Enhanced lithological mapping of Durba‑Araba basement blocks, along the eastern margin of the Central Gulf of Suez Rift, Egypt, using Landsat‑8 Data

The eastern margin of the central Gulf of Suez rift exposes unique rotated basement blocks of Gabal Durba-Araba. According to geological maps, these blocks are composed of Dokhan volcanic, older, and younger granites cut by basaltic dikes and overlayed by a tilted sedimentary succession of Paleozoic to quaternary. Herein, we propose a workflow of stepwise preprocessing and classification procedures for Landsat-8 (OLI) data verified by control points from previous maps to update and improve lithological discrimination in converging areas that require further field mapping. The workflow applies digital image-processing techniques, including spectral signature analysis, band-ratio transformation, maximum likelihood, and Mahalanobis supervised classifications. The accuracy assessment of the lithological mapping reaches 86.6% and 75.7% for maximum likelihood and Mahalanobis classifications, respectively. By running the Tactical Hyperspectral Operations Resource (THOR) algorithm on the classified images, we could accurately map and modify the lithological boundaries for the following rock units: (1) gneiss-amphibolite-schist, tonalite, and granodiorite at Gabal Araba; (2) granitic gneiss, granodiorite, biotite-muscovite-leucogranite, and riebeckite-syenite-albitite at Gabal Abu Haswa; (3) Dokhan volcanic, granite biotite-muscovite-leucogranite, granites riebeckite-syenite-albitite, Pyroclastic breccia and lava flow, and NW tertiary dike at Gabal Abu Durba.


Introduction
The Gulf of Suez rift is a significant geological province with a long history of oil and gas production and mining exploration. Its Neogene opening led to the development of several tilted fault blocks with three transverse provinces (north, central, and southern province; Moustafa 1976). For a long time, the unique exposures of the rotated basement block structures and stratigraphy on both sides of the Gulf of Suez have attracted geologists worldwide (Ungerer et al. 1984;Montenat et al. 1988;Moustafa and El-Raey 1993;Abdelkareem et al. 2021). The study area contains three exposed rotational fault blocks cored with the Neoproterozoic basement. The basement cores (including Gebel El Zeit and Esh El Malaha on the southwestern margin and Gabal Durba-Araba on the central-eastern margin) represent a surface analog model for the unconventional subsurface basement reservoir in the rift area. There is a need for detailed mapping of the basement cores to understand their subsurface extensions and discover new areas suitable for energy and mineral development.
In arid and semi-arid environments, as the study area, remote sensing datasets and their superior technology are effective for geological mapping due to the exposed bedrock and lack of soil or vegetation coverage (Elhebiry et al. 2020;Abdelaal et al. 2021). Owing to their spatial, spectral, and radiometric properties, remote sensing techniques are efficient in lithological mapping. The digital format of remote sensing data enables advanced image processing techniques, which, in turn, enhance the extraction of information for mapping (Ghrefat et al. 2021). Remote sensing is a significant tool for tracing lithological boundaries, and it is widely used in regional and detailed mapping, geomorphological and structural interpretation, and mineral prospecting (Crosta 1989;Hassan et al. 2015;Hassan and Sadek 2017;Zakeri et al. 2017;Gaber et al. 2017;Beiranvand Pour et al. 2018;Abu El-Leil et al. 2019; Rajan Girija and Mayappan 2019; Hamimi et al. 2020;Abdelkareem et al. 2021). The progress made so far in the optical Landsat-8 (OLI) data is appropriate for remote lithological mapping and reducing the cost and time of field surveys and displays an array of multispectral data spanning a large area (Zhu and Abdelkareem 2021). The use of the Landsat-8 OLI multispectral dataset enhances structural, lithologic, and geomorphic remote sensing mapping (El-Din and Abdelkareem 2018; Abdelkareem et al. 2020). It comprises high radiometric resolution (16 bits) that covers a wide spectrum of nine spectral bands with a spatial resolution of 30 m for bands 1-7 and 9 divided into shortwave infrared (SWIR), thermal infrared (TIR), and visible segments (Table 1).
Employing remote sensing techniques and field verification, Seleim and Hammed (2016) and Abdelkareem et al. (2021) provided a detailed lithological map for southwestern rotated blocks in the Gulf of Suez rift. Most studies on the rotated basement blocks of Gabel Durba-Araba were based on field observations, geochemistry, and geochronology sampling (e.g., Egyptian Geological Survey 1994; El-Bialy 1999; Anbar et al. 2005;and Geological Survey 2015). These mapping attempts produced large-scale maps (Egyptian Geological Survey 1994;Geological Survey 2015), yet military restrictions and limited accessibility hindered sample collection from the study area. Lithological mapping depending solely on regional field surveys and geochemical analysis assisted with aerial photos of low spatial resolution results in geological maps with poorly defined boundaries (e.g., Eyal and Hezkiyahu 1980;Gharib and Obeid 2004;Azer 2006). Thus, there is a need to discriminate rock units and accurately define their contacts by applying remote sensing techniques using high-resolution satellite datasets.
The study area is a part of the Afro-Arabian Nubian Shield, which was intensively mapped using various remote sensing sensors and field verified (El-Din and Abdelkareem 2018;Abu El-Leil et al. 2019;Asran and Hassan 2021;Shokry et al. 2021). The efficiency of OLI sensors in lithological discrimination has been tested and validated in the same terrain (Seleim and Hammed 2016). Therefore, in this study, we employed advanced image processing techniques on multispectral OLI datasets to enhance previous maps and differentiate the lithologies of Gabal Durba-Araba (Fig. 1). The following techniques were used in this study: 1. Optimum Index Factor (OIF), a statistical approach widely used to define the most advantageous three-band combinations (Qaid and Basavarajappa 2008) 2. Band composite image 3. Detailed spectral signature analysis for material identification and characterization (Govender et al. 2007) 4. Band rationing (BR), which has been widely used for mapping mineral abundance from multispectral Landsat datasets (San et al. 2004). In this study, BR was created from pixel reflectance instead of the digital number (DN), which is a more related characteristic of identified rock units

