Multi-Sensor Remote Sensing Data and GIS Modeling for Mapping Groundwater Possibilities: A Case Study at the Western Side of Assiut Governorate, Egypt

Abstract


Introduction
In the last few decades, the water demands in Egypt is strongly increased due to the overpopulation and the ambition of reclamation projects all over the country (i.e., the New Valley, Toshka, New Delta, and the 1.5 million feddan El Salam reclamation projects) (Salem et al., 2016).In fact, groundwater has special importance due to the desert conditions prevailing in Egypt.With a limited share of water and deserts covering about 96% of the total area, Egypt relies on groundwater, which represents about 20% of the total potential water resources (Allam et al., 2003).Therefore, groundwater occurrence, assessment, management, and protection became significant and strategic task.Nowadays, the water shortage in Egypt pushes the government to invest and establish several water recycling projects, such as the saline water plants and wastewater management plants.These projects utilize the latest technology to produce/reuse fresh water from saline or wastewater, but this operation is expensive and consume huge amount of energy.According to the UN, Egypt has entered the water poverty area where the water per capita amounted is less than 500 m 3 /year (Mahrous et al., 2016).Hence, the Egyptian government pays more attention to locate and define alternative renewable water resources, which would be the base for establishing new development and communities in Egypt.The study area locates in the western part of the Assiut Governorate and covers an area of about 635 km 2 in the arid zone, which is defined as hot and dry climatic conditions in the summer and temperature is low in the winter with rare rainfall.
Traditional techniques, such as geophysical tools, have been used successfully for mapping groundwater potentialities, especially in deserts (Elbeih et al., 2015).However, it is time and money consuming, complicated and only effective in local (small) areas.Therefore, the main goal of this study is to detect the high groundwater probability zones based on investigating the effect of physiographic and morphological features of sub-basins (catchments) that contribute positively in groundwater recharge and storage.Furthermore, field measurements were analyzed to validate the multi-sensor remote sensing data results to increase the confidence of the produced Groundwater Potential (GWP) map in the study area.The produced map can be used later as a supplemented tool with conventional geophysical methods.This relatively new technique was applied successfully in regions with similar climatic conditions (Amer et al., 2013;Elbeih, 2014;Abrams et al., 2018, andZein El-Din et al., 2018).
The Relatively limited previous studies on the study area are summarized in the following .El Meligy (2004) classified the area of Assiut Governorate into four main hydrogeological provinces, each has its own geologic, hydrographic, and hydrologic characteristics; Eastern province includes the eastern plateau, southern province includes (Abu Tieg, Assiut), northern province extends widely from Assiut to Dayrout and east-Ibrahimia province in which the groundwater is directly recharged from the Ibrahimia Canal.Ahmed (2013) evaluated groundwater for drinking and domestic uses in Dayrout district and concluded that the TDS concentrations in groundwater at Dayrout district increased toward the desert fringe and decreased toward the floodplain areas.Moreover, Mahmod (2015) studied groundwater evaluation for drinking and domestic uses in Assiut Governorate, Egypt.
ElDakar et al. ( 2016) introduced a distribution map for the TDS concentration using GIS at Assiut, Egypt, and studied the interaction between surface water (SW) and groundwater (GW) and its impact on the groundwater salinity of quaternary aquifer in the area between Assiut and Dayrout, Egypt.
According to El Meligy (2004), the Eocene limestone aquifer is the primary aquifer in the region, which has a limited supply of water and is made up of fractured carbonate rocks of Samalut, Minia, and Drunka formations.This aquifer overlies the Nubia sandstone (Pre-Cenomanian) aquifer, the Quaternary and Plio-Pleistocene aquifers underlying this aquifer.It is a paleokarstified, low to moderately productive aquifer with minimal recharge from deep aquifers and surface runoff.
The salinity of the groundwater varies between 200 and 900 ppm, demonstrating the freshness of the water and this aquifer's groundwater depth is between 90 and 200 m, the transmissivity of the Eocene limestone aquifer ranges between 18.3 and 1758, This large fluctuation relates to the variation in voids and fractures intensity.The Quaternary aquifer is situated on the study area's east edge and has a variety of compositions, from graded Pleistocene sand and gravel intercalated with clayey lenses taking up one area.Then comes the Plio-Pleistocene sand and clay intercalations with gravel encompassing an area with a width of 1-3 km at the edges (El Meligy, 2004).This aquifer's groundwater often has unconfined conditions (Said, 1962).
The current study will employ the powerful coverage of remote sensing data and geospatial technique to indirectly detect zones with high groundwater probabilities within the study area.
The research area is a portion of the limestone plateau's reclaimed land, west of the governorate of Assiut.The region's development situation is greatly influenced by the geological setting.The types of rocks, the degree of porosity, Surface features, geologic structures, the composition of the sediment, the rarity of floods and the degrees of decline play a valuable role in the impact on the geo-environmental.
In essence, within the study area, the sedimentary sequence exposed consists of Quaternary rocks and sediments and Eocene.From bottom to top, the Eocene rocks are divided into two formations: Drunka and Minia (Fig. 1b).

Data and Methodology
The study depends mainly on remote sensing data and GIS techniques with field validation, which can provide modern tools to determine sustainable groundwater resources for optimal uses.The remote sensing data included optical, radar and thermal satellite images for mapping GWP, especially in light of the scarcity of surface flow data in the study area.
To represent the physiographic impact on groundwater pools, the data types were selected an approved methodology was followed using the weighted index overlay analysis technique for computing the variable weights and producing the GWP map.These variables were extracted from remotely sensed data to make the proper integration between them to predict the shallow groundwater potential sites.The workflow for the utilized data and methods consists of two main procedures; first, variable extraction using the compiled remote sensing data products, and second, GWP analysis and modeling (Fig. 2).

Remote Sensing Data
Remote sensing technology has advanced to advanced levels of practicality, with regard to optical, radar, and thermal infrared images.Optical images can be used to map surface features, present land cover/use, and vegetation distribution patterns in addition to lithological details of the study area.On the other hand, space borne radar images are a powerful tool in mapping surface roughness and characteristics of rivers buried in sand, such as ancient river beds and lake basins, e.g., Mccauley et al., 1982;Ghoneim and El-Baz, 2007;and AbuBakr et al., 2013.In addition, the Shuttle Radar Topography Mission (SRTM) is the most common source of the Digital Elevation Model (DEM) and is strong in fixing surface drainage networks (e.g., Ghoneim et al., 2007).Thermal infrared (TIR) can be used to track the accumulations of moisture content near the surface after precipitation events, especially in sandy deserts (e.g., Ghoneim, 2008), Like the area under investigation.
In this study, optical, radar and thermal data were utilized, including (ALOS PALSAR DEM 12.5 m), Operational Land Imager (OLI), Moderate Resolution Imaging Spectroradiometer (MODIS), Shuttle Radar Topography Mission (SRTM), Tropical Rainfall Measuring Mission (TRMM), and Radarsat-1.These wide set of data were downloaded from the USGS and NASA's websites, which are open sources (Table 1).

Methodology
According to the workflow in Fig. 2, the adopted methodology was divided into two main parts, first building the required thematic layers and second, applying the modeling techniques to detect groundwater Potentiality.

Building the required thematic layers
We have created ten thematic layers (variants), which are mainly extracted from satellite data.These layers are elevation, slope, curvature, drainage density, Topographic Wetness Index (TWI), surface roughness, frequency of thermal anomaly, accumulated precipitation, Land Use / Land Cover (LULC) and lineament density.These variables will be mentioned in detail in the following sections, based on their data sources.

