LithoSFR Model for Mapping Groundwater Potential Zones Using Remote Sensing and GIS

Abstract

Groundwater is a significant source of water supply, especially with depleted and quality-deteriorated surface water. The number of drilled boreholes for groundwater has been increased, but erroneous results often occur while selecting sites for digging boreholes. This makes it necessary to follow a science-based method indicating potential zones for groundwater storage. The LithoSFR Model is a systematic approach we built to create an indicative map with various categories for potential groundwater sites. It is based mainly on retrieved geospatial data from satellite images and from available thematic maps, plus borehole data. The geospatial data were systematically manipulated in a GIS with multi-criteria applications. The novelty of this model includes the empirical calculation of the level each controlling factor (i.e., weights and rates), as well as the LithoSFR Model, adopting new factors in its design. This study was applied on a representative Mediterranean region, i.e., Lebanon. Results showed that 44% of the studied region is characterized by a very high to high potentiality for groundwater storage, mainly in areas with fractured and karstified carbonate rocks. The obtained results from the produced map were compared with datasets which were surveyed from representative boreholes to identify the discharge in the dug boreholes, and then to compare them with the potential zones in the produced map The reliability of the produced map exceeded 87%, making it a significant tool to identify potential zones for groundwater investment.

1. Introduction

Groundwater is an invisible source of water that is stored at a depth within rock layers. This makes it necessary to analyze surficial geological and hydrological features in order to identify hydrogeological clues for water flow/storage into the rock stratum below. Searching for groundwater and its investment is often expensive; therefore, digging an unsuccessful borehole (i.e., a dug water well) is a financial loss.

Many regions are suffering from water shortage as a result of low precipitation rates and high potential evapotranspiration. Even it is characterized by slight humidity, Lebanon is a typical example of water shortages in the Mediterranean Region, and recently it became a water-stressed country as a result of climate change and an increased population with new water demands, as well as the unreliable management approaches applied due to the uncertainty in the available hydrologic and climatic data. This has resulted in decrease in the discharge from water sources, plus a deterioration in water quality, whether in surface or groundwater resources. The discharge from rivers and springs in Lebanon has declined by more than 60% over the last five decades, while the rate of snowmelt has increased from 4 months to 2.5 months, besides a lowering in the groundwater table in the major aquifers estimated between 20 and 25 m [1].

Groundwater is widely tapped worldwide, and aquiferous rock formations are subjected to unprecedented exhaustion, of which Lebanon is a typical example [2,3]. In this respect, the number of groundwater boreholes, which are chaotically drilled, has been largely increased in different regions, notably in the agricultural and urbanized ones; in Lebanon, this is averaging to about 500–600 wells/km². For example, the density of the number of boreholes in the suburbs of the capital, Beirut, has been increased from 500 to 1450 wells/km² over the last 20 years, while the number of boreholes has been increased by 3.5 times over the last three decades (i.e., an estimated density of about 200 wells/km²) in the Hermel region (a rural area in North Lebanon). Most of these boreholes are private (and illegal) ones, and the ratio between public and private groundwater boreholes is about 1:25 [1]. Thus, the over-pumping from groundwater in Lebanon has resulted in abrupt depletion in the water table, piezometric level, pumping rate, and discharge from the surrounding springs, as well as most of the coastal aquifers witnessing seawater intrusion and other aspects of contamination.

The lack of hydrological data and uncertainty of field measurements are often major challenges in managing the water sector; therefore, this results in inaccurate calculations in water budgets. This is well pronounced in the contradictory results between different studies. In this regard, the estimation of renewable groundwater volumes in Lebanon sharply varies between different sources. For example, renewable freshwater in Lebanon was estimated at 0.5 billion m³ by the Ministry of Power and Water [4], while it has been mentioned at 3.2 billion m³ by FAO [5], and 4.84 billion m³ by UNDP [6]. There are also contradictions between the percentage of surface water to groundwater, which is often considered as 60% to surface water and 40% to groundwater [7], and a water volume of about 3.65 billion m³ has been estimated in rivers and springs.

Groundwater abstraction from wells is often undetermined and uncontrolled and contradictory estimations are always raised. In Lebanon, for example, the total annual groundwater abstraction from groundwater boreholes is estimated at 0.7 billion m³ [1]. Under normal climatic conditions, it is considered that there is a yearly deficit of 0.2 billion m³ in groundwater, and this in turn does not give a precise number, and it shows the uncertainty of such estimations [8]. The increased population and the new aspects of water demand, as well as the changing climatic conditions, are negatively controlling the balance in water supply/demand, and thus water becomes a comparative commodity. The lack of management controls has resulted in many geo-environmental problems, such as water hazards (e.g., floods, erosion, etc.), while water contamination is a widespread phenomenon. In this respect, surface water resources are most influenced by quality deterioration, and it is rare that a surface water source in Lebanon is found to be pure, while the environmental flow is poor enough. All these problems in surface water make it necessary to invest in groundwater resources.

There are several models (i.e., mathematical, numerical, analytical, and fuzzy) and methods applied to identify potential locations for groundwater storage [9,10,11,12,13,14,15]. Among these models are APLIS, GRACE, FEFLOW, SVFLUX, MODFLOW, etc. Nevertheless, positive results from these models and methods were either not validated by dug boreholes or uncertain results were reached [16,17,18]. This might be attributed to the imprecise weighting given for the integrated factors in the applied models [19,20], or the lack of certain effective factors in these models [21,22,23].