Geological setting
The Pan-African basement subsegments of granitoid, volcanic, and dikes are exposed in the central part of the Gulf of Suez named from north to south Gabal Durba, Gabal Abu Haswa, and Gabal Araba (Fig. 1a). Egyptian Geological Survey (1994) classified the Neoproterozoic of Gabal Durba-Araba into four rock units, and Gabal Araba was divided into gneiss-amphibolite-Schist, tonalite, diorite, granodiorite, and biotite-muscovite-leucogranite. Gabal Abu Haswa is classified into two (biotitemuscovite-leucogranite and riebeckite-syenite-albitite), and Gabal Durba is classified as one rock unit of granite riebeckite-syenite-albitite (Fig. 1b). The basement rocks of Abu Durba Abu Haswa comprise the following rock units: older granites, phase II younger granites, phase III younger granites, post-granitic dykes, acidic porphyries, pyroclastic, and lava flow of tertiary basalts (El-Bialy 1999). Fieldwork and petrography reported by Anbar et al. (2005) indicate that the northern part of Gabal Abu Durba is occupied by phase II younger granites (granodiorites), intrudes into volcanics of the Dokhan type, and sends off-shoots and apophyses into the Dokhan volcanic. The Dokhan volcanic and younger granites are dissected by post-granite basic dikes of tertiary age trending NW. The phase III younger granites (monzogranites) located at the south-eastern part of Gabal Abu Durba intrude into volcano of the Dokhan type. The Gabal Abu Haswa, on the other hand, is composed mainly of phase II younger granites (granodiorites), which are intruded by phase III younger granites (monzogranites and alkali feldspar granites). Geological and Survey 2015 divided the basement  (Hammed 2002). The two faults are in the study area, and both are hard-linked normal fault segments in zigzag arrays and of two dominant strikes, Gulf parallel NW-trending and Gulf oblique NNE-trending faults (Morley 1995). The plain area between the two basement blocks is a sedimentary depocenter known as the El Qaa basin (Fig. 1a)

Dataset and preprocessing
The approach used in this study updates the lithological map of the study region. The dataset was derived from a Landsat-8 scene (175/040) row, UTM projection of zone 36, and WGS-84 datum taken in October 2019 with scene ID "LC81750402019225LGN00." The data were obtained from the US Geological Survey (USGS). Many computational processes, such as data preparation, feature extraction, model construction, and model assessment, are required in modern data science. For land use/ land cover mapping, even a simple correction method can give satisfactory results, providing pure pixels and spectral signatures corresponding to specific land cover types for classification (Chrysoulakis et al. 2010). Preprocessing steps are required to improve the dataset used in the compilation of a remote predictive lithological map of the study area, including layer stack, subset by region of interest (ROI). Moreover, several procedures are necessary to prepare data for conversion from digital number (DN) to reflectance data, including masking the water band, radiometric correction, and fast lineof-sight atmospheric analysis of hypercube algorithm (FLAASH), which was used herein for atmospheric correction (Cooley et al. 2002). Some pixels may fall outside this range; they generally correlate to highly reflecting surfaces that cause saturation, or they might be dark, negative values seen in deep water or shadows. When the radiance is low, the reflectance values fall outside the normal range (Table 2). Hence, there is a need to rescale the radiance data, where the Band Math tool is used to divide the pixel values by 10,000 to convert them to floating-point values ranging from 0 to 1.0 using the post FLAASH equation (Matthew et al. 2000) for every 7 bands (Table 3): Comparing images before and after FLAASH preprocessing, we find that the processed images are sharper and more accurate (Fig. 2).

Data processing and methodology
The remote-sensing technique used in this work is described in detail in Fig. 3. It consists of seven major stages.
(1) OIF was utilized to define the most beneficial three-band composite image.
(3) The spectral signature for the image's different rock units was extracted using ENVI 5.3 software ( Fig. 6). (4) The collected spectral signatures of the examined rock units were used as a guide to select the best spectral BR transformation images. (5) THOR was used for detecting unknown rock units in ENVI 5.3 software by comparing the target spectral signature to the reference spectral applied for the first time in the study area. (6) Supervised image classification was performed.

Fig. 2
Gabal Abu Haswa in band composite images 7 6 2 RGB above and 7 5 1 RGB below before (left) and after (right) FLAASH atmospheric correction (7) All remote-sensing datasets and processing efforts in Sinai were verified by previously published geology and topography maps. ENVI 5.3, ESRI ArcGIS 10.5, and ILWIS software were used to perform image preprocessing and analysis (Fig. 3).

Color composite images
The satellite image, which appears in a distinct band as a black and white band, indicates a low and hazy contact between the lithological rock groups. When presented as red, green, and blue imagery, the comparison of spectral properties of land features in several bands (color composites) shows improved separation/contrast across distinct surfaces. In the case of Landsat-8, 4-3-2 RGB bands, the combination of the same spectral bands, such as the visible band, is a true-color composite image (Fig. 1a). When the image is merged between multiple spectral bands, it is referred to as a false-color composite (FCC), and it maintains its morphological characteristics and presents distinct lithological units in different hues (Al-Nahmi et al. 2016). Herein, we employed many FCC of the OLI band as 7, 6, 2 RGB and 7, 5, 1 RGB combination based on OIF to select the optimal band combination of Landsat OLI data using ELWIs 3.4 software (Table 4). Then, the images were used to extract the spectral signature of rock unit as they characterize different rock boundaries (Figs. 4 and 5).

Band ratio
BR is employed in remote sensing to efficiently depict spectral fluctuation and increase material contrast. It was easily applied by dividing the DN of each pixel in one band by that of another (Drury and Walker 1987;Sabine 1999). Depending on the research purpose, BR can also convey spectral or color information. Different forms of BR can be used for a range of purposes, such as lithological mapping or alteration discrimination. The properties of image components change in scene light (Omwenga 2018). After eliminating the atmospheric state, several BR composite images were created using OLI data. Spectral band rationing is particularly effective for highlighting certain features or materials that are not visible in digital numerical data (Ourhzif et al. 2019).
In this study, the lithological map was enhanced by analyzing the Landsat-8 data to distinguish lithological rock units based on spectral characterization derived from band combinations of 7, 6, 2 RGB and 7, 5, 1 RGB data images.
The spectral contrast between the bands shows the following: • A low reflectance characteristic with visible bands-2, 3, and 4 having a medium reflectance value • The band-6 (SWIR1), on the other hand, has typical reflectance with a differentiation characteristic between the lithological rock groups described by the study area. Band-7 (SWIR2) also has a high absorption rating ( Fig. 6a and b) • The spectral properties of Neoproterozoic rocks of the study area have a few different characteristics around bands 2 and 3, with an emphasized reflectance feature around bands 5 and 7 and typical differences in reflectance features for band 6 • Neoproterozoic rocks represented by gneiss-amphibolite-schist, granitic gneiss, tonalite, diorite, granodiorite, granite G1, Dokhan volcanic, granites G2 and G3, pyroclastic breccia, and tertiary dikes reveal that maximum reflectance occurs at 1.609 µm of the SWIR region of band 6 and the minimum reflectance at 0.4826 µm of the visible region of band 2, except for tonalite and granite G3, where the minimum  show a low differentiation between VINR, NIR, and SWIR electromagnetic waves ( Fig. 6a and b) • The maximum reflectance can be used as the denominator or numerator to differentiate lithological rock units of the study area (e.g., 7/6, 6/7, 6/5, 6/4, 2/6, 5/6, 5/4, 4/6, 4/3, 5/7, 4/5, 4/2, 3/2, 6/5).

Material identification
Many spectral matching algorithms, ranging from the traditional clustering techniques to more recent automated matching models, have evolved (Shanmugam and Srinivasaperumal 2014). Comparing it to every signature in THOR-integrated spectral libraries, THOR tools automatically compute the correct scale factor required to scale the unknown signature to the same scale as reflectance libraries. Libraries in THOR include the Imaging Spectrometer Data Analysis System (ISDAS), developed by the Canadian Centre for Remote Sensing in 1995, and the USGS's Tricorder (Clark et al. 1990). These systems provide basic data input and output, interactive display, and data analysis for imaging spectrometers. Single algorithms, such as SAM in SIPS, Fig. 4 The study area and clipping of a Gabal Ab Durba; b Gabal Abu Haswa; c Gabal Araba and its characterized rock units including Dokhan volcanic, different types of granites, volcanic breccia, and dikes shown in false color 7, 6, 2 RGB SFF in Tricorder, and MRSFF in Tetracorder, govern spectral matching for most mineral mapping applications (Shanmugam and Srinivasaperumal 2014). The matching results are sorted with the best matches on top (the smaller spectral angels are good (less than 0.08), fair (0.081-014), poor (0.1141-0.25), and very poor (greater than 0.251) (Material Identification 2021). In this study, the THOR technique was used to detect and recognize different rock unit samples classified as BR 3/4, 4/6, 2/1 RGB, 7/6, 2/6, 5/6 RGB, and 5/2, 3/2, 7/2 RGB, and the results were compared with previous geological mapping of the study area and then used as ground trust ROI for supervised classification.

Supervised classification
Supervised classification requires a significant amount of input from an image analyst and information about the sorts of surfaces observed in the study area. This information may be collected through maps or fieldwork, in which various surface classes are recognized and then entered into the software as ROI. It can also be achieved by collecting end members or acquiring a unique spectrum for each rock segment. Various supervised categories, including classifications based on Mahalanobis distance and maximum likelihood, were tested.

