Identifying groundwater recharge potential zone using analytical hierarchy process (AHP) in the semi-arid Shinile watershed, Eastern Ethiopia

Water is essential for ensuring health, food security, and livelihoods, as well as for sustainable economic development. Ethiopia has long been susceptible to water concerns even though it is referred to as ‘the water tower of Africa’. Water resources are unevenly distributed, with over 70% of them being in the West while the majority of the country's water consumption is currently in the eastern regions (Tolche 2021). In addition to being naturally vulnerable to extremely variable rainfall, Ethiopia is now experiencing regular droughts and floods as a result of climate change (Sitotaw & Hailu 2019). In addition to accounting for a significant portion of the current water supply in Ethiopia, groundwater has been identified as a prime mechanism to combat seasonal variability and climate change-related water shocks. However, information related to the distribution, depth, and quality of groundwater in Ethiopia is extremely scarce leading to poor management, over-extraction, and high levels of depletion across the country (Tilahun & Merkel 2009).

The analysis of groundwater flow networks and the computation of the water budget in the context of the hydrologic cycle depend heavily on the recharge and outflow of groundwater. Recharge is a key component of developing water supplies in semi-arid and arid areas, balancing water demand, and managing water resources. Due to rising freshwater needs for domestic, agricultural, and industrial demand, groundwater has become an even more indispensable resource. However, surface water as a source of recharge is limited in dry and semi-arid countries (Lentswe & Molwalefhe 2020; Zghibi et al. 2020).

In Ethiopia, groundwater is a significant source of drinking water, accounting for 80% of total consumption (Tilahun & Merkel 2009; Lulu et al. 2019). The contribution of groundwater is even larger in the drier parts of southeastern Ethiopia. Currently, deep boreholes, hand-dug wells, and springs provide Dire Dawa and other large towns like Harar, Haramaya, Adele, and Aweday with the entire water supply for drinking, agricultural, and industrial uses (WWDSE 2002). The main challenges for this eastern region include reduced recharge rates related to landscape degradation, over-extraction, depletion, high levels of calcium carbonate from the deeper rock formation, and urbanization-related groundwater pollution.

Groundwater resources are under significant stress due to the growing demand, which also poses a risk to many communities' ability to calculate their water budgets. In arid areas with high temperatures and low rainfall, groundwater recharge dynamics are determined by the soil, slope, vegetation, and landscape. Rainwater infiltration into the watershed and subsequent storage in the aquifer is also significantly influenced by the geological properties and general soil profile characteristics. The body of literature on groundwater recharge demonstrates the conceptual complexity of geospatial analytics as a bottleneck to a sufficiently accurate recharge estimation (Xu & Beekman 2003; Scanlon et al. 2006).

The determination of the groundwater recharge zone is essential for groundwater management, including the development of artificial recharge programs and the avoidance of groundwater depletion. Groundwater recharge is impacted by anthropogenic activities and natural factors, such as climate, surface, and within the aquifer systems. Rainfall, lineament density, topography, lithology, drainage density, soil type, soil depth, land use/land cover (LULC), geomorphology, and the characteristics of the unsaturated zone all have an impact on where groundwater occurs and moves (Abdul Bari & Vennila 2013; Etikala et al. 2019; Zghibi et al. 2020). To identify the surface and groundwater over vast areas, geographic information systems (GIS) and remote sensing (RS) are effective techniques (Etikala et al. 2019). Different methods are used to map groundwater recharge zones, e.g. Lentswe & Molwalefhe (2020). The weights of evidence (WOE), multi-influencing factors (MIF), frequency ratio (FR), certainty factor (CF), and fuzzy logic index models are among the approaches used to map recharge zones. Due to their speed, accuracy, and affordability in making decisions based on methodical expert judgment, the analytical hierarchy process (AHP) and MIF are the methodologies that have received the most attention in groundwater recharge potential zoning (Abijith et al. 2020). By allocating weights based on professional judgment, the AHP multi-criteria decision-making (MCDM) technique compares geographical parameters in a paired manner (Al-Ruzouq et al. 2019).

This study uses weighted overlay analysis based on the AHP technique to integrate six influencing parameters (geology, lineament density, slope, soil texture, drainage density, and LULC) and determine the spatial distribution of groundwater recharge in the Shinile Watershed. By displaying the borehole, hand pump, and stream ground control points (GCP) on the high-potential zone, the AHP's accuracy is assessed.

Shinile Watershed provides water to major towns in its vicinity including Dire Dawa, where water is rationed with each household getting water once every 2 weeks (DDWSSA 2021). Water scarcity and the subsequent rationing have been progressively worse in this area and can be directly correlated to the decline in its groundwater sources. In this semi-arid Shinile Watershed, groundwater was accessible at less than 100 m depth in the 1980 and 1990s. Currently, groundwater depth exceeds 500 m indicating an approximately 13 m per year decline rate. The cost of drilling and pumping ever-increasing borehole depths is draining the city's limited financial resources. The severity of groundwater decline is also causing high levels of calcium bicarbonate related to the deeper rock formation. Calcium bicarbonate-related water quality issues are increasing health risks, increasing water treatment costs, and damage to water infrastructure from mineral deposition over time. The decline in groundwater depths in the watershed is related to growing water demands and reduced groundwater recharge rates. Because of the topography and the limited watershed development activities, the watershed is known for flash floods and limited time for subsurface infiltration during the rainy season. While interventions to reduce flash floods and improve groundwater recharge are known, financial constraints are a limiting factor to restoring large swaths of the watershed. Therefore, it is imperative to identify high-potential groundwater recharge zones to prioritize restoration areas and optimize available resources. This research will provide substantial input to decision-makers involved in watershed management to consider high recharge potential locations in soil and water conservation activities of the watershed.

While some potential recharge studies have been conducted, the majority of the studies in this watershed have focused on soil loss, LULC, floods, and groundwater potential (Tilahun & Merkel 2009; Sitotaw & Hailu 2019). But, there is no research conducted in the area that identifies potential groundwater recharge zones. Thus, the main objective of this study is to identify high-potential groundwater recharge zones using geospatial technology and the AHP approach. Decision-makers, government agencies, and the commercial sector will be able to use the findings to design sustainable groundwater management strategies and prioritize intervention activities in high-impact areas of the watershed.

The topographic features of the study area exhibit a wide range of spatial variability. Mountainous topography is found in the southern, southwest, and southeast regions, which have an altitude of 2,400 masl and slopes of up to 74°. The majority of the territory's northern, northeastern, and northwest regions are flat plains with slopes up to 0–6° at an altitude of 900–1,000 masl. The Dire Dawa town is situated in the plain terrain part of the watershed. The study area's lithology is made up of a variety of rock formations. Precambrian (undivided) sandstone, Triassic and Permian Adigrate sandstone (lower sandstone), Jurassic Antalo limestone, and Cretaceous upper sandstone make up the bedrock. Holocene and Tertiary were intrusive. The alluvial deposit extends across the vast majority of the northern plain.

A digital elevation model (DEM) with a resolution of 12.5 m was acquired from Alaska to create slope and drainage density maps using the ArcGIS tool. This section covers the data sources, formats, and data processing methods used for each parameter in the assessment of prospective groundwater recharge zones.

Sink filling, flow direction, flow accumulation, and stream network extraction can all result in drainage density. The drainage density map was made using a line density analysis program. The United States Geological Society (USGS) created lineament density and LULC maps using the Landsat 8 Operational Land Imager (OLI) satellite picture and used them in this work (USGS 2019). Lineament was automatically extracted using the PCI Geomatica line module. The Ethiopian Ministry of Water and Energy provided the groundwater borehole location data, while the global geologic maps website (https://certmapper.cr.usgs.gov/data/apps/world-maps/) provided the geological data. Soil maps for the research area were created using information from the Harmonized World Soil Database (HWSD). After the theme map was generated, weights were given to factors affecting the presence and movement of groundwater.

Similar to this, a set of competitive criteria for groundwater recharge investigations can be employed to use AHP to identify potential recharge regions. To determine the likely rechargeable locations within the Shinile watershed, the AHP methodology is implemented in four steps: (1) identification of criteria affecting groundwater recharge zones, (2) pairwise comparison matrix, (3) relative weight evaluation, and (4) evaluating matrix consistency. Each stage is described in depth separately as follows.

The selection of the main affecting elements and their characteristics is the initial step in the AHP approach. A pyramid structure made up of the purpose, influencing variables, and various potential solutions to the problem is created by the researcher by selecting the factors (Saaty 1980a, 1980b; Boroushaki & Malczewski 2008; Brunneli 2015). The research's primary goal forms the pyramid's apex, followed by factor attribute and choice solutions at the base.

The primary goal of this study is to define the Shinile catchment's prospective groundwater recharge zone. Factor qualities made up of thematic layers include aquifer geologic formation, soil type, drainage density, lineament density, LULC, and slope were collected, analyzed, and classified. The classification layer of the influencing variables identifies the potential recharging zones based on each factor. As stated in Table 1, the relative significance of the parameter on groundwater recharge in this study was rated on a nine-point scale. A score of 1 indicates that two aspects are equally important, whereas a score of 9 indicates that the rowing motif is significantly more important than the column factor (Saaty 1980a, 1980b).

It is based entirely on the wide range of entering criteria for the definition of recharging capability zones, the pairwise contrast matrix A (meter*meter) was constructed using the concept of Saaty (1980a, 1980b). Thematic layer maps that are mentioned in the Results and discussion section have been categorized and added to the inputs. Each matrix's entries show how the column aspect and row aspect interact. Using Saaty's one-to-nine factor scale of Table 1, factor influences influence each other.

According to Freeze & Cherry (1979), geology plays a crucial role in the incidence and distribution of groundwater in any terrain, so the selection and weighting of factors for the development of the pairwise comparison matrix (PCM) (Table 2) for delineating recharge zones changed to be based primarily on geology and methods affecting recharge in the Shinile watershed (Dabrala et al. 2014). Because water would directly recharge aquifers, lithology (outcrop) was converted into the first parameter and recorded in row 1 and column 1 of the matrix. The slope was selected as the second most important parameter because it has an impact on how water gets to the surface, which was one of the criteria used to reclassify the recharge potential parameter. The third option was lineament density because a high lineament density results in porosity and permeability. The soil type is ranked fourth in the hierarchy because it affects permeability and infiltration capacity, whereas soil thickness affects infiltration volume. The lowest value was chosen because the drainage density depends on the other variables that were examined.

One of the factors used to reclassify the recharge potential parameter, the slope, was chosen as the second most crucial characteristic since it affects how water reaches the surface. Lineament density was the third possibility since a high lineament density causes porosity and permeability. Since soil type impacts permeability and infiltration capacity while soil thickness affects infiltration volume, soil type is rated fourth in the hierarchy. Because the drainage density depends on the other factors that were looked at, the lowest value was picked. The lithology/drainage density pair was given the number 9 because, in the absence of aquifer outcrops, water can reach the lineaments in locations with concave slopes, thick soils, and good drainage. To parameters of equal importance, one was allocated.

To reduce over-fitting and reduce noise in the statistics, AHP eliminates functions that have a strong connection between them. It also uses the opinions of subject-matter experts and the eigenvector and eigenvalue principle (Dabrala et al. 2014). Primary eigenvalue techniques are used to rank the elements, and expert judgement and eigenvector to determine the component weight. (Malczewski 2006; Taylor & Doug 2007; Hajkowicz & Higgins 2008).

The eigenvector is a parameter ordering that influences recharge by distributing weights (Saaty 2003). The relative weights of each parameter in the direction of recharging were computed using the eigenvector (Brunneli 2015). The eigenvectors in column 8 in Table 3 were calculated by first dividing column values by the column sum in Table 2 and then averaging row values (Saaty 1980a, 1980b).

The major eigenvalue (max), which is the sum of all the eigenvalues, is a measure of the degree of matrices deviating from consistency, and it is used by AHP to prioritize the relevance of parameters to recharge (Brunneli 2015). A pairwise comparison matrix is said to be steady simplest if the major eigenvalue (max) is more than or equal to the number of parameters being evaluated (n); otherwise, a new matrix is needed. The product column sums in the pairwise matrix (Table 2) and eigenvector (Table 3) were used to get the main eigenvalue for the last row in Table 4. The calculation of the consistency index (CI) was made possible by the completion of the 6.52 major eigenvalue for a 66 matrix.

CR is less than 10%, hence, it is acceptable and thus weighted overlay analysis was performed in ArcGIS software to integrate the weighted map layers and to map the spatial distribution of recharge zones.

Weighted overlay analysis (Esