Cherif Kessar,Yamina Benkesmia,Bilal Blissag,Lahsen Wahib Kébir

Agence Spatiale Algérienne,Centre des Techniques Spatiales,Arzew,Algérie.

Abstract:Sustainable management of groundwater resources has now become an obligation,especially in arid and semi-arid regions given the socio-economic importance of this resource.The optimization in zoning for groundwater exploitation helps in planning and managing groundwater supply works such as boreholes and wells in the catchment.The objective of this study is to use remote sensing and GIS-based Analytical Hierarchy Process (AHP) techniques to evaluate the groundwater potential of Wadi Saida Watershed.Spatial analysis such as geostatistics was also used to validate results and ensure more accuracy.Through the GIS tools and remote sensing technique,earth observation data were converted into thematic layers such as lineament density,geology,drainage density,slope,land use and rainfall,which were combined to delineate groundwater potential zones.Based on their respective impact on groundwater potential,the AHP approach was adopted to assign weights on multi-influencing factors.These results will enable decision-makers to optimize hydrogeological exploration in large-scale catchment areas and map areas.According to the results,the southern part of the Wadi Saida Watershed is characterized as a higher groundwater potential area,where 32%of the total surface area falls in the excellent and good class of groundwater potential. The validation process revealed a 71% agreement between the estimated and actual yield of the existing boreholes in the study area.

Keywords:Wadi Saida Watershed;Groundwater potential mapping;Remote Sensing;GIS;Analytical Hierarchical Process;Geostatistic

Introduction

Groundwater is a dynamic and important replenishable natural freshwater resource (Shekhar and Pandey,2014).Globally,groundwater represents one-third of all freshwater exploitation,providing an expected 36%,42%,and 27% of the water used for domestic,agricultural,and industrial purposes respectively (Döellet al.2012).Demand for freshwater resources in the world is critically increasing as a result of rapid growth of industrial activities and population.Consequently,proper groundwater investigation and extraction is now an important part of groundwater management(Senanayakeet al.2015).The United Nations Environment Program estimates that more than two billion people will live with high water stress by 2050,which would limit the development of many countries around the world (Sekar and Randhir,2007).In the last century,global water use has increased six times and continues to grow by 1% every year attaining 30% by 2050,which is responding to excessive demand related to population growth and economic development(WWDR,2019).According to Fieldet al.(2014),by 2020,between 75 million and 250 million people in Africa would be exposed to increased water stress,and in some regions,rainfed agriculture yields would be reduced by 50%.Access to food could be severely compromised in such situation.

In Algeria,where the population was estimated at 46 million in 2020,water requirements were estimated at 5 billion m3/a.The demand for freshwater grows every year by 4% to 5%,while current mobilization is barely 2 billion m3/a (Kettab,2001).

Defined as the amount of groundwater available in an area,groundwater potential is dependent on several hydrological and hydrogeological factors(Gdouraet al.2015).At a different spatial scale,it has become necessary to utilize appropriate new techniques for prospecting groundwater(Le Pageet al.2012).Geophysical prospecting and experimental techniques are among the best-known techniques for achieving this goal.However,proper water resource management requires the application of new approaches for decision supporting.Due to the strict policies on groundwater management in semi-arid regions and the high cost of investigations regarding the development of new boreholes,new approaches have been developed and widely used with data availability and result validation.Countless researchers around the world have conducted studies on the applications of remote sensing and GIS for the exploration of groundwater potential areas,and it has been found that the determination of these areas is influenced by the various factors,which needs a consequent validation (Mageshet al.2012)The results are evaluated based on field review and change of geo-environmental conditions.Currently,AHP techniques based on GIS and remote sensing is the most popular approach in mapping groundwater potential areas and several such studies have been conducted in the past years (Pankajet al.2006;Ould Cherif Ahmedet al.2008;Chowdhuryet al.2009;Suja Rose and Krishnan,2009;Jhaet al.2010;Adiatet al.2012;Alaa and Ayseer,2015;Hachemet al.2015;Rajasekharet al.2019;Nithyaaet al.2019;Gebruet al.2020;Ghoshet al.2020).The availability of spatial,spectral and temporal remote sensing data provides immediate convenience for studies over large and relatively inaccessible areas.Remote sensing has,therefore,become an adequate method for assessing,monitoring and conserving groundwater resources (Balasubramani,2018).Such operational challenges remain,until now,the main hindrance for the authorities to make decision on a proper location for groundwater extraction,and solving such difficulties is the main aim of this research,for which a multi-criteria decision-making tool (MCDM) such as analytical hierarchical method (AHP) is employed (Kessaret al.2020).

The AHP,developed by Saaty (1980),offers a strong and readily understood way of evaluating complicated problems (Alaa and Ayser,2015),to ensures the optimal use of groundwater resources and sustainable management of this wealth.This research aims to build a groundwater potential map of the Wadi Saida Watershed,northwest of Algeria using integrated RS,GIS and AHP techniques.The contribution of this study focuses on the utilization of spatial analysis techniques to visualize,validate and propose areas with good groundwater potential.The heterogeneity of the data from the sources,the hierarchical process and the validation of results using geostatistical analysis of existing boreholes are considered as excellent contributions if the field data are available.

1 Study area

The watershed of Wadi Saida (coded 11-11)is geographically situated between X=223 110,X=250 600 and Y=3 889 100,Y=3 844 370(Coordinate system:WGS 84-UTM).It is part of the larger Macta basin,which stretches northwest of Algeria,in the last tabular uplands of the southern side of the Tellian Atlas,composed by the Tlemcen,Daïa and Saida Mountains (Kessaret al.2020),and on the edge of the high steppe plains (Fig.1).With a surface area of 624.66 km2,it is located between the ends of the mountains of Daïa in the north and the highlands region in the south.It is limited from the West side by;the Sidi Ahmed Zeggaï Mountain,the Sidi Abdelkader Mountain to the south,the Saida Mountain to the east and the Djebel Tiffrit Mountain,which peak at 1 200 m.The Wadi Saida Watershed is characterized by semi-arid climate,during the year by two major seasons are observed.A cold season characterized by an average minimum temperature values of 10°C (between December and February),seasonal temperatures drop sometimes below 0°C,resulting in the occurrence of frost.The average monthly temperature (Saida Station) is high between June and September with a maximum of 27.5°C observed in July and August.The average annual potential evaporation is 835 mm (Yles,2014).The wadi Saida Watershed is subject to the influence of two opposite seasonal regimes.The first is dominated by Mediterranean climate with marine influence causing heavy rainfall in winter.The second is characterized by the storms in the summer season.The analysis of the rainfall series from five available gauging stations (1975-2014)shows that 1997 was the rainiest year with an annual rainfall of 516.12 mm,while the lowest is 199.24 mm in the year of 1999.

The watershed is well drained by a fairly dense hydrographic network,small tributaries arise from high areas and rainwater collects in the valley where it is crossed by the main wadi of Saida.It extends from the north end of the city of Saida,crosses the cities of Rebahia,Sidi Amar and Sidi Boubkeur to arrive at Wadi Taria in Mascara province.The main tributaries are Wadi Massil in the east of Hammam Rabi city,Wadi Nazreg in the east of Rebahia city and Wadi Tebouda crossing Ain el Hadjar in the south of Saida province (Fig.1).Overall,the surface flow generally follows a northward direction which has an outlet.

The wadi Saida Basin has a complex and diversified geology (Medjber and Berkane,2016).It is part of the Tlemcenian domain and a subset of the atlasic domain (Fig.3).The Jurassic formations are the dominant in the study area,in which the Lower Jurassic consist of dolomitized limestones and marls and the Upper Jurassic consists of clays and sandstone limestones.Dolomitic rocks and limestones are generally highly karstified.In the depressions as well as the valleys and wadi beds,lands of continental origin (Fluviatile and Aeolians)of tertiary age are often found and Mio-Pliocene and Quaternary are mostly undifferentiated.

Fig.1 Location map of the study area (Kessar et al .2020)

2 Materials and method

2.1 Thematic map generation

In this study,six components were used to generate the groundwater potential zone map(geology,slope,lineament density,drainage density,land use and rainfall).The methodology is presented in the following flowchart (Fig.2).For the Geology map,we have used a digitized map extracted from the 1/200 000 geological map of National Agency of Hydraulic Resources (NAHR,2008).The lineament density map represents the total length of all recorded and inferred lineaments divided by the surface area under study (Edetet al.1998).An extraction and analysis process is required to determine the linear network before producing the density map.The lineaments network delineated in this study is captured from Landsat 8 (OLI),The Operational Land Imager/ (TIRS),Thermal Infrared Sensor image(LC81970362018264LGN00) and SRTM Global 1 arc-second (The Shuttle Radar Topography Mission (SRTM) Collection User Guide,2015;Shuttle Radar Topography Global Mission 1arc second V003;https://earthexplorer.usgs.gov/),processed by image enhancement and directional filtering (Fig.4 and Fig.5).The lithological and structural discontinuities corresponding to structural lineaments were captured on screen by visual analysis.This methodology allowed us to extract lineaments from the study area(Kanohinet al.2012).To interpret and validate the structural lineaments map,we have used the 1/200 000 geological map of the National Agency of Hydraulic Resources (NAHR,2008) and the 1/50 000 geological map of the NAGMC (National Agency for Geology and Mining Control) (Saprikhine and Riabenko,1978).After these steps,the lineament density map was produced using the following equation:

Where:liis the total lineament length (km) andAis the area of the grid (km2).The density was classified into different intervals (Table 4).For the highest lineament density interval,the highestranking was designated (Pankaj and Amit,2006).Depending on the understanding of its capacity to promote the occurrence of groundwater,a rank from 1 to 5 is allowed to each lineament buffer zone (Riloet al.2019).

Streams in the study area are regarded as linear features and displayed as drainage pattern (Adiatet al.2012).To convert the drainage pattern to measurable quantity,Drainage density (Dd) was derived from the drainage pattern by adopting steps similar to those described by Grennbaum(1985).

Where:Diis the total stream length (km) and A is the area of the grid (km2).A low infiltration rate is usually due to high runoff with high drainage density of the area,whereas,an area with low drainage density thus represents low runoff and high infiltration (Prasadet al.2008).

For the rainfall map,the yearly average rainfall of 38 years on five stations from the National Agency of Hydraulic Resources was used(NAHR,2014);rainfall map was then created using inverse distance weight (IDW) interpolation,(Leroux R,2007),which is widely used by earth scientists (Barier and Keller,1996) and is one of the most commonly used functions in the spatial data interpolation (Hanquiezet al.2014).This interpolation method is particularly suitable for variable and scattered data such as data from geological surveys or environmental measurements(Huisman and Rolf,2009).The method is based on the assumption that the rate of correlation and similarity between neighbouring points is proportional to the distance between them,which can be defined as an inverse function of the distance of each point to neighbouring points.Regarding the factors considered important in this method,the neighbouring radius and the power associated with the inverse distance function can be included in the following equation:

Where:Z0is the estimated value of variable Z at point i,Ziis the sample value at point i,diis the distance between the sample point and the estimated point,n is the coefficient that determines weight based on distance and N is the total number of predictions for each validation case.From 30 m resolution SRTM GL1 (Shuttle Radar Topography Global Mission 1arc second V003),the slope map was generated using the spatial analyst tool in ArcGIS 10.4 software.

Finally,the land use types of the basin were derived from the map generated by the National Institute of Soils,Irrigation and Drainage (NISID,2009).All of these thematic layers are converted to raster file with the same spatial resolution of (30 m) and integrated into ArcGIS 10.4 software to identify groundwater potential zones.

Fig.2 Flowchart of methodology adopted for delineation of groundwater potential zones

2.2 Assignment and normalization of weights

Analytical Hierarchy Process (AHP) is an organized strategy utilized to investigate complex problems,where an extensive number of interrelated targets or criteria are included (Saaty,1980).According to the field experience and expert evaluation,the weights are assigned and normalized for different thematic maps using Saaty AHP concept (Saaty,1980).The subjectivity associated with the assigned weight to different thematic maps as well as their characteristics decreases with the standardization process.Saaty(1980) suggested the calculation of consistency ratio (CR) using the following steps;

Step 1:Principal eigenvalue (λ),is computed by the eigenvector technique.

Step 2:Consistency Index (CI),is calculated from the following equation (Saaty,1980):

Where:fis the number of criteria or factors.

Step 3:Finally,consistency ratio (CR),is calculated as (Saaty,1980):

Where:RIis the random indexes whose values depend on the order of the matrix.In our case,a matrix of 6 factors or criteria are used and the RI values are adopted from Alonso and Lamata (2006)(Table 1).The value of RI was obtained from the Saaty scale from 1 to 9.The value of CR should be less than 10% (Prasadet al.2008) for consistent weights.After calculation using an AHP Excel template with multiple inputs and five decisionmakers analysis (Klaus,2013),the consistency ratio CR was 3.3%,then the relative weights can be re-evaluated to prevent confusion (Shekhar and Pandey,2014).A denotative 9-point scale was used for the prioritization operation and to generate the pair-wise comparison matrix,where 1,3,5,7 and 9 represent important,moderately important,strongly important,very strongly important and extremely important (Table 2).The intermediate values 2,4,6 and 8 could be used when compromise is needed (Gdouraet al.2015).For each previous layer a new rate is given to each category class and this rate indicates the ranges of groundwater potential associated with each factor.Rates were assigned to each class according to the order of the influence of the class on groundwater potential (Table 4),ratings of 1 to 5 were adopted,representing very low,low,medium,high and very high groundwater potential (Kumaret al.2014).

Table 1 Random indices for matrices of various sizes (Alonso and Lamata, 2006)

Table 2 Calculation of effects and rates of factors affecting groundwater potentiality (Saaty, 1980)

2.3 Delineation of groundwater potential zones

For the delineation of the groundwater potential zones,the groundwater potential index (GWPI)was used.Thematic maps from various sources were collected,including remote sensing and conventional data.All the themes and features of the study area were considered and integrated(Kessaret al.2020).To estimate the groundwater potential index (GWPI) (Rao and Briz-Kishore,1991;Malczewski,1999;Shekhar and Pandey,2014;Rahmatiet al.2015;Maity and Mandal,2017;Nithyaaet al.2019),a weighted linear combination method was used as follows:

Where:Wirepresents the normalized weight of thei-th thematic layer,Xjis the rank value of each class of thejlayer,mis the total number of thematic layers,andnis the sum number of classes in the thematic layer.In this study,GWPI is calculated as:

Where:Ldwis the weight of lineament density andLdrrepresents its attributed rank;Ddwis the weight of drainage density andDdrepresents its attributed rank;Swis the weight of slope andSrrepresents its attributed rank;Rwis the weight of rainfall andRrrepresents its attributed rank;Luwis the weight of land use andLurrepresents its attributed rank;Gwrepresents the weight of geology classes andGris its attributed rank.The Raster Calculator tool allows us to create and run the map algebra expression of (GWPI) to generate the groundwater potential map.After assigning the rank to the rating classes,each thematic raster is multiplied by the normalized weights and the result is saved as a new raster.The addition of these resulting raster layers will provide us with the final map of groundwater potential zones.

2.4 Validation using geostatistic analysis

Geostatistics is a branch of statistics which studies the spatial connections of datasets and variables in order to discover and analyze any quantifiable phenomenon.Geostatistical study is based on the theory of spatial or spatiotemporal variables,and its primary objective is to highlight,when it exists,the spatial structuring of the phenomenon and to estimate or predict this phenomenon.The mathematical tool of geostatistics is the experimental variogram or semivariogram(Koudouet al.2014).Kriging is one of the geostatistical techniques for spatial modelling,which takes spatially dispersed data as input to obtain a generalized representation of this information (Hennequi,2010).Kriging refers to the semivariogram,which is calculated using the following formula:

Where:Z(xi)is the value of the variableZat location ofN(h),his the lag,andN(h)is the number of pairs of sample points separated by h.For random sampling,it has little chance that the distance between the sample pairs to be exactly h,wherefore h is usually represented as a distance range.Calculating the semivariogram values at several lags can help produce a semivariogram plot.These values are then fitted with a theoretical model:Spherical,exponential,or Gaussian.These models provide information on the input parameters for the Kriging interpolation as well as on the spatial structure of variables.Kriging is a type of weighted moving average method,which is also an optimal spatial interpolation method (Nas,2009):

3 Results

3.1 Factors influencing groundwater potential

3.1.1 Geology (G)

Groundwater potential is significantly dominated by lithology (Das and Pal,2019).According to Fig.3,four geological classes are identified in the study area based on their respective importance and effect on groundwater potential.The major geological types are Alluvium,Sandstone,Limestone and Granites.Alluviums play a significant role in groundwater supply than other geological formations.They are characterized by good permeability.In the study area,alluvium is widely distributed in the wadi Saida valley,which represents the rainwater collection area.The surface drainage network promotes the accumulation of surface flows in this area.Limestones are well exposed in the extreme northern part of Sidi Boubkeur and Sidi Amar and also the southern part of the watershed in Ain al Hadjar.Sandstone formation covers both the eastern and western part of the wadi of Saida.These formations are moderately favourable for groundwater recharge.The granite formations situated in the north of Hammam Rabi generally have very low groundwater potential except where they are fractured.

Fig.3 Geological map of Wadi Saida Watershed (After NAHR,2008)

The study area is marked by the presence of two water tables,a shallow water table and a karst water table.The first one is located exclusively in the valley of the wadi Saida and in the upper layers of the Saida formation and the plio-quaternary deposits (clay sands,calcareous clays and conglomerates).The karstic aquifer is situated in the carbonate sediments of the Lower and Middle Jurassic with a captive part in the valley and a free part in the rest of the aquifer (mainly Nador carbonate formation).

3.1.2 Lineament density (Ld)

Lineaments are linear features of tectonic origin that are long,narrow and relatively straight alignments.They are visible on satellite images as tonal differences compared to other terrain features(Klaus,2013).The distinguished lineaments may be the result of fracturing or fault system which indicates the possibility of increased porosity and permeability zone in hard rock and is consequently of importance in groundwater studies (Adiatet al.2012).

Lineament extractions-the maps of the lineaments network in this study are interpreted from Landsat 8 OLI/TIRS through image processing,such as lithological discrimination by the processing of composite band,band ratios,and principal component analysis (Fig.4). Directional filtering of the SRTM GL1 (Shuttle Radar Topography Global Mission 1arc second V003) was also used (Fig.5).Lineaments derived only from natural topographical features are indicators of groundwater flow-paths,therefore their hydrogeological importance needs to be evaluated(Jordanet al.2005;Sander,2007).The boundary contacts of two or more geological formations are also important in identifying high groundwater potential zones.The validation process consists of removing anthropic activities to leave only the contours and lines effectively corresponding to the lineaments (Koitaet al.2010),and observing the fracturing network already existing or assumed in the geological maps and the lineaments of geological contact or lithological and structural discontinuities.The application of the directional filter (5×5) allowed the mapping of 77 lineaments in Wadi Saida Watershed concentrated in the southern part of the basin in close proximity of the city of Saida.The filtering was carried out in different directions,i.e.0°,45°,9°,135°,180°,225°,270°,315°,only the directions of 45°,90°,135° and 225° are identified and merged (Fig.5).

Statistical Analysis of lineament was performed by the development of two-directional rosettes;the first represents the percentage of lineaments regarding the total length,and the second indicates the number of lineaments to the total number (Fig.6 and Fig.7).The statistical analysis of lineaments has been explored by several authors to study the geometry of the lineament network and to identify the dominant directions at the regional scale(Pretorius and Partridge,1974).

The conventional method consists of producing directional rosettes against the cumulative length of the lineaments in classes of 10° of orientation.The two directional rosettes show a nearly even distribution of percentage of the lineaments in both total length and total number of lineaments.The results show that 49.35% of the lineaments’directions are in the range of 10°～70° whereas 10.39% of them are SW-NE within the range of 60°～70°) which is the most dominant direction.Regarding the length of the lineaments,the ones with the SW-NE direction are also dominant,and 21% of the total length of the lineaments have the direction between 40° and 80°.

The statistical analysis of the lineament network shows that this direction is secondary because it represents only 18% of the total number of lineaments and 28% of the total length.The lineament density of the watershed varies from 0 to 2.5 km/km2(Fig.8) and was reclassified into five classes,i.e.＜0.5 km/km2,0.5～1 km/km2,1～1.5 km/km2,1.5～2 km/km2and ＞2 km/km2,where the first two classes represent 85% of the basin area(Fig.8).The comparison between the lineaments’direction and the piezometric analysis by Dahmani(2016) shows that the underground flow direction and that of the lineaments are identical,implying strong possibility of recharge of karstic aquifer by water percolated through the faults network.

3.1.3 Rainfall (R)

Rainfall is one of the main sources of groundwater recharge.More rainfall in any given region indicates higher groundwater potential recharge(Machiwalet al.2011).Intensity and duration of rainfall significantly affect groundwater recharge and potential (Wondifraw et al.2019).In the Wadi Saida Basin the annual average rainfall ranges from 300 mm to 330 mm,and the southern part of the basin receives higher precipitation but is of stormy character (Fig.9).

Fig.4 Different lithological discrimination process applied to the Landsat 8 image:(a) RGB (432),(b)RGB (742),(c) RGB (753),(d) Band ratios (6/2,5/3,4/2),(e) PCA (PC1,PC2,PC3)

Fig.5 Directional fi ltering applied to the SRTM image (45°,90°,135° and 225°)

Fig.6 Directional rosette of lineaments as percent of total number of lineaments

Fig.7 Directional rosette of lineaments length as percent of total lineament length

Fig.8 Lineament density map

Fig.9 Rainfall map

3.1.4 Slope (S)

Slope is presented as a degree,indicating the capacity of surface water to flow and reach the groundwater reservoir.This factor is extremely necessary for the identification of favourable groundwater recharge sites,as it directly affects run-off intensity,surface water infiltration and recharge (Nouayti et al.2017).Slope is an indicator for the suitable groundwater prospect(Klaus,2013).

The high degree of slope results in a rapid downward flow of surface water and a short infiltration time,indicating a low recharge level.However,the low degree of slope tends to increase recharge because of the increased retention of rainwater (Nouayti et al.2019).The slope of the Wadi Saida Watershed was divided into fi ve classes:＜3°,slight slope (3°～7°),moderate slope (7°～12.5°),steep slope (12.5°～25°),and very steep slope (＞25°) (Fig.10).The areas with steeper slope cause relatively high run-off andlow infiltration,and hence are categorized as poor and have less groundwater accumulation(Jha et al.2010).

3.1.5 Drainage density (Dd)

The drainage density of the Waidi Saida

Fig.10 Slope map

3.1.6 Land use (Lu)

Land use map was generated by the National Institute of Soils,Irrigation and Drainage (NISID,2009) to reclassify and rank the new land use categories of the study area (Fig.12).The new groups were reclassified as follows:Forest class,including dense maquis,dense forest and clear forest;built class,including agglomeration,airport and built-up;and shrubland class,including brush and clear maquis,and Agriculture and barren soil class,which remained the same.

3.2 Groundwater potential zones

The application of AHP techniques on the weighted components has allowed the mapping of groundwater potential zones in raster format in ArcGIS.In the pair-wise comparison matrix Watershed ranges between 2.22 km/km2and 3.33 km/km2(Fig.11),which was reclassi fi ed into three classes,viz,low (＜2.5 km/km2),moderate (2.5～3 km/km2),and high (＞3 km/km2).52% of the study area falls in the moderate class of drainage density,followed by low (36.4%) and high drainage density class (11.6%) (Fig.11).

Fig.11 Drainage density map

Fig.12 Land use map

(Table 3),factors of influence are reciprocals of each other and a normalized weight estimated in percent is given.

The calculation of the normalized weight of the layers revealed that lineament density had the greatest weight (44.8%),which is justified by the fact that the zone is characterized by the presence of a karstic aquifer where fracturing plays an important role in groundwater recharge.Precipitation is also an important factor and a weight of 20.44% is applied.

The weight of the geology factor is 14.47% and it shows that the alluvial and limestone-dolomitic zones have the highest rating (Table 4).The final groundwater potential map of the Wadi Saida Watershed using the AHP method shows four areas(Fig.13),i.e.32.39% of the total area,or 202 km2,is classified as good and very good groundwater potential zone;65.76% of the area is classified as moderately potential zone and only 1.85% as the low potential zone (Table 5).

Table 3 Pairwise comparison matrix and percent normalized weights of used criteria

Table 4 Classification of influencing factors for groundwater potential zones

Table 5 Area statistics of groundwater potential zones

3.3 Accuracy assessment and validation

Geostatistic approach was used to validate the groundwater potential map.It involves questions about how data in different layers can relate to each other,and how it varies in space (Huisman and Rolf,2009).

In the present work,we applied the ordinary kriging method,of which the stationary variable of the mean is unknown and the variogramγ(h)reaches a plateau where it grows less rapidly thanh2.The nugget effect represents the abrupt variation of the measured parameter (Table 6) (Fig.14).The correlation clouds compare the measured values (in abscissas) against the estimated values(in ordinates),and a good correlation coefficient of 0.87 has been achieved (Table 7).According to the model obtained from the estimation error (Fig.15),the adjusted ordinary kriging is more accurate because the average error is closer to zero and the standard deviation of the estimation errors is small(about 41%).

Table 6 Model characteristics

Fig.14 Semi-Variogram of estimated flows

Fig.15 Regression of the measured values according to the estimated values

Table 7 Correlation function characteristics

The spatial distribution of borehole yield was generated using ordinary kriging on a set of 47 boreholes with yield data.Three zones were identified to fill the areas which were not delineated before we added neighbourhood boreholes to the watershed (Fig.12).The borehole yield is classified into four class:Poor (0 m3/h to 10 m3/h),moderate (10 m3/h to 50 m3/h),good (50 m3/h to 75 m3/h) and very good (＞75 m3/h) (Fig.16).An accuracy assessment to validate the results was performed by comparing the actual yield of 21 boreholes in the field and the estimated optimum yield from the two maps produced (Ordinary Kriging map of existing borehole yield and the GWP map) (Fig.13 and Fig.16).

The procedure can be described as follows,if the classification results are the same or similar from the three sources of input (actual yield from the field,estimated yield from the kriging map and estimated yield from the GWP map),we may conclude that an agreement status has been reached between the estimated and the actual yield.On the other hand,if these classes are different,we may state that there is a disagreement between the estimated and actual yield.A partial agreement can be presented when close values are found between two sources of inputs of the three (Table 8).Out of the 21 wells,optimum yield of 15 wells,or 71%of the total number of wells,suggests that there is a partial or full agreement between the estimated yield using Kriging map (Fig.16) or Groundwater potential map (Fig.13) and the actual yield of boreholes.The rest 6 boreholes show an absolute disagreement between the estimated and actual yield.

Fig.16 Ordinary Kriging map of existing borehole yield

Table 8 Accuracy assessment of groundwater potential zone map with optimum yield data

4 Discussion

The Wadi Saida Watershed is recognized as the major target for groundwater exploitation since it is the main source of water supply in the region. Boreholes are generally deep (more than 150 m) with yields ranging from several m3/h up to more than 100 m3/h.The application of Analytical hierarchy process (AHP) method has shown that some criteria have more influence than others.In this study,the lineament density carries the highest weight,followed by geology.Rainfall in the study area is characterized by irregular and intense events in the autumn,which disturbs the groundwater recharge and causes high groundwater demand in irrigation and other activities.The slope also plays an important role where runoff and infiltration have the great impact on groundwater recharge.

According to the directional filtering of lineaments,the dominant direction is the SW-NE,in which the latter is due to the alpine orogeny.This type of tectonics affects the ancient formations characterized by large folds giving rise to the Saida Mountains.A dense lineament network is present where the most important lineaments are located along the border of the Saida city towards the northwest (faults of Zeboudj) and the southeast.In the southeastern part of the watershed,a geological contact between the karst aquifer system and other aquifers is located.Multiple faults are presented in this area allowing the discharge of water to the deepest aquifer in contact.Those aquifers in the Rebahia zone can provide a good supply of groundwater.

According to Bencherki (2008),the structures of the Viseean orogeny are corresponding to faults and folds with the directions of NE and NE-SW.The uplift also caused the deformation of the Jurassic dolomitic cover,while a collapse ditch along N-S axis was developed in the Saida valley.These differentiated uplift movements have generated brittle tectonics in the region.The principal directions are approximately SW-NE to NNE-SSW.Tectonic events are mostly sub-vertical(Bencherki,2008).

The karstic aquifer extends over the Saida Mountains,part of which is located on the southeastern side of the watershed and the other part is outside the basin.A vertical impermeable flow boundary presents in the southwest generated by two perpendicular faults to the west.These faults form an impermeable substratum which limits the water table to its northwest.Thus,it is bordered by a discharge line along the Triasopaleozoic basement aquifer and the edge of the Ain Soltane and Ain Balloul plateaus,and an impermeable contour in the Djebel Mozdbab area with a cut in the aquifer in the north sector of Mozdbab horst.To the northeast,the diagonal fault of the eastern flank of the Tiffrit mole is an impermeable boundary.According to Djidiet al.(2008),the saida valley is devoided of epikarst as evidenced by the intensely fractured outcrops as well as the presence of caves and chasms with several meters in diameter (Boukhors,Ain Zerga,Vieux Saida,Bir H’mam).The karstic system can be divided into two formations:One is in the surface mainly occurred in Tidernatine plateau as sinkholes,losses and backslides giving rise to funnels like Ghar Ouled Amira,Ghar Slouguia;the other one is at the bottom where sinkholes may evolve with collapse of the underlying karstic void (Bir H’mam well) and other deep ones whose system is drained by the source of Ain Zerga,the underground river of Bir H’mam and the loss of Ghar Ouled Amira.This river could be the principal collector of the karstic system of Tidernatine.The losses could feed the underground river of Bir H’mam and the resurgence of Ain Zerga.

Previous studies (Pitaud,1973 in Dahmani,2016 and Djidiet al.2008) have shown that the series of Jurassic carbonates resting on the impermeable substratum of Triassic clays form the main aquifer in the region.With local aquifers,Upper Jurassic and Lower Cretaceous sandstones can be drained down towards the underlying carbonate aquifer while the network of existing faults can play the interconnectivity role between these aquifers,which demonstrates the importance of lineament density in the study of the potential of groundwater.Also,as our study has proven,these previous studies showed that the Jurassic carbonate formations are influenced by faults oriented NESW especially in the Wadi Saida valley,where the lineament density is higher than other parts of the watershed.The spatial coverage of precipitation around the slope gradient of water points in conjunction with the hydrographic network substantially influences the infiltration rate.Runoffincreases and affects negatively the potential groundwater areas (Kumaret al.2014).Land use analysis provides important indicators of the extent of groundwater demands and its usage,as well as an important indicator in the selection of sites with good groundwater potential.Demographic data and new urban areas as well as regions with significant agricultural potential are also needed in the planning of new borehole sinking projects.

From the Groundwater Potential map using AHP model (Fig.15),areas with high estimated borehole yields are located in the northeast of the Wadi Saida Watershed,west of Ain Soltane and Hammam Rabi and the region of Ain el Hadjar which is characterized by the presence of discontinuous aquifers with cracks and karstic features composed mainly by upper Cretaceous limestones and dolomitized and marl-limestones from the lower Cretaceous.Estimated average flow rates are observed in the West of Sidi Boubkeur and Doui Thabet,characterized by the presence of carbonate sediments from discontinuous aquifers of the upper and middle Jurassic.Low values are concentrated in the city of Saida and the extreme eastern zone of the study area,which represent the boundary between the continuous aquifers of the quarter-century and the discontinuous karst aquifers.While,the Ordinary Kriging map of existing borehole yields (Fig.16) show also that,the very good to good estimated flow class were located in the north-eastern part of the basin on Sidi Boubkeur region,the area of Hammam Rabi and Rebahia,as well as the south-west part of Ain el Hadjar.

Geostatistical analysis of 47 wells was used in this study to validate of the groundwater potential map produced with the data from existing boreholes.Variographic analysis showed a good correlation between the actual values and those estimated.According to the results,the correlation between the estimation zones (groundwater potential zone map) and the spatial distribution of actual borehole yields (Ordinary Kriging map of existing borehole yields) is concluded.The results indicate that the southern part of the basin is more favorable for groundwater exploitation,the north of Sidi Boubkeur and Sidi Amar and the south of Ain El Hadjar and the far west of Doui Thabet.

As a result,the mapping of groundwater potential through the Remote Sensing and GISbased AHP approach could be applied with satisfaction.Consequently,these results can be utilized for better planning and management of groundwater resources,borehole siting,and maintaining the sustainable exploitation of areas.

5 Conclusions

In this study,an integrated approach of GISbased Analytical Hierarchical Process is used to delineate the groundwater potential zone in Wadi Saida watershed.Six layers were selected as influencing factors (geology,lineament density,rainfall,drainage density,slope and land use).The processing and interpretation of SRTM and Landsat 8 OLI images has provided good results in the mapping of geological lineaments present in the Saida region.Groundwater potential in the study area is mainly controlled by lineament density,geological settings and rainfall,while slope,drainage density and land use are considered as secondary factors.The groundwater potential map using AHP shows that the whole study area can be classified into four groundwater potential zones,such as Very good (2.26% of the area),Good (30.13% of the area),Moderate (65.76% of the area),and Poor (1.85% of the area).

This study confirms the efficiency of using remote sensing and GIS for the identification of groundwater potential zone through AHP techniques and geostatistic analysis in semi-arid regions.The current study provides highly valuable knowledge to decision-makers for planning and framing new strategies for sustainable groundwater development and water resources management.