Results
Various remote-sensing image-processing techniques were employed in this study. They are categorized as follows: a) Color composite images The statistical OIF technique was used to pick the most informative color image from the Landsat-8 band composite images. The best band composite image was chosen with a high percentage to extract useful information about the land surface, such as hydrothermally spectral analyst discrimination (Table 4). Band composite images 7, 6, 2 RGB and 7, 5, 1 RGB are more varied and suitable for the rock units in the study area and provide an excellent base map in which rock units are easily discriminated (Figs. 4 and 5).

Discussion
The study area, located on the eastern edge of the central Gulf of Suez rift, exposes elongated basement blocks of Gabal Durba-Abu Haswa and Araba. Previous research has been limited to field surveys, chronology, and chemical analysis (e.g., Egyptian Geological Survey 1994 Fig. 13 a;El-Bialy 1999;Anbar et al. 2005) or remote-sensing techniques (e.g., Geological and Survey 2015: Fig. 14). To compare distinct lithological rock units, both approaches give low-resolution geological maps. In the study area, Landsat-8 data preprocessing and remote-sensing technology interpretations for building BRs for enhanced lithological-unit categorization are effective. All three tested BR combinations could separate at least the principal exposed rock segments in previous research. Two of the combinations, including BR 3/4, 4/6, 2/1 RGB and 7/6, 2/6, 5/6 RGB of Seleim and Hammed (2016), showed the best results. In particular, BR 6/7, 6/4, and 4/2 RGB of Landsat-8 intruded by Zeinelabdein et al. (2014) gained an acceptable contrast and good separability between different types of older granite and Dokhan volcanic, but it could not divide the various types of younger granite. BR 5/2, 3/2, and 7/2 RGB could discriminate between basic, intermediate, and acidic rocks. All tested BR combinations described in previous studies and extracted according to spectral signature analysis could separate at least the primary exposed rock units (e.g., gneiss-amphibolite-schist, granitic gneiss, granite G1, Dokhan volcanic, granite G2, and granite G3, Pyroclastic breccia, and tertiary dikes). The methods used to process and analyze Landsat-8 images could divide rock units in the study area more We compare our results with those of previous studies in three sections: 1. Gabal Durba is a single geological block that comprises granite G3 (Egyptian Geological Survey 1994), and it is divided into three geological units: Dokhan volcanic on the western side, granite G2 on the eastern side, and granite G3 on the northern side (El-Bialy 1999 andAnbar et al. 2005). Pyroclastic and lava flows occur only on the western side of Gabal Abu Durba, principally as a relatively large mass of liver-like shape in addition to other minor smaller bodies (El-Bialy 1999). Geological Survey (2015 reported that Gabal Abu Durba has two rock units: granite G2 and volcanic breccia intruded by tertiary dikes. BR images were used in this study. 6/7, 6/4, 4/2 RGB (Fig. 7) after Zeinelabdein et al. (2014) and 3/4, 4/6, 2/1 RGB (Fig. 8) and BR 7/6, 2/6, 5/6 RGB ( Fig. 9) after Seleim and Hammed (2016) showed that the Gabal Abu Durba is classified into four rock units with clear contact: the eastern part is identified as granite G2, the western part as Dokhan volcanic, and the most eastern part as granite G3, which is introduced by tertiary dikes. Also, pyroclastic and lava flow occur only on the western side. This is contrary to the findings of the Geological Survey (2015). Thus, there is a need to define the kind of rock units using a THOR material identification approach, which revealed that most parts of the western side of Gabal Durba are characterized by Dokhan volcanic rocks with rhyolitic and basaltic andesite compositions (Fig. 11c, d). Furthermore, the most northern portion is distinguished by granite G3 alkali syenite (Fig. 11a) and the eastern part by granite G2 muscovite composition. 2. To the south, the Neoproterozoic of Gabal Abu Haswa contains granitic G2 (El-Bialy 1999), which is divided into two groups: granite G2 and G3 (Egyptian Geological Survey 1994 and Anbar et al. 2005). However, it was classified as granitic gneiss by the Geological Survey (2015). BR 3/4, 4/6, 2/1 RGB and 7/6, 2/6, 5/6 RGB revealed three distinct contacts between rock units, including granitic gneiss, granite G2, and granite G3. El-Bialy (1999) reported that both gneissose and large granites are exposed at the west of Gabal Abu Haswa where they exhibit somewhat dispersed contact, which contradicts the results obtained here. BR 3/4, 4/6, 2/1 RGB and 7/6, 2/6, 5/6 RGB showed that granitic gneiss located at the eastern side is mostly occupied by gneiss. The BR 5/2, 3/2, and 7/2 RGB images also divide the Abu Haswa basement into two rock units shown in green and violet. The THOR material identification approach revealed that the green rock unit is a granitic gneiss (Fig. 11f). 3. Moving south of Gabal Araba, the Egyptian Geological Survey (1994) and Geological Survey (2015) categorized Araba basement rocks into gneisses, migmatites, and amphibolite; granite G1; and granite G2, with differences in contact among the rock groups (Figs. 13a and 14). 6/7, 6/4, 4/2 RGB, BR 3/4, 4/6, 2/1 RGB, and 7/6, 2/6, 5/6 RGB showed clear contact between several rock units in Gabal Araba where the rock unit contact has been reclassified (e.g., tonalite diorite and granodiorite: Fig. 13b).
Following the selection of training data, THOR material identification was effective for the supervised classification process, giving the overall best classification accuracy. The prepared lithological map shows clear improvements and variations from the earlier version. Generally, many predicted borders between the lithological units were set up newly, and others were confirmed. Also, the discrimination between several rock types was enhanced, e.g., the granitic gneiss of Gabal Abu Haswa and the Dokhan volcanic of Gabal Abu Durba. Even though the results are satisfactory, some regions in the study area show vague results. Classifications, such as confusion between granitoid and biotitemuscovite-leucogranite at Gabal Abu Haswa, also classified Paleozoic S.S as Dokhan volcanic at the eastern part of Gabal Abu Durba. Herein, additional field data would be needed to define further training areas for lithological units. Combining the classification results with field verification, a new lithological map could be created (Fig. 13b).

