Keywords

1 Introduction

Groundwater resource is an important natural resource for its use in agriculture, industries and domestic purposes. It is unevenly distributed both spatially and temporally all over the world. In northern Ethiopia, where rainfall is scarce and unevenly distributed; groundwater is the main source to meet various water demands [4]. The demand for groundwater in Lake Tana basin (LTB) has been increasing over the years [48]. Development of groundwater resources is essential to satisfy the increasing demand [47]. In this area, groundwater recharge varies significantly in space and time due to the difference in distribution and amount of rainfall, permeability of rocks, geomorphology and availability of surface water bodies close to major unconfined and semi-confined aquifers that feed the groundwater [4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37].

There are several conventional methods, used to assess groundwater potential, some are: Ground surveys, Geophysical Methods such as resistivity, and ground penetrating radar; Probabilistic Models like Logistic Regression Methods [3] are unreliable and time consuming. There are also various statistical methods that were adopted for groundwater potential mapping such as frequency ratio [13,14,15,16,17,18,19,20,21,22,23,24,25,26,27], multi-criteria decision evaluation [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43], weights-of-evidence model [29]; random forest model [24] maximum entropy model [30] boosted regression tree, classification and regression tree [24]. Remote sensing and Geographic Information System (GIS) techniques have become an easier, effective and time efficient in the assessment, monitoring and conservation of groundwater resources [45]. These days a number of researchers, [2, 10, 15, 19, 28] applied remote sensing and GIS through Analytical Hierarchy Process (AHP) for the assessment of groundwater potential.

Even though the importance of the groundwater resources is essential for development, limited groundwater studies have been carried out in the basin. The potential is not well defined. However, the government of Ethiopia has declared the Lake Tana – Beles as a growth corridor. Therefore, the objective of this study is to assess the groundwater potential in Lake Tana basin using Geographical Information System (GIS) and Remote Sensing (RS) based MCDA techniques, and to produce a groundwater potential map.

2 Methods and Materials

2.1 Description of the Study Area

Lake Tana basin is located in northwestern Ethiopia (Fig. 1), which feed to the Blue Nile River. Geographically, it is located between 10°57ʹ–12°47ʹN and 36°38ʹ–38°14ʹE. The basin landscape is part of the western plateau of Ethiopia and includes the escarpments of Gondar and Gojjam, the lower plains surrounding the lake that forms extensive wetlands of Dembiya, Fogera, and Kunzila plains located in the north, east, and southwest of the lake, respectively. The basin has a total area of 15,320 km2.

Fig. 1.
figure 1

Location map of Lake Tana basin

Full size image

2.2 Data Collection and Preparation

In this study eight thematic layers (drainage density, rainfall, slope steepness, land cover/use, lineaments, soil and lithology) were used. The rainfall map was prepared using Inverse Distance Weighting (IDW) interpolation from 26 stations long term mean annual rainfall data. The lithology, soil, and land use are collected from Ethiopian Geological survey. These thematic layers were prepared using Arc GIS 10.1 platform. The drainage and lineament density maps have been prepared using the line density analysis tool in Arc GIS 10.1. The slope and drainage density were derived from Shuttle Radar Terrain Mapping Digital Elevation Model (STRM DEM) of 30 m resolution. Arc Hydro tool was used to generate stream network and drainage density was calculated in Arc GIS 10.1 using Eq. 1 [12].

$$ {\text{DD}}\, = \,\Upsigma {\text{Li}}/{\text{A}} $$
(1)

Where, DD is drainage density; Li denotes the total length of streams (L) and A denotes the unit area.