The identified constraints in the investigated models were useful to calibrate a new model after addressing the problems of giving various weights for each factor involved, as well as by adding new factors. Therefore, we built the LithoSFR Model by integrating six controlling factors on GWP which were not integrated together before in one model, and this gives novelty to our study. These factors are: Lithology, Streams’ density, Streams’ connectivity, Slope, Fractures and Recharge rate. Therefore, we studied 34 methods and models related to groundwater storage and recharge in order to identify the factors used and the adopted weights.

For identifying potential sites for groundwater storage, there are many hydrogeological elements that must be primarily identified, and these elements vary between different regions. However, the use of a remote sensing and geographic information system (GIS) becomes significant to acquiring the geospatial data needed for analysis and for the systematic manipulation. Hence, many studies have adopted these techniques, and they have become widely performed in many regions worldwide [13,24,25,26,27,28,29,30,31].

This study aims at producing an indicative groundwater potential (GWP) map based mainly on the digital processing of satellite images in integration with thematic maps and data records from boreholes. This is supported by various GIS applications for geospatial data manipulation and visualization. The produced map shows geographic zones with various probabilities to store groundwater. Even though this study has been applied to a selected region in Lebanon, it can be applied elsewhere in coastal regions dominant with carbonate rocks, mostly in the Mediterranean region. The produced map represents a supportive document for decision makers and water-concerned stakeholders to benefit from it while planning for water management projects.

2. Materials

2.1. Study Area

The selected study area from Lebanon, located along the Eastern Mediterranean, is dominant with carbonate rocks (i.e., limestone and dolomite). This area encompasses a variety of natural characteristics which are identical to the entire coastal regions of the Mediterranean region where it interlinks slightly humid regions with semi-arid and arid regions. This includes the climate, geomorphology, hydrology, and geology. The selected study area is about 504 km², representing about 5% of the Lebanese territory, and it is situated along the southern part of the country, where it extends from the coast to the mountainous regions in the east. It is allocated between the following geographic coordinates (Figure 1): 35°10′05′′ E and 35°28′50′′ E, and 33°06′55′′ N and 33°17′10′′ N.

The average rainfall rate in the study area varies from 650 to 900 mm, and snow cover often occurs between 5 to 10 days a year, while the average temperature ranges between 18 °C in winter and 27 °C in summer. The area is almost a moderately-mountainous region and is usually described as a plateau [32,33], with a narrow coastal ribbon that does not exceed 2 km. It extends from the coast to about 25 km to the east until the highest altitude of about 820 m; therefore, comprising a slope gradient of about 30–35 m/km. The average altitude ranges between 250 and 400 m, and it gradationally increases eastward with no remarkable cliffs, except where valleys span seaward.

There are no river flows within the area of study and the Litani River (i.e., the largest Lebanese rivers) is almost several kilometers to the north of this area. Nevertheless, seasonal streams are widespread and they run-off water during wet seasons, which last for a few months. The geology of the study area is characterized by dominant carbonate rocks (i.e., limestone, dolomitic limestone, and dolomite) that are distributed into four rock formations, where they are often exposed in intervening with argillaceous materials (e.g., marl, marly limestone, and chalky limestone). These rocks are interrupted by many rock deformations, including mainly the strike-slip faults that cross several tens of kilometers from the mountains to the sea and then result in intensive fracture systems (i.e., joints and fissures), as well as rock strata folding.

2.2. Hydrogeology of the Study Area

Groundwater in the Mediterranean region is mainly found in carbonate rocks which are characterized by high permeability due to developed karstification and intensive rock deformations. Four rock formations are exposed in the study area, where the oldest one is from the Middle Cretaceous age (Cenomanian–Turonian) with rock sequences belonging to the carbonate rocks. These rock formations are uniformly overlying on each other. The total thickness of all rock formations in the study area is of about 2300 m. These are of the Quaternary, Tertiary (Eocene), and Upper Cretaceous (Senonian and Cenomanian–Turonian) ages. The geologic structures mainly govern the geographic distribution of these rock lithologies. The hydro-lithological profile of the exposed rock formation can be summarized (from top to bottom) as follows:

• Quaternary deposits: These are detrital materials composed mainly of alluvial deposits and sometimes mixed with rock debris and unconsolidated sediments from the surrounding highlands. They are often found in the low-lands and depressions. These deposits occur also in the coastal plain where sandy and loamy deposits are widespread. This rock formation has neither uniform thickness nor unified lithology.
• Eocene (Lower Tertiary–Lutetian) limestone: It comprises well-bedded carbonate rocks with a maximum 800 m thickness of thin to moderately thick and fractured, jointed, partly karstified beds of Nummulitic limestone with chert nodules, interbedded with marly and chalky limestone. This rock formation is exposed mainly in the NE of the study area, and it is considered as an aquifer due to the dominant spacing between its bedding planes and the common karstification and fractures, but the presence of chalk and marl, in some instances, reduces its aquiferous property [1].
• Senonian (Upper Cretaceous) Marl: This rock formation is composed largely of marl and some marly limestone, changing from massive, jointed, and fractured to soft and friable rocks. In the study area, it is exposed in different localities as elongated patches between the Tertiary and Cretaceous rocks. Nevertheless, the considerable clayey content, with minimal permeability, makes it unfavorable for groundwater flow, whether in the vertical or horizontal direction; therefore, it is considered as an “Aquiclude” [1].
• Cenomanian–Turonian (Middle Cretaceous) Limestone: The Turonian rocks (in the upper part of this formation) have marly limestone facies, have limited exposures, and are usually joining part of the Cenomanian rock formation. The latter is composed mainly of massive, thick to moderate thickness, well-bedded limestone and dolomitic limestone. This rock formation is highly fractured with developed karstic features (e.g., karstic conduits and galleries, etc.). It is well observed in the SW part of the study area. This rock formation is considered as an excellent aquifer, because it is characterized by secondary porosity and high permeability due to the abundant fracture systems, including fissure and multiple-set joints. In addition, karstification is well developed and represented by galleries and subsurface conduits, shafts, and grottos which are mostly filled with groundwater [34]. Therefore, the study area occupies two main aquifers (i.e., Eocene and Cenomanian), where the Cenomanian aquifer is known as the most productive groundwater reservoir not only in the study area but also in the entirety of Lebanon. The hydraulic characteristics of these aquiferous rock formations are shown in