• Elevation and DEM derived layers
The Digital Elevation Model (DEM) data were created by mosaicking two tiles of SRTM1Arc-Second (30 m), and then this mosaic was cropped to match the boundary of the study area.The spatial analyst tools in ArcGIS10.5 were used to produce the slope, standard curvature, and drainage density layers (Fig. 3a, b, c and d).
The slope and curvature provide an explanation of the infiltration conditions in water catchments (basins).The slope measures the change of relief per degree (raster unit) and thus measures the runoff acceleration from upstream to downstream.The slope (degrees) layer was derived from the DEM using the surface tools in ArcGIS.The SRTM-DEM was also used to produce the standard curvature layer, which gives an interpretation to the concavity and convexity of the ground surface as positive and negative values, respectively.
According to Jenson and Dominque (1988), the surface flow direction was computed using the eight directions model (D3), then the flow accumulation was calculated, and the drainage network was determined using a threshold at 300.Ultimately, a density map was extracted based on the produced drainage patterns and calculated in a square kilometers.

• Topographic wetness index (TWI)
The Topographic Wetness Index (TWI) was developed by Beven and Kirkby (1979) to expresse the steady state of wetness, which is usually applied to quantify the topographic control on hydrological processes (Sørensen et al., 2006, Abrams et al., 2018).TWI is usually applied to assess the effect of topography and soil moisture depending on the formerly created slope and flow accumulation layers.The TWI layer was calculated using the following equation: ln (a/ tan b) (1) Where, a is the upslope area of a given cell and b is the slope in radians.Using the above equation the TWI was created in the raster calculator tool in ArcGIS software (Fig. 4a).

• Surface roughness
According to Gaber et al., 2011, Surface Roughness (SR) can be extracted based on radar back scattering waves, which has a strong effect on microwave signals.According to Lee and Pottier, 2009, the L-band with its longer wavelength has more penetration depth and could be less sensitive to slight changes in wadi surface roughness (micro-roughness) than the microwave with shorter wavelength of C-band.Therefore, Radarsat-1 satellite images (C-band) were used to create the surface roughness map of the study area to classify the texture of surface sediments into five classes: very rough, rough, medium, fine, and very fine materials (Fig. 4b).
• Land surface temperature (LST) According to AbuBakr, 2014, Monitoring thermal change in the bare sandy surface after precipitation can provide indirect indications of groundwater accumulations.We used thermal remote sensing data here to identify and examine thermal anomalies in the study area to identify possible surface groundwater infiltration zones.In other meaning, it tries to uncover the possible link between these anomalies and groundwater leakage.This method was used successfully to detect groundwater in similar arid lands, such as the North Sinai Peninsula and the United Arab Emirates (AbuBakr, 2014;Ghoneim, 2008).
A systematic analysis for thermal infrared (TIR) data of daily MODIS LST were extracted from the MODIS MYD11A1 V6 image, which provides per-pixel land surface temperature (LST) in a 1200 x 1200 kilometer grid (Wan et al., 2015).MODIS tiles were re-projected from their original Sinusoidal (ISIN) projection to UTM-36N projection and then combined and clipped to the study area boundary, and finally the pixel values were converted from Kelvin to Celsius degrees (Fig. 4c).

• Accumulated Precipitation
Due to the lack of rainfall monitoring stations in the study area, the Tropical Rainfall Measuring Mission (TRMM) was used to estimate the accumulated precipitation.TRMM is a joint space mission between NASA and Japan Aerospace Exploration Agency (JAXA) designed to monitor and study tropical rainfall.This mission provides global rainfall data with a 0.25° × 0.25° footprint since 1997.Despite the coarse resolution of the satellite rainfall products, they are used in a wide range of applications due to their being cost effective and available for the globe (Harris et al., 2007).Thus, TRMM was employed here to estimate the amount of rainfall events that were occurred between 2012 and 2018.
The precipitation data was extracted using TRMM 3B43 V7 for 6 years (from 2012 to2018).The TRMM data were converted from net cdf format to vector points then an interpolation process was applied using spline method to convert back the point vectors to interpolated raster, which was then clipped and adjusted to the extent of the investigated area (Fig. 4d).