Cloud free Landsat 8 Operational Land Imager (OLI) images (Row 52/Path 169, Row 51/Path 170, and Row 52/Path 170), dated June 17 and 24, 2017 were downloaded from USGS website (https://earthexplorer.usgs.gov) for lineament extraction. After applying principal component analysis (PCA) technique using ENVI 5.1 to produce uncorrelated output bands, to segregate noise components, and to reduce the dimensionality of data sets [25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43], lineaments were extracted automatically using PCI Geomatica 2017 trial version from Landsat 8 [23]. Lineament density was calculated in Arc GIS 10.1 using Eq. 2 [13].

$$ {\text{LD}} = \Upsigma {\text{Li}}/{\text{A}} $$
(2)

Where, LD is lineament density; Li is denotes the total length of lineaments (L) and A is denotes the unit area.

2.3 Weighting and Ranking of Each Layers (AHP)

Because all parameters do not have equal effect on groundwater distribution, ranking of each parameter is needed. Weighting each factor and a pair-wise comparison matrix were prepared for each map based on [34]. The rate/rank for classes in a layer and weights for each factor was assigned and a pair-wise comparison matrix has been prepared for each map by using Multi Criteria Decision Analysis technique based on Analytical Hierarchy Process (AHP) [35] by considering seven factors (lithology, lineament density, soil, and drainage, rainfall, slope and land use/cover). Square matrices are constructed, where each criterion is compared with the other criteria, relative to its importance, on Saaty’s scale from 1 to 9 (Table 2). The rate/rank for classes in a layer and weights for thematic layers of each factor was calculated as the following steps:

  1. 1.

    Square matrix A = aij (Eq. 3) the element of row i column j was produced and the lower triangular matrix was Completed by taking the reciprocal values of the upper diagonal using the formula aij = 1/aij

    $$ A = \left( {\begin{array}{*{20}l} {\frac{p1}{P1}} \hfill & \cdots \hfill & \cdots \hfill & {\frac{P1}{Pj}} \hfill & \cdots \hfill & \cdots \hfill & {\frac{P1}{pn}} \hfill \\ {\frac{pi}{P1}} \hfill & \cdots \hfill & \cdots \hfill & 1 \hfill & \cdots \hfill & \cdots \hfill & {\frac{Pi}{pn}} \hfill \\ {\frac{pn}{P1}} \hfill & \cdots \hfill & \cdots \hfill & {\frac{Pn}{Pj}} \hfill & \cdots \hfill & \cdots \hfill & {1\frac{Pn}{Pn}} \hfill \\ \end{array} } \right) $$
    (3)
  2. 2.

    Summation of all columns j matrix from (Eq. 3) using (Eq. 4)

    $$ \frac{ p1}{Pi} + \ldots \frac{Pi}{Pj} + \ldots \frac{Pn}{Pn} = (\mathop \sum \nolimits_{i = 1}^{n} Pi)/Pi $$
    (4)
  3. 3.

    Divide each element of matrix aij = pi/pj (Eq. 3) by (Eq. 4) to get normalized relative weight (Eq. 5)

    $$ \frac{{\frac{Pi}{Pj}}}{{(\mathop \sum \nolimits_{i = 1}^{n} Pi)/Pj}} = \frac{Pi}{Pj} *\frac{Pj}{{\mathop \sum \nolimits_{i = 1}^{n} Pi }} = \frac{Pi}{{\mathop \sum \nolimits_{i = 1}^{n} Pi }} $$
    (5)
  4. 4.

    Averaging across the rows (Eq. 6) to get the normalized principal Eigen vector (priority vector) i.e. Rate (Ri) or weight of row ‘i’(Wi). Since it is normalized, the sum of all elements in priority vector should be one.

    $$ {\text{W}}_{\text{i}} /{\text{R}}_{{{\text{i}} = }} \left( {\frac{Pi}{{\mathop \sum \nolimits_{i = 1}^{n} Pi }} + \ldots + \frac{Pi}{{\mathop \sum \nolimits_{i = 1}^{n} Pi }}} \right)* 1/{\text{n}} $$
    (6)

The judgments of the pair-wise comparison within each thematic layer were checked by the consistency ratio [35]. The consistency ratio (CR) is a measure of consistency of the pair-wise comparison matrix. Scholars [21,22,23,24,25,26,27,28,29,30,31,32,33,34,35] recommended that for matrices with CR rating greater than 0.1, the process is repeated until the desired value of CR < 0.1 is achieved. The value of CR is computed as (Eq. 7)

$$ {\text{CR}} = {\text{CI}}/{\text{R}} $$
(7)

where, RI is Saaty’s ratio index; CI is Consistency index.