  Table 1. Hydraulic properties of the existing aquifers in the study area (Shaban [1] and other sources).

  Rock Formation	Maximum Thickness *	Hydraulic Properties
  		Porosity (Ø)	Permeability (P) **	Moisture Content (ɱ)	Hydraulic Conductivity (k)	Specific Yield (Sy)	Transmissivity (T)
  	(m)	%	-	%	m/s	%	m2/s
  Lutetian	550	4–6	M-H	7–9	10–5–10 **	14–24	1.2 × 10−6 −1.4 × 10−6
  Cenomanian	700	8–12	H-VH	6–8	10–5–10 **	16–26	1.8 × 10−6 −2.4 × 10−6

  Notes: * Maximum thickness in the study area. ** L = Low, M = Moderate, H = High, VH = Very high.

  . These estimations have been adopted either from field measures or from different sources, e.g., [34,35,36,37].

2.3. Satellite Data Sources

Several materials were used in this study in order to enable the gathering of all geospatial datasets needed and to assure preparing them in a digital format. The majority of the geospatial datasets represent the factors controlling the GWP property. Therefore, determining these factors is a prerequisite step. For this purpose, remote sensing techniques, and more specifically satellite images, were the main source of geospatial data and these techniques have proved their reliability in many studies [38,39,40,41,42,43,44,45,46]. The retrieved data from the processed satellite images were integrated by the analyzed thematic maps as well as the data records and field observation.

The adopted satellite images and their characteristics are shown in

Table 2. Used satellite images in this study, and their specifications.

Satellite	No of Bands	Spatial Resolution	Revisit Time	Swath Width (km)	Year of Retrieve
IKONOS	5	0.82 m	3 days	11.3 × 11.3	May 2021; July 2021,
Aster	14	15 VNIR, 30 m SWIR, 90 m TIR.	16 days	60 × 60	March 2006
Landsat 7 ETM+	8	30 m, 120 m thermal, 15 m pan	16 days	183 × 183	April 2008
SRTM	1	30 m, 90 m	one mission	225 × 233	July 2017

. They were retrieved from the following sources:

-

IKONOS: https://www.satimagingcorp.com/gallery/ikonos/ (accessed on 13 March 2023).

-

Aster: https://asterweb.jpl.nasa.gov/data.asp (accessed on 22 May 2022).

-

Landsat 7 ETM⁺: https://geonarrative.usgs.gov/landsat-7/ (accessed on 8 June 2022).

-

SRTM:https://search.earthdata.nasa.gov/search?q=SRTM&ff=Map%20Imagery&lat=34.14729739882451&long=-3.375 (accessed on 15 February 2023).

The used satellite images encompass a variety of temporal and spatial resolution, as well as an uneven number of spectral bands, including thermal ones. The diversity in the used satellite images was useful to overcome any missing data or difficulty of acquiring certain terrain features, which may not have been obtained if only one type of satellite images was depended upon.

For satellite image processing and GIS applications, two types of software were used. These are

-

ERDAS-Imagine 11 for satellite image processing (Leica Product, Atlanta, GA, USA).

-

Arc-GIS 10.2 for the manipulation of geospatial data (ESRI Product, Environmental System Research Institute, Redlands, CA, USA).

In addition, a number of thematic maps are also used in this study, and they were re-produced in digital form using a GIS system. The used thematic maps in this study are

-

Geological maps of 1:50,000 scale. Sheets of Tyr-Nabatieh and Naqoura-Bent Jbeil [32].

-

Topographic maps of 1:20,000 scale with a 25 m contour interval. Sheets of Tyr, Ma’arakeh, Srifa, Rachidieh, Jouya, Tibnine, Naqoura, Aita Chaab, Bent Jbiel, Rmayesh, Yaroun [46].

-

Recharge potential map of the occidental zone of Lebanon with a 1: 1,000,000 scale [47].

-

Water wells map (collected from different sources and from field surveys, e.g., [6,48,49,50].

3. Methods

The generated LithoSFR Model aims to integrate all influencing factors in a map form, whether they are linear shapes or polygons, where the sum of the impact from these factors is worked in a unified trend towards determining the level of impact on the GWP. In other words, when these factors are integrated together, they will result in the cumulative impact on GWP property (i.e., high, moderate, low, etc.). This requires applying a systematic approach for data analysis, manipulation, and overlapping, and it also requires preparing all factors in a unified digital format. For this purpose, geospatial data retrieved from satellite images were converted into digital forms in the GIS system (i.e., shape-files). Except the lithological characteristics, which were classified into 4 classes in the LithoSFR Model, the other factors were classified into 5 classes, each representing their influence on the GWP (Figure 2). The degree of impact of each factor was described as a “Weight”, and the classes branched from these factors were also given a sub-level of impact and they were attributed to “Rates”. The calculation of weights in this study was based on the following:

• Results of adopted weights used by the authors in previous studies and from other studies done [1,28,32,44,50,51,52,53], where the calculated weights proved their credibility.
• An empirical method which depended mainly on the number of studies that mentioned any of the factors (or relevant elements to those factors). For this purpose, we implemented a detailed survey on the previous studies done on GWP mapping or any relevant groundwater recharge topic, and this is also one of the novelties of the LithoSFR Model. In this survey, 34 previous studies were analyzed and the number of used factors in each study was calculated (

  Table 3. Surveyed previous studies and the matched factors with the factors mentioned in this study.

  #	Author(s)	Mentioned Factors *
  		L	F	Sd	Sc	Ts	Rp
  1	Meisler, H. 1963 [54]		X
  2	Rauch, H; LaRiccia, M. 1978 [37]		X
  3	Taylor, L. 1980 [55]		X
  4	El Shazly et al., 1983 [38]					X
  5	Seelan, K. 1983 [9]	X				X
  6	Salman, A. 1983 [56]			X	X
  7	Ahmed et al., 1984 [57]		X	X
  8	El-Baz, F. 1992 [58]		X	X		X
  	Savane et al., 1996 [59]	X	X
  9	Gustafsson, P. 1994 [25]		X
  10	Teeuw, R. 1995 [10]		X
  11	Per Sandra et al., 1996 [26]	X	X	X
  12
  13	Edet et al., 1998 [60]		X
  14	Robinson et al., 1999 [39]		X	X
  15	Das, D. 2000 [61]	X	X
  16	Murthy, K. 2000 [62]		X	X		X
  17	Bilal, A.; Ammar, O. 2002 [63]	X	X	X
  18	Doll et al., 2002 [52]					X	X
  19	Sener et al., 2005 [40]	X	X
  20	Shaban et al., 2006 [47]	X	X	X
  21	Kumar et al., 2007 [64]	X	X	X		X
  22	Ganapuram et al., 2008 [28]	X	X			X
  23	Joycee, D.; Santhi, M. 2014 [65]	X	X	X
  24	Ali-Ahmad, 2015 [48]	X	X	X	X
  25	Elbeih, S. 2015 [42]	X			X
  26	Dinesan et al., 2015 [66]	X		X		X
  27	Deepa et al., 2016 [67]	X	X	X		X
  28	Senanayake et al., 2016 [30]						X
  29	Arshad et al., 2019 [12]			X		X	X
  30	Nigussie et al., 2019 [43]	X	X	X		X
  31	Andualem, T; Demeke. G. 2019 [68]	X	X	X	X	X
  32	Chenini et al., 2019 [69]	X					X
  33	Aslan andÇelik, 2021 [70]	X	X	X		X
  34	Samida et al., 2022 [45]	X	X	X
  Total		19	25	17	3	13	4

  Notes: * L = lithology, F = fractures, Sd = streams’ density, Sc = streams’ connectivity, Ts = Terrain slope, Rp = Recharge potential.

  ).

In the LithoSFR Model, the total number of the mentioned factors were calculated and considered as the total value (Tv). Consequently, the number of mentions for each factor in the previous studies (Mf) was divided by Tv to deduce the percentage of the weight (Wt) for each factor, as in the formula below. The results are illustrated in

Table 4. Determining factors weight according to previous studies.

Factor	Mentions in the Surveyed Studies (Mf)	Percentage of the Weight (Wt)
Lithology	19	23%
Fractures	25	31%
Streams’ density	17	21%
Streams’ connectivity	3	4%
Terrain slope	13	16%
Recharge rate	4	5%
Sum value (Sv)	81	100%

.

Wt = Mf/Tv.

For the rates of each factor, the percentage of their influence was also calculated by dividing the number of determined classes over 100 (as in Figure 2 and

Table 5. Interpolated percentages of weights and rates in LithoSFR Model for GWP mapping.

Factor	Wt (%)	Classes (Sub-Factor)	Impact on GWP	Rt (%)	Ri (%)	Rip (%)
Lithology (Descriptive)	19	Cenomanian–Turonian Limestone	Very high	25	4.75	5.86
		Eocene Limestone	High	25	4.75	5.86
		Quaternary deposits	Moderate	25	4.75	5.86
		Senonian Marl	Low	25	4.75	5.86
Fractures’ density (Lineaments/25 km2)	25	>26	Very high	20	5	6.17
		26–20	High	20	5	6.17
		19–13	Moderate	20	5	6.17
		12–6	Low	20	5	6.17
		<6	Very low	20	5	6.17
Streams’ density (km/km2)	17	8.62–6.90	Very high	20	3.4	4.19
		6.90–5.17	High	20	3.4	4.19
		5.17–3.44	Moderate	20	3.4	4.19
		3.44–1.[	Low	20	3.4	4.19
		<1.72	Very low	20	3.4	4.19
Streams’ connectivity (connecting points/km2)	3	4.92–3.15	Very high	20	0.6	0.74
		3.15–2.5	High	20	0.6	0.74
		2.5–2.03	Moderate	20	0.6	0.74
		2.03–1.06	Low	20	0.6	0.74
		<1.6	Very low	20	0.6	0.74
Terrain slope (Degree °)	13	>35	Very high	20	2.6	3.20
		35–20	High	20	2.6	3.20
		20–10	Moderate	20	2.6	3.20
		10–5	Low	20	2.6	3.20
		<5	Very low	20	2.6	3.20
Recharge rate (%)	4	50–45	Very high	20	0.8	0.99
		35–30	High	20	0.8	0.99
		20–10	Moderate	20	0.8	0.99
		10–5	Low	20	0.8	0.99
		<5	Very low	20	0.8	0.99
Sum					81	100

). The resulting weights and rates were multiplied to calculate the ratio of impact (R_i) for each factor; this can be achieved using the following formula, which is illustrated in Table 3:

Ratio of impact = % of factor influence × [% of factor sub-influence-1 + % of sub-influence-2 impact +...]

R_i = % Wt × [% sub-influence-1 + % sub-influence-2 +…]

The ratio of impact R_i for each factor was converted into a percentage (R_ip) as shown in Table 5; thus, the resultant values (R_ip) were digitally represented and, therefore, systematically emerged in Arc-GIS to produce the final GWP map for the area of study.

4. Results and Validation

The LithoSFR Model was applied in a representative coastal Mediterranean region which encompasses similar natural conditions, including the climate, geology, and geomorphology, with a special emphasis on distinguished elements, including the hydrogeology of coastal zones, developed karstification, and groundwater over-pumping, which are common in the entire Mediterranean region. Thus, the analysis of the selected region from Lebanon was considered to be a pilot, from where the LithoSFR Model can then be adopted in other regions.

Groundwater storage is a hydrogeological process that occurs in various aspects within rock voids, spacing, and fractures where they are characterized by a diverse volume of stored water at depths. There are many factors controlling groundwater storage in rocks, governed by the natural characteristics of the studied area. In other words, some factors do not appear in an area, but they are developed in other areas. However, the lithological characteristics are always counted as a significant element in groundwater flow/storage regimes.

In this respect, there are many studies applied for mapping groundwater potential zones, but not all of these studies used similar factors to be modelled, and these factors are given different levels of impact on groundwater storage, which is described as “weights”. For example, Per Sander at al. [25] integrated the factors of elevation, drainage, lineaments, slope, lithology, and soil, which are typical for GWP mapping; however, elevation and slope can be interlinked as one factor and they might act with similar impacts on GWP. Meanwhile, Ganapuram et al. [28] considered other factors, including the hydrogeomorphology, geology, geologic structures, drainage, slope, and land use/cover; however, the hydrogeomorphology must represent drainage and slope and they are not necessarily accounted for separately. Further, Smida et al. [30] involved only the lineaments (i.e., fractures) and geology, and these two factors represent only the geological controls on the GWP, where they must be supported by other acting factors. The selection of the controlling factors is significant to reaching accurate results. One novelty in this study is the selection of factors, which was considered to include all elements governing water percolation (i.e., recharge) into the substratum, groundwater flow, and storage into deep rocks. Therefore, the following six factors were selected in this study, but they were not used together in previous studies.

4.1. Geospatial Data for the LithoSFR Model

1.

Lithology

The lithological characteristics of the exposed rocks have a major role in groundwater recharge, where they act in two aspects; first is the percolation of surface water into the rock layers beneath, and second is the ability to store groundwater into rock voids, spacing, and fissures and between bedding planes. Therefore, all previous studies, such as El-Baz and Himida [51], accounted for rock lithology in mapping or determining GWP zones. This is also the case for studying recharge potential zones, where lithological characteristics are often considered as the main acting factor [52], and the identification of recharge potential zones is often a primary step for the mapping of groundwater storage. For the study area, the identification of lithologies was done using the available geological maps, e.g., [32,33]. The lithological distribution, as polygons, were digitized using Arc-GIS 10.5 software. Four classes were assigned according to the existing rock formations, which are attributed to Quaternary deposits, Eocene Limestone, Senonian Marl and Cenomanian–Turonian Limestone (Figure 3).

2.

Fractures

Rock deformations, including mainly fractures, are one of the most significant hydrogeological clues used while searching for groundwater resources, and they have been considered in many studies, e.g., [32]. Notably, fractures are able to create spacing into rocks and thus increase porosity and permeability. Fractures are well developed in consolidated rocks, such as the carbonate rocks which exist in Lebanon, and characterized by secondary porosity that permits water to easily percolate and then store in these multiple-set fissures. In this respect, fracture systems may exist locally (in a limited area) and result in high-infiltration domains, or as they can be one-set elongated fractures that span laterally for long distances (i.e., several kilometers) and cut vertically into different rocks until impermeable layers exist and then groundwater accumulates. Groundwater-bearing zones are always associated with high fracture systems. This has been proved in many studies, such as by Shaban et al. [26].

In this study, fractures were traced to generate the “lineament map”, where lineaments are defined as linear geological fractures that can be observed/detected mainly in satellite images and aerial photos. The density of lineaments is significant to produce GWP maps, and this has been commonly adopted in several studies, e.g., [26,43,44,58].