• Land use/land cover (LULC)
The LULC is significant for mapping the variation in soil types and anthropogenic activities in the plain area, which is critical for runoff infiltration.These layers were produced from the OLI sensor, which was mounted on the Landsat-8.Several pre-processing analyses were applied, including atmospheric correction, which removes the effects of the atmosphere (scattering and absorption) and the DN was ultimately converted to surface radiance.After that, a Principal Component Analysis (PCA) was carried out as an image transformation technique to enhance the spectral resolution of the images to reduce the redundancy of the data before going through the image classification (Lillesand & Kiefer, 1987).Another step was performed prior image classification, which is the extraction and masking the vegetation in the study area through applying the Normalized Difference Vegetation Index (NDVI) (Weier and Herring, 2000).The resulted vegetation vector layer is compared with the OLI image using a false band combination (bands 5, 4, 3 in RGB, respectively) as a verification step before proceeding to the classification process.A K-Means classification method was adopted to classify the land cover in the study area.The technical classification parameters were comprised of a threshold of 5%, and maximum iterations of 50 to produce 20 classes.The spectrally similar classes were merged together according to the surface lithology and the type of land use with the aid of the published geological map (EGSMA, 1994) and high resolution imagery in Google Earth.The final product involved 5 classes of LULC: basement rocks, old alluvial deposits, recent and wadi deposits, sedimentary rocks, urban, roads and vegetation (Fig. 5a).

• Lineament density
According to (Smith and Wise, 2007) the precision of lineament extraction from optical satellite images relied on sensor characteristics, spatial and immaterial separation, and weather board-brightness qualification.Therefore, lineaments were extracted from theLandsat-8imagesafter applying the Gram-Schmidt transformation to enhance the spatial resolution of the multispectral OLI image (Laben and Brower, 2000).The are lineaments then exported to Google Earth software for usage to validate them and remove false lineaments, such as roads, trail, and other manmade activities.Ultimately, the lineaments with high degree of confidant were selected and a line density layer was created using the same procedure groove of the drainage density layer (Fig. 5b).

Groundwater potentiality indexing and modelling
It is pointless to perform groundwater potentiality modeling directly on the extracted thematic layers without a weighted index connecting them together based on their groundwater contribution.Therefore, a semiquantitative weighted index (GWP indexing) was used to divide each thematic layer into only 5 classes.
• Semi-Quantitative Weighted Index This index was selected based on previous studies on regions with similar conditions (e.g., AbuBakr et al., 2013;Abrams et al., 2018).Furthermore, field investigations in the study area and literature review significantly contributed to the assignment of the GWP indexing.
The layers were approximately changeless and classified into five classes using the natural break (Jenks) method (Jenks, 1967), to increase the resolution across the study area (Table 2).
Each class was assigned to a common scale value, which could frequent from 0 to 100 as a value of contribution in the GWP for propagate the conclusive formula of the thematic layers (Figs. 6,7 and 8).Larger scale value indicative to higher GWP and imperfection versa.For token, the very supercilious height and steep slope represent very low contribution (0-20) in GWP, while very low elevation and bland sloping contemplate very high to high contribution (80-100) in GWP (Table 2).• Simple additive weight method (SAW) There are several techniques of GWP overlay modelling, most of them have been standard and estimated, such as Probabilistic Frequency Ratios (PFR) (Manap et al., 2013;Naghibi et al., 2015), the Analytical Hierarchy Process (AHP) (Adiat et al., 2012;Mallick et al., 2015), and the Simple Additive Weight (SAW) (Abrams et al., 2018;Al-Ruzouq et al., 2015).
According to these studies, especially Abrams et al., (2018), the Simple Additive Weight method (SAW) is one of the effective methods for modeling groundwater potential.The SAW technique is the most primitive of the semiquantitative index-overlay techniques.It assigns equal weight to all variables (layers) and assumes that the assigned GWP values are adequate to represent potentiality (Abrams et al., 2018).Iraqi Geological Journal Khattab et al. 2023, 56 (1C), 176-190 186 Therefore the 10 weighted thematic layers were summed together using the following SAW formula: GWP SAW =∑ GWP E , GWP S , GWP DD , GWP LD , GWP C ,GWP LC ,GWP SR ,GWP LST , GWP TWI , GWP AP (2) Where, GWP S , is the SAW technique for all groundwater potential, GWP E potential for elevation, GWP S potential for slope, GWP DD potential for drainage density , GWP LD potential for lineament density, GWP C potential for curvature, GWP LC potential for land cover, GWP SR potential for surface roughness, GWP LST potential for land surface temperature, GWP TWI potentialfor topographic wetness index and GWP AP is the potential for accumulated precipitation.
The interpretation of the GWP map reveals that most of the very high (VH) zone was defined along the north side of the study area.In other words, the very high GWP zone is extended along the cultivated land at the western bank of the Nile River.The extension of this zone is reasonable and can be attributed to the massive runoff that originates from the southern part and flows toward the northern part, where the flow velocity slows down due to the gentle slope and provides optimum conditions for the infiltration process.On the other hand, the high (H) GWP zone is distributed in regions with high lineament density and surface roughness.The Moderate (M) zone occupies a relatively smaller portion of the plain.The low and very low (L and VL) zones are predominately in the southern part of the study area due to the steep slope and low infiltration conditions.These outcomes conclude that the VH and H GWP were generally associated with low terrain, high surface ruggedness, dense drainage, and lineament, relatively close to thermal anomalies, and fluvial wadi deposits or sandy areas.