Consistency index is a measure of consistency or degree of consistency of the judgment [21,22,23,24,25,26,27,28,29,30,31,32,33,34] and computed using (Eq. 8).

$$ {\text{CI}} = \left( {\lambda { \hbox{max} } - {\text{n}}} \right)/\left( {{\text{n}} - 1} \right) $$
(8)

where, n is the number criterion, the value of RI for n criteria as indicated in Table 1.

Table 1. Random consistency index (RI)
Full size table
Table 2. Scale of preference between two parameters in AHP [34]
Full size table
Table 3. Pair-wise comparison matrix between all thematic layers (CR = 0.03)
Full size table

To compute λmax, first multiply the normalized value by the respective Weight (Eq. 6) and then, the values of the product are added together to get λmax (Eq. 9)

$$ \lambda { \hbox{max} } = \mathop \sum \limits_{{{\text{i}} = 1 }}^{\text{n}} (Wi *(Pi/ \mathop \sum \limits_{i = 1}^{n} Pi)) $$
(9)

where, λmax, the largest Eigen value of the pair wise comparison matrix. Pi is the priority of the alternative and wi is the assigned rate/weight.

2.4 Overlay Analysis

After assigning the rates for classes in a layer and weights for thematic layers each factors the final map (Fig. 9) was obtained by overlaying all thematic maps using weighted overlay methods [19,20,21], using the spatial analysis tool in are G.I.S 10.1 as shown (Eq. 10)

$$ GWP = \mathop \sum \limits_{{{\text{i}} = 1 }}^{\text{n}} Wi *Ri $$
(10)

Where, GWP is groundwater potential zone; Wi is weight for each thematic layer and Ri; rates for the classes within a thematic layer derived from AHP.

2.5 Sensitivity Analysis

The map-removal sensitivity analysis was used to examine the impacts of removing one of the thematic layers in the computation of the GWP distribution map. The map removal sensitivity analyses [5,6,7,8,9,10,11,12,13,14,15,16,17] were used to identify the factors that affect the groundwater potential, details are shown in Table 4. In this method, each of the thematic layers are removed, and a new GWP map is obtained each time for the remaining layers overlaying each other. Map removal shows the sensitivity of the GWP map to the removal of one of the thematic layers. This analysis is expressed in terms of an index of sensitivity as shown (Eq. 11).

Table 4. Statistics of map removal sensitive analysis
Full size table
$$ {\text{S}} = \, \left[ {\left( {\left( {{\text{GWP}}/{\text{N}}} \right) - \left( {{\text{GWP}}^{\prime } /{\text{n}}} \right)} \right)/{\text{GWP}}} \right]* 100 $$
(11)

where S is the index of sensitivity associated with the removal of one map; GWP is the groundwater potential index computed using all the thematic layers; GWPˈ is the groundwater potential index computed by excluding one thematic layer at a time; N and n are the number of thematic layers used to compute GWP and GWP′, respectively.

3 Results and Discussion

3.1 Design and Preparation of Thematic Maps

Drainage Density (DD)

Drainage Density is expressed as the total length of all streams or rivers per unit drainage area [12, 13]. In areas where a higher drainage density is dominant, ground-water potential is generally poor, because less time for infiltration. Whereas, areas of low drainage density allow more infiltration to recharge the groundwater system [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22]. Stream network derived from 30 m DEM, mainly six drainage density categories have been identified and mapped as shown in Fig. 2. The highest drainage density was found in the southern, eastern and northeastern part of the Lake Tana basin. However, moderate and low drainage density concentrates in the southern and central part of the lake basin. By using pairwise comparison, the rate was set based on groundwater potential for areas with low drainage density were assigned the higher rate, and high drainage density were assigned the lowest rate due to low groundwater potential. The lowest drainage density (0–0.06 km/km2) were assigned the highest rate (0.41) and areas having higher drainage density 0.35–0.5 and 0.5–0.8 km/km2 were assigned the lowest rate of 0.05 and 0.03, respectively.

Fig. 2.
figure 2

Stream network and drainage density map of the Lake Tana basin

Full size image

Lineament Density (LD)

Lineaments are the simple and complex linear properties of geological structures like: faults, cleavages, fractures, and various surfaces of discontinuities (terraces and ridges) arranged in a straight line or a slight curve that are identified by remote sensing [25]. Lineaments were extracted automatically using Geomatica module, which is extracted from a single image of principal component analysis (PCA) and record the output as polyline with vector segment. Due to multispectral imaging bands are often highly correlated; Principal Components used to produce uncorrelated output bands, to segregate noise components, and to reduce the dimensionality of data sets [23,24,25]. This is done by finding a new set of orthogonal axes that have their origin at the data mean and that are rotated so the data variance is maximized. After applying Principal component analysis of Landsat images eight-bit gray scale lineaments are extracted automatically (Fig. 3). Areas with high lineament density are assumed good for groundwater potential [10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41]. From the result, the highest lineament density was observed in the North Eastern escarpments and South tip of the basin. As shown in Fig. 3, the highest lineament density (0.63–1.10 km/km2) was assigned the maximum rank (0.39) and the lowest rate/rank (0.02) was assigned to the lineament density class of 0–0.083.

Fig. 3.
figure 3

Lineament density map

Full size image

Slope

Slope defines the variation of elevation in a particular area which influences the runoff [24]. In gentle slopes, the movement of water is generally slow, which allows more infiltration. Whereas, in steep slopes runoff is favored and lower infiltration. Slope map has been derived from 30 m resolution DEM (Fig. 4). Slope varies from 0 to 72.8° and classified based on natural break classification method [26]. It classified in to six classes viz., 0–2.6, 2.6–6.7, 6.7–12.6, 12.6–19.4, 19.4–28.3, and 28.3–72.8. The highest rate (0.39) was assigned for the lowest slope gradient which is observed in central parts of the basin surrounding to Lake Tana and the lowest rate (0.02) was assigned for the highest slope gradient which found in northeastern, eastern and northern tip of the basin.

Fig. 4.
figure 4

Slope gradient of Lake Tana basin

Full size image

Rainfall

Rainfall is another factor that governs the amount of water that would be available to infiltrate and joins the groundwater zones, and thus it controls groundwater potential [42]. The rainfall map is classified into six classes using equal-interval classes [32]. Long term mean annual rainfall map of the area is prepared using IDW interpolation from 26 meteorological stations (Fig. 5). Based on the mean annual rainfall and its contribution for groundwater recharge; the relative weight (rate) was given for each rainfall classes. The rate was given based on the contribution to groundwater recharge, higher rate was assigned for high rainfall; and low rainfall areas were assigned the lowest rate for low groundwater potential [16]. Mean annual rainfall between 1713–2231 mm were assigned the highest rate (0.41), and 1507–1712 mm rainfall were assigned the next higher rate of 0.25. Areas having mean annual rainfall of 1039–1205 mm and less than 1000 mm mean annual rainfall were assigned the lowest rate as 0.05 and 0.03, respectively.

Fig. 5.
figure 5

Meteorological stations and rainfall map of Lake Tana basin

Full size image

Lithology

Lithology plays an important role in the occurrence and distribution of groundwater. According to Ethiopian Geological survey, the major lithological units of Lake Tana basin are Tarmaber basalts (39.8%), Basalts related to volcanic center (26.5%), alluvium deposit (4.7%), lacustrine deposits (2.3%), colluvium (1.9%), Amba Aiba Basalts (0.07%), Lateriteon Amba Alaji Rhyolite (0.003%), marsh soil (4.7%), and Lake (20%). The basin covered by Basalts related to volcanic center mainly containing scoriaceous basalt widely found at southern part and the Tarmaber Basalts at northern, western and the southern end of the basin (6–7). The lowland covered by a thick Quaternary alluvial deposit (clay to gravel grain size), lacustrine deposits and Colluvium formed the major aquifer zones. Based on their age, the aquifer systems classified in to three major groups. Those are: the Tertiary Volcanic (Tarmaber formation, Amba Aiba Basalts and Lateriteon Amba Alaji Rhyolite), the Quaternary Fractured Vesicular Basalt (quaternary basalt related to volcanic centers) and the Quaternary Alluvial deposits such as Alluvium, Colluvium and Lacustrine deposits) [39, 40]. As shown as Fig. 6 the south part with Quaternary Basalt overlain by Alluvial Deposit displays the highest productive aquifer. However, the northern, southern and eastern mountainous regions, mostly covered by highly degraded tuff and massive basalt, are characterized by lowest productivity [40]. Thus, the highest rate was given for the highest productive aquifer, and the lowest rate was given for lowest productivity.

Fig. 6.
figure 6

(source: Ethiopian Geological survey)

Lithological map of Lake Tana basin

Full size image

Land Use/Land Cover

Land use cover significantly affects hydrological processes; i.e. interception, infiltration capacity of soil, runoff generation mechanism, are affected by the land use/land cover type. This in turn affect groundwater recharge and discharge [33]. Most of the study area is covered by cropland with sparse woodlands while only few limited areas of highlands are forested (less than 1% of the basin). The major land cover types are cropland 45%, wood land 18%, water body (20%), Grass land (13%), and bare land (2%) Forest (1%), and urban and built 0.2% [47]. As shown in Fig. 7 forests were assigned the highest rank due to less runoff because leaves and trees slow down the rain before it hits the ground, giving plant root’s time to infiltrate and join into the groundwater system. From the pair-wise comparison, the rate was given based on contributions for groundwater recharge [38]. Highest rates were assigned for areas cover with water bodies (0.26) and swampy (0.19). Whereas, the lowest rates were given for cultivated (0.02) and urban areas (0.01).

Fig. 7.
figure 7

Land use/land cover map of Lake Tana basin

Full size image

Soil Group

Soil is an important factor for delineating the groundwater potential areas. Soil characteristics invariably control rates of infiltration; percolation and permeability [38,39,40,41,42,43,44]. Tana basin is covered by a wide range of soil groups. As shown Fig. 8, the major soils are: Haplic Luvisols which covers (26%), Chromic Luvisols (20%), Eutric Leptosols (16%), Eutric Vertisols (15%), Eutric Fluvisols (12%), Haplic Alisols (6%), Lithic Leptosols (3%), Haplic Nitisols as medium portion, and Eutric Regosols and Eutric Cambisols less than 1% of the basin. The southern part is covered by Hablic Luvisols and Chromic Livisols which have higher infiltration infers high groundwater potential [8]. The floodplain to the north and east of Lake Tana is covered by Eutric Leptosols and Eutric Vertisols which have the lowest infiltration rate. The major soil properties of Lake Tana basin are classified based on its texture, drainage condition and infiltration capacity. Thus, Eutric Regisols which are characterized by sandy loam to loam texture; It excessively drain and are categorized as soil group A as highly permeable and soil group B such as: Haplic Livisols, Chromic Livosols, Eutric Fluvisols and Haplic Nitosols are characterize as well drained to moderately drain condition due to having textural class of clay to silt clay [8]. Therefore, Eutric Regosols, Hablic Luvisols and chromic Livisols which have higher infiltration were given the relative weight (rate) of 0.18, 0.14 and 0.10, respectively. However, soil group like Eutric vertisols and Lithic Leptosols which contain clay texture and characterized by poorly drained condition and assigned the lowest rate of 0.02 (Fig. 8).