In this study, the extraction of geologic-related linear features (i.e., lineaments) was done using Landsat 7 ETM⁺ satellite images, where the thermal band (i.e., band No. 6) with a 120 m × 120 m spatial resolution was useful for lineament detection. For this purpose, ERDAS-Imagine 11 software was used, since it has advantages in detecting linear features. These advantages include its band combination (e.g., bands 2, 5, and 3), edge detection, color slicing, filtering, etc. The application of these digital advantages enabled the detection of all elongated fractures (i.e., faults) which are filled with soil and loss sediments, and thus showed higher witness than the surrounding rocks. Thus, thermal bands can identify these fractures, from which the lineament map can be then produced (Figure 4).

The representation of lineaments’ impact on the GWP was performed by the density of lineaments, which was projected from the vector shapes of the linear features in the satellite images and then converted into raster shapes. The concept of performing the density of lineaments followed the “Sliding Windows” method which was done by Shaban et al. [47] and Shaban et al., 2007 [26]. It can be applied by classifying the area of interest into frames and then calculating the number of lineaments in each frame; therefore, the average value of the lineaments in each of the four adjacent frames will be considered as an indicative value to build the contour map representing different domains of lineament density.

In the GIS system, the “Sliding Windows” method was obtained by constructing a circle with a 20-pixel radius adopted in each raster, and the number of pixels belonging to a fracture in each circle was calculated. Therefore, a lineament density map with five classes was generated in this study, where contour lines with different domains of lineament density were produced (Figure 5). These classes were divided as >26, 26–20, 19–13, 12–6, and <6 lineament per each 25 km² for the very high, high, moderate, low, and very low lineament density categories, respectively. The selection of 25 km² areas was attributed to the total surface area of the study.

3.

Streams’ density

Streams, as dominant geomorphological and hydrological features, control the horizontal flow and run-off, and partially the vertical flow (i.e., infiltration), of surface water. However, their morphometric characteristics are always considered as an indicative element for groundwater storage into the underlying rocks. In geomorphology, stream patterns are defined as structural and lithological terrain features, while the stream density reflects the infiltration (i.e., recharge) rate of water from the surface to the subsurface rocks. Hence, the denser the streams, the less infiltration, and vice versa.

The extraction of streams in this study was done from Digital Elevation Models (DEMs), while topographic maps (1: 20,000 scale) were also used to assure the names and orientation of the existing streams. For this purpose, SRTM-DEM (30 m mesh size) was used, and Triangulated Irregulated Networks (TINs) were constructed using the Spatial Analyst in Toolbox of Arc-GIS to automatically generate streams in the study area (Figure 6). The streams’ density map was obtained by following the same approach applied for lineaments’ density (i.e., Sliding Windows); therefore, five classes were produced, as follows: 8.62–6.95, 6.95–5.17, 5.17–3.44, 3.44–1.72, and <1.72 km/km² for the very high, high, moderate, low, and very low stream density categories (Figure 7).

4.

Streams’ connectivity

This factor is a novelty of the LithoSFR Model. According to Herron and Wilson [59], the property of hydrologic connectivity mainly refers to the efficiency of run-off movement through a basin. The joining between two or more streams has never been considered in the previous hydrological studies applied to determine potential zones for groundwater, while emphasis must be shed on the difference between a region with dense stream connectivity (i.e., whether confluences or diversions) and another region with low stream connectivity. Denser connection between streams evidences a lower infiltration rate, and vice versa. The connectivity of streams is represented by the number of points that connect one or more streams together, whether they are confluences or diversions.

The connectivity of the streams was adopted from the stream map (Figure 5), while their density map was obtained by plotting the points on the contacts between different streams, and then using the same method used for lineament and stream density; thus, a connectivity density map was produced (Figure 8). Five classes were obtained, as follows: 4.92–3.15, 3.15–2.5, 2.5–2.03, 2.03–1.6 and <1.6 connectivity point/km² for the very high, high, moderate, low, and very low stream density categories, respectively.

5.

Slope

Terrain slope has been tackled in few studies about potential groundwater storage; especially those that studied areas that were not flat. However, slope is a function of the run-off velocity and the overland flow; therefore, a steep slope reduces the rate of groundwater recharge [58]. Along steep sloping lands, the water flow energy will be high and this does not give sufficient time for water to percolate into the phreatic zones, while along gentle slopes, the flow rate of surface water is lower, allowing sufficient time for it to infiltrate into the rocks beneath, and then rock reservoirs can be recharged with more water.

The generation of the slope map was done using SRTM (Shuttle Radar Topography Mission), with a one-arc second spatial resolution (30 m). Arc-GIS was used to determine and draw the slope map (Figure 9). The five resulting classes for terrain slope were as follows: >35, 35–20, 20–10, 20–10, 10–5, and <5 degree for the very high, high, moderate, low, and very low slope categories, respectively.

6.

Recharge rate

This factor is a novelty of the LithoSFR Model. It is a supportive factor from previously measured/assessed recharge rates which evidences the hydrogeological regime of water flow to the underlying rock lithologies. It has been rarely used in previously applied models and methods. In this respect, the involvement of this factor is helpful to support the manipulation of the adopted factors for groundwater storage mapping. A groundwater recharge potential map for the entirety of Occidental (i.e., western) Lebanon is available at a scale of 1:1000.000 [32]; therefore, it was used for this study. It encompasses five classes (from very high to very low potentiality for the recharge property) representing the percentage of recharged water resulting from precipitated water. Therefore, sub-setting of the recharge potential zones from the available map was performed to localize the study area using Arc-GIS (Figure 10).

According to Shaban et al. [47], there are five classes dividing the study area in terms of recharge potential rate, which represents the percentage of infiltrated water from precipitation. These are very high (50–45%), high (35–30%), moderate (20–10%), low (10–5%), and very low (<5%).

4.2. LithoSFR-Based GWP Map

The retrieval and manipulation of the geospatial data required for a creditable model entailed determining various controlling factors on groundwater flow/storage, which were integrated in a systematic approach to model these factors. The produced GWP map shows the distribution of different GWP zones, and each zone was descriptively assigned (Figure 11). There were some limitations raised while producing the LithoSFR Model; in addition to the determination of weights, there was the selection of the influencing factors, especially factors that were not known in previous studies, such as streams’ density and connectivity, as well as the recharge rate. However, our experience in similar applications on groundwater assessment indicated it was of utmost significance to include these factors. Moreover, the integration of linear shapes with polygons was another limitation, but this was also addressed by converting the density of linear shapes into polygons by constructing a circle with a 20-pixel radius which was adopted in each raster; from this, the number of pixels belonging to a fracture in each circle was calculated. Results show that the study area, which is a typical Mediterranean coastal region, is characterized by two equal percentages of GWP properties. Thus, about 44% of this area encompasses zones with high to very high GWP, alongside 42% of low to very low GWP (

Table 6. Groundwater potential zones for the study area, as resulting from the LithoSFR Model.

GWP	Area (km2)	Percentage of the Total Area	Remarks
Very high	57.2	11.34%	Very high-GWP zones exist as elongated geographic patches on the Cenomanian rock formation, and they obviously correspond to several geologic structures and also the fault alignments.
High	165.1	32.75%	These zones are almost coincident with the geographic distribution of the Cenomanian rock formation which extends through the middle and lower parts of the study area.
Moderate	69.3	13.75%	These GWP zones are located mainly between the high- and low-GWP zones, and they do not have a defined geographic distribution, but they often sharply exist with straight lines, evidencing their control with the geologic structural controls.
Low	109.9	21.80%	The low-GWP zones are situated mainly in the eastern and western flanks of the study area, but they are totally absent in the lower part of it where fractured and karstified carbonate rocks are present.
Very low	102.5	20.33%	These GWP zones exist as elongated stretches in the eastern part of the study areas, and they are mainly situated along the coastal zone to the west where the argillaceous materials (e.g., marls) are mixed with limestone rocks.

). The map shows that high-GWP zones are mostly located in the middle and lower parts of the study area, where limestone and dolomitic limestone of the Cenomanian rock formation are situated and characterized by developed fracture systems and well-developed karstification. Moreover, the geographic distribution of high-GWP zones is clearly trending in the NE–SW direction, since they are controlled by the diagonal strike-slip faults that align with the geographic distribution of the limestone rock masses.

There was a constraint to apply a comparative analysis with other existing groundwater potential mapping, because the studies done in this respect are rarely validated and it would be difficult to assure their reliability, while the studies with validated results and even lower numbers of integrated factors were used and showed favorable coincidence with our study [47,48,51].

The geographic distribution of different GWP zones, as they are plotted on the produced map, is rational because the majority of high- and very high-GWP zones are largely found in the exposed carbonate rocks, where 53% are located in the Cenomanian limestone rock formation and 39% in the Eocene marly and chalky limestone rock formation, and both are known as major aquiferous rock formations in Lebanon [1]. In addition, the general orientation of high-GWP zones in the NE–SW direction reflects the control of the existing sets of strike-slip faults which are dominant in the middle and southern part of Lebanon [32]. However, validation of the LithoSFR Model’s results is significant to assuring the credibility of the adopted work, whether for the used tools of analysis or the adopted methodologies.

For this purpose, data from dug groundwater boreholes were surveyed specifically for this study, in order to project different water boreholes with their productivity on the different GWP zones; and then to verify the coincidence between the productive wells and zones with high and very high GWP (Figure 11). The data on the productivity of water wells were reported in l/s. In this regard, 64 groundwater boreholes were surveyed in the study area, and the relevant data were illustrated in the Supplementary (Table S1) which shows the location (and coordinates) of boreholes, their productivity, and the belonging rock formations. The number of surveyed boreholes was reasonable, because these boreholes were selected from diverse localities based on different altitudes, proximity from the coast, hydro-stratigraphic characteristics, water table depth, and pumping rates.

For water productivity (i.e., measured discharge rate) in the surveyed boreholes, it was assigned into descriptive categories in order to classify the reported productivity into specific ranges. The highest and lowest reported discharge rates from water wells in the study areas were considered for this categorization, and then the following categorization was adopted:

-

Discharge rate greater than 12 L/s = Very high productivity.

-

Discharge rate between 12 and 8 L/s = High productivity.

-

Discharge rate between 8 and 4 L/s = Moderate productivity.

-

Discharge rate between 4 and 1 L/s = Low productivity.

-

Discharge rate less than 1 L/s = Very low productivity.

4.3. Results Validation

Using Arc-GIS to apply the projection of groundwater boreholes on different GWP zones (Figure 11) shows that the majority of the productive wells (i.e., very high and high discharge rate) coincided with 87.5% of the very high- and high-GWP potentiality, as shown in the produced map (Figure 11), and this evidences the credibility of the LithoSFR Model in this study. In addition, linear regression was also applied to figure out the relationship between the two variables (i.e., location of groundwater boreholes and high-GWP zones) by fitting a linear equation with the measured datasets. Hence, the actual discharge (L/s) was plotted on the Y axis, while the mapped productivity classes were plotted along the X axis (Figure 12). It is obvious that the discharge increases linearly within the high- and very high-GWP zones with a calculated ration of R² = 0.81. This clearly evidences the credibility of the applied model, which can be adopted for mapping GWP zones in areas where a carbonate rock stratum exists, like in the case of Lebanon.

5. Discussion

Groundwater management is usually tackled on the national level in regions under water stress. This includes the exploration of new resources and adaptive and conservation measures. The selected area for this study encompasses coastal and mountainous zones, and it is characterized by diverse hydrology and geomorphology, representing a typical Mediterranean zone. With the exacerbation of various geo-environmental problems, especially the pollution of surface water resources, concerns have been diverted to groundwater resources. According to Shaban (2020) [1], the increased pumping from groundwater reservoirs has been reported in many studies since the beginning of 1985. Therefore, the percentage of groundwater pumping to surface water use in Lebanon has been raised from about 17% to about 47%. Recently, this percentage has been slightly reduced after a governmental control, represented by an obligatory license for groundwater drilling as imposed by the Ministry of Power and Water. Nevertheless, the dramatic increase in the number of private (and illegal) water wells is still witnessed, and chaotic groundwater pumping has been widened, representing one of the main challenges on water resources’ management. This is also hindered by the lack of data from the private water wells, which represent more than 80% of wells in Lebanon.

The developed karstification elements associated with intensive rock deformations (e.g., fracture systems, open conduits, etc.) are always misleading while exploring sites for potential groundwater. This is well pronounced in many coastal zones of the Mediterranean region, where karstification is dominant with the intensive rock deformations. Therefore, the erroneous selection of sites for digging water wells has become a challenge, notably when the dug boreholes are found empty, and this results in unfavorable socioeconomic conditions; in particular, the failure to reach groundwater reservoirs exhausts financial resources. Therefore, following science-based approaches for groundwater exploration becomes a must.

The impact of the adopted factors was critically evaluated based on our previous knowledge and on the studies applied before. Therefore, lithology is a main controlling factor, where its characteristics control the behavior of groundwater storage and flow among diverse rock stratum, and this has been evidenced in all surveyed studies, while fractures are significant to enlarge the spacing into rocks and then to enable these rocks to store groundwater, and this has been well demonstrated in groundwater mapping. The density of streams and the connectivity between these streams are a function of water infiltration into subsurface rock layers. From the slope, a factor was considered where sloping terrain accelerates the overland lows and reduces the infiltration that feeds groundwater aquifers; whereas the availability of a groundwater recharge potential zone was an added value to support the criteria of water infiltration into the underlying rock layers.

A credible approach is needed for the successful positioning of potential groundwater localities, and there are several hydrogeological clues used to identify GWP sites, such as the proximity to fractures [27] or the boundary aspects of geological rock formations [1]. However, these hydrogeological clues cannot be determined by unexperienced people (e.g., farmers, decision makers, etc.). For this reason, a creditable model for mapping GWP zones is required, and it will represent an easy way to locate promising localities for groundwater storage. The produced indicative document in this study (i.e., groundwater potential map) can be easily used by different levels of stakeholders, including grassroots, expert, and governmental stakeholders, and more specifically the decision makers who can use this map in groundwater management plans.

There are some limitations raised while implementing this study, and it was constrained mainly by the acquisition of data from dug water wells, notably from the private wells where owners of these wells do not cooperate in providing data on their wells, if they exist. Hence, the issue was overcome by using the supported datasets from the Ministry of Power and Water, who provided the study with data logs for several dug wells. Another major limitation for performing this study was the abrupt fluctuations in the discharge in some of the surveyed wells, and this was attributed to the continuous pumping of groundwater. Therefore, detailed well logging was applied when a steady state of discharge was assured.

6. Conclusions

Based on the applied LithoSFR Model, a GWP map was produced to visualize the potential geographic localities for groundwater storage in the selected typical Mediterranean area. This was determined by the integration of several controlling factors, each with its own weight and rate of influence on the storage process. Determining the numeric values of these weights was a major challenge in this study, as well as in many previous studies, but that was addressed based on the long experience of the authors in this field [1,29,44,47,50], as well as we utilization of the weights determined by several applied studies on similar topic.

The validation of the LithoSFR Model was implemented by assessing the coincidence between the geographic distribution of potential groundwater zones resulting from the produced map (Figure 11) and the measured water productivity in existing (operational) dug wells. The results were promising, with 87.5% coincidence. This was also supported by the linear regression, which also showed a coincidence ratio of 81%. The results of this study have a substantial consistency with previous applied studies done by different authors, including [33,34,36,71,72]. However, the results of this study proved to be more accurate, exceeding 87%, and this was not reached by the previously done studies. Therefore, the GWP map produced from the LithoSFR Model is a successful example of a supportive document that can be used for successful groundwater positioning. It has the advantage of easy reading and understanding for the symbolized elements on the map. It can be performed with a smaller scale to cover the entire Lebanese territory and many other coastal zones in the Mediterranean region.