Field Validation
The fieldwork of the present study was carried out in April, 2018 to confirm the remote sensing data results and the accuracy of the GWP map (Fig. 9), which included several confined pumping and recovery tests.
Fieldwork also included projections of some well locations on the final GWP map to validate model assumptions for the five different GWP zones (Fig. 9).The results showed that 100% of the wells were located in the high GWP zone.Most of these wells were drilled according to geophysical surveys (mainly geo-resistivity), which produced good water quality for industrial and agricultural purposes.Therefore, based on the distribution of the investigated wells in the study area, the overall accuracy of 2023, 56 (1C), 176-190 188 the GWP is roughly estimated to be around 81%.However, further geophysical studies of the delineated VH and H GWP belts in the study area are recommended.

Conclusions
This study integrates remote sensing data and geospatial techniques for mapping groundwater potential in one of the most promising locations for development in Upper Egypt.The significance of utilizing remote sensing data instead of conventional land-based techniques came to produce a continuous surface map with higher resolutions and consistent scales than previously published maps, and to overcome the scarce of surface flow data.The key objective of this work was identifying physiographic variables that affecting groundwater recharge and storage for detecting renewable ground water accumulation.A huge variety of remote sensing data, which includes optical, radar, and thermal sensors have been used to assemble numerous geospatial thematic layers (variables).These variables consist of elevation, slope, curvature, drainage density, topographic wetness index, floor roughness, frequency of thermal anomaly, gathered precipitation, Land Use/Land Cover (LULC) and lineament density.All variables have been organized and weighted primarily based on their contributions in groundwater recharge via infiltration and percolation processes.The Simple Additive Weight (SAW) technique turned into followed for computing the variable weights and generating the GWP map.
Fieldwork was conducted along the study area to validate the remote sensing data results and the accuracy of the produced GWP map through pumping test, collecting water well samples, mapping farms and other vegetation patterns in the study area.The fieldwork shows that most of the water wells are located in the high GWP zones, Therefore, the general accuracy of the produced GWP changed into the kind of anticipated at approximately 81%.However, for vital discipline validation to the result, geophysical investigations need to be performed on the provided VH and H GWP zones.In conclusion, the existing provides a geospatial-based tool to define zones with excessive groundwater opportunity for downscaling the exploration efforts via traditional land-primarily based totally techniques and might be carried out somewhere else in arid regions.

Fig. 1 .
Fig. 1.A)Location map overlaid with topographical Digital Elevation Model (DEM) for the study area; B) Geologic map of the study area (modified after Conoco, 1987)

Fig. 2 .
Fig. 2. Simplified work flow for the remote sensing data and the methodology; 1) Construction of thematic layers; 2) groundwater potentiality modeling.

Fig. 9 .
Fig. 9. Groundwater potential map with locations of field verification sites Table1.Remote sensing data description, sources and associated layers.

Table 2 .
All variables and weights according to contribution to GW