Conclusions
In this study, we employed specific data processing techniques on Multispectral Landsat-8 datasets for the automated lithological classification of the study area. The proposed workflow resulted in a derivative dataset with enhanced information relevant for lithological discrimination of Neoproterozoic rock in Gabal Durba, Abu Haswa, and Gabal Araba. Most geological studies on the study area depend on the published traditional geological maps developed based on sample collection, geochronological, and geochemical processes. Due to the difficulty in the geographic location, sharp topography, and inaccessibility to all places, geological maps of limited accuracy have been developed.
In this study, we employed OIF, a spectral signature analyst; image transformation generated from spectral band rationing; THOR material identification; and ultimately supervised classification. These techniques, in conjunction with previous geological mapping, were employed to update the geological map of the studied area. The revised lithological map (Fig. 13bb) shows more distinct improvements and variances than the previous map. The proposed workflow resulted in higher-resolution maps with accurate rock unit boundaries. Many lithological units have a new set of borders, whereas others are retained. Some rock units are subdivided into smaller and separate units (e.g., the granitoid of the Gabal Araba is divided into four phases, and the granitoids of Gabal Abu Haswa and Gabal Abu Durba are divided into three phases). Despite the high-accuracy assessment results of this study (kappa values 0.7192-0.8634), further field mapping would be useful to establish the rock variations as new classes. Since there is no comprehensive detailed field mapping for large regions, the new geological map cannot guarantee 100% accuracy for some places. However, with a few exceptions, the findings are considered safe due to the high classification quality and outstanding outcomes. Final confirmation of this result can only be obtained by extending the field mapping to locations not previously studied.
Funding Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB).

Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.