,r.J., 2001: Application of ASTER band- Solaf metamorphic belt, Sinai Peninsula: impli- ratio images for geological mapping in arid re- cation for the closure of Mozambique Ocean. – gions; the Neoproterozoic Allaqi Suture, Egypt. Precambrian Research 123: 269–293. – Geological Society of America 3 (3): 289. el sHafei, M.K., KHawasiK, s.M. & el gHawaBy, Bedell, r.l., 2001: Geological mapping with AS- M.a., 1992: Deformational styles in the tec- TER satellite: new global satellite data that is a tonites of Wadi Sa`al area, South Sinai. – 3rd signiicant leap in remote sensing geologic and Conference Geological Sinai Development: 1–8, alteration mapping. – Special Publication Geo- Ismailia, Egypt. logical Society of Nevada 33: 329–334. farina, p., catani, f., coloMBo, d., fuMagalli, a., clarK, r.n., swayze, g.a., wise, r., livo, K.e., KuKav icic, M., MarKs, f. & Moretti, s., 2005: Hoefen, t.M., KoKaly, r.f. & sutley, s.J., Remote sensing: a tool for landslide investiga-

Fig. 14: Final lithological map of Dahab basin area. 726 Photogrammetrie • Fernerkundung • Geoinformation 6/2012

tions at a basin scale. – Geophysical Research Area, Saudi Arabia. – Arabian Journal for Sci- Abstracts 7: 10157–10168. ence and Engineering 33 (1C): 99–116. gad, s.& KusKy, t.M., 2007: ASTER spectral ra- r eda, a., KusKy, t. & gHulaM, a., 2010: Litho- tioing for lithological mapping in the Arabian- logical mapping in the Central Eastern Desert of Nubian shield, the Neoproterozoic Wadi Kid Egypt using ASTER data. – Journal of African area, Sinai, Egypt. – Gondwana Research 11 (3): Earth Science 56: 75–82. 326–335. rowan, l.c. & Mars, J.c., 2003: Lithologic map- goMez, c., delacourt, c., alleMand, p.,ledru, p. ping in the Mountain Pass, California area using & wacKerle, r., 2005: Using ASTER remote Advanced Spaceborne Thermal Emission and sensing data set for geological mapping in Na- Relection Radiometer (ASTER) data. – Remote mibia. – Physics and Chemistry of the Earth 30: Sensing of Environment 84: 350–366. 97–108. said, r., 1962: The Geology of Egypt. – 377 p., El- Hassen, i.s., iBraHiM, s.K.& eleMer, p.M., 2004: sevier,Amsterdam. Evolution and origin of the metavolcanics at salisBury, J.w., walter, l.s. & vergo, n., 1987: Wadi Saâl area, south Sinai, Egypt. – Annals Mid-infrared (2.1–25μm) Spectra of Minerals. Geological Survey Egypt 27: 61–78. – United States Geological Survey, Open File Hassen, i.s., el sHafei, M.K. & stüwe, K., 2007: Report 1: 87–263, Reston. Late Proterozoic crustal evolution in the Arabi- soliMan, f.a. 1986: Geology of Wadi Sa’al area an-Nubian Shield of Wadi Zaghra tectonites, with special emphasis of metamorphism and tec- South Sinai, Egypt. – Annals Geological Survey tonics, Central Sinai of Egypt. – 240 p., Ph.D. Egypt V.XXIX: 77–93. thesis, Faculty of Science, Canal Universi- Hewson, r.d., cudaHy, t.J. & Huntington, J.f., ty, Ismailia, Egypt. 2001: Geologic and alteration mapping at Mt Fit- waldHoff, g.,BuBenzer, o.,Bolten, a., Koppe,w. ton, South Australia, using ASTER satellite- & BaretH, g., 2008: Spectral analysis of AS- borne data. – International Geosciences and Re- TER, Hyperion and QuickBird data for geomor- mote Sensing Symposium 2: 724–726. phological and geological research in Egypt HuMe, w.f., 1906: The Topography and Geology of (, Western Desert). – XXI ISPRS the Peninsula of Sinai (south-eastern portion). – Congress, WG VIII/12: 1201–1206, Beijing. 280 p., Egyptian Survey Department, . yaMagucHi, y.& naito, c., 2003: Spectral indices Hunt, g.r., 1977: Spectral signatures of particulate for lithologic discrimination and mapping by us- minerals in the visible and near infrared. – Geo- ing the ASTER SWIR bands. – International physics 42: 501–513. Journal of Remote Sensing 24 (22): 4311–4323. Hunt, g.r., 1979: Near infrared (1.3–2.4 μm) spec- zalata, a.a.,el Metwally, a.a.,el aassy, i.e. & tra of alteration minerals – Potential for use in el sayed, a.a., 1997: Evolution and Geochem- remote sensing. – Geophysics 44: 1974–1986. istry of the Basement rocks of west Dahab area, KoBayasHi, c., oriHasHi, y., Hiarata, d., naranJo, Southeastern Sinai. – Third Conference on Geo- J.a., KoBayasHi, M.& anMa,r., 2010:Compo- chemistry, University: 1–15. sitional variations revealed by ASTER image analysis of the Viedma Volcano, southern Andes Addresses of the Authors: Volcanic Zone. – Andean Geology 37 (2): 433– adel oMran and Prof. Dr. volKer HocHscHild, In- 441. stitute of Geoscience, Tübingen University, Rüme- Kora, M. & genedi, a., 1995: Lithostratigraphy linstr.19–23, D-72070 Tübingen, e-mail: adel- and facies development of Upper Cretaceous [email protected], carbonates in East Central Sinai, Egypt. – Facies [email protected] 32: 223–236. Knipling, e.B., 1970: Physical and physiological Prof. Dr. MicHael HaHn, Faculty of Geomatics, basis for the relectance of visible and near-in- Computer Science and Mathematics, University of frared radiation from vegetation. – Remote Applied Sciences Stuttgart, Schellingstraße 24, Sensing of Environment 1: 155–159. D-70174 Stuttgart, e-mail: michael.hahn@hft- Madani, a. & eMaM, a.a., 2009: SWIR ASTER stuttgart.de band ratios for lithological mapping and mineral Associ. Prof. Dr. aHMed el-rayes and Prof. Dr. Mo- exploration: A case study from El Hudi area, HaMed geriesH, Faculty of Science, Suez Canal Uni- southeastern desert, Egypt. – Arabian Journal of versity, University,Ismailia, e-mail:ahmed_elrais@ Geosciences 4: 45–52. science.suez.edu.eg, [email protected] Qari, M., Madani, a., MatsaH, M. & HaMiMi, z., 2008: Utilization of Aster and Landsat data in Manuskript eingereicht: März 2012 geologic mapping of basement rocks of Arafat Angenommen: August 2012