Fig. 8.
figure 8

(source: Ethiopian Geological Survey)

Soil map of the Tana basin

Full size image

3.2 Combination of Thematic Layers

Since for all layers, the value of consistency ratio (CR) is less than 0.1 and, hence, the judgments of the pair-wise comparison within each thematic layer (Table 3) were acceptable [35]. Therefore, groundwater potential areas were obtained by the combination of all layers using overlay analysis method in Arc GIS 10.1 [22] using Eq. (12).

$$ {\text{GWPZ}} = 0. 3 8*{\text{LT}} + 0. 2 4*{\text{LD}} + 0. 1 6*{\text{Soil}} + 0. 10*{\text{SL}} + 0.0 6*{\text{DD}} + 0.0 4*{\text{RF}} + 0.0 2*{\text{LU}}/{\text{C}} $$
(12)

As shown in Fig. 9, the southern and eastern part of the basin shows very high groundwater potential zones due to presence of Quaternary basalts related to volcanic centers which are highly productive, alluvial plain and gentle slope. Similar studies have been reported [18 36 & 38]. The least groundwater potential areas are found in south eastern and northern escarpments (tips) of study area. The result indicated that, about 15% (1765 km2) classified as “very high,” 26.5% (3151 km2) as “high,” 31% (3592 km2) as “medium,” and 27.5% (3224 km2) were classified as “low” groundwater potential areas in the basin.

Fig. 9.
figure 9

Groundwater potential map of Lake Tana basin

Full size image

3.3 Validation of Groundwater Potential Map

Validation of the groundwater potential zones were conducted using existing groundwater yield data of hand dug wells and/or bore holes and springs [9]. The groundwater potential map was verified with 195 existing boreholes and springs data (Fig. 10). Out of 195 existing water points (boreholes and springs) in which 149 (76%) were matched with its corresponding classes of groundwater potential zones (GWPZ). This indicates that, the groundwater potential map and actual borehole data shows good agreement.

Fig. 10.
figure 10

Distribution of boreholes and springs with its yield in GWP zone

Full size image

3.4 Sensitivity Analysis

Sensitivity analysis is important to know the influence of each factor on a ground-water potential map. For this study map removal, sensitivity analysis was used based on [5,6,7,8,9,10,11,12,13,14,15,16,17]. It indicates that the index variation of thematic layers by removing one thematic layer at a time using Eq. 6. As shown in the Table 4, the statistical summary of variation index resulting from removing each thematic layer was identified. The highest variation index with mean 2.65% associated to lithology, which is the highest sensitive parameter of groundwater potential. However, it is less sensitive to soil with mean variation index of 0.44%.

4 Conclusions

The result of this study indicated that from the total area of the basin; about 27.5% (3224 km2) classified as: “low”, 30.6% (3592 km2) “moderate”, 26.9% (3151 km2) “high”, and 15% (1765 km2) “very high” groundwater potential zones, excluding the lake area. Out of 195 existing (borehole and springs) data for validation; 149 (76.4%) were agreed with corresponding groundwater potential classes. High ground-water potential was found at the southern and eastern side of Lake Tana. Groundwater can be exploited in these areas. On the other side, the least potential areas were found in northern, eastern and southern escarpments (tips) of the basin, respectively. This was because of the characterization by high gradient mountains, steep slope, high drainage density, hills, and hard-rock formation, which contribute high run off and less infiltration. Hence, in order to meet the increasing demand, new well fields may be formed in high groundwater potential zones, especially in the southern and eastern part of Lake Tana basin. But, in low groundwater areas, soil and water conservation measures and integrated watershed management should be applied to improve groundwater recharge. Future detail geophysical survey investigation in the basin is required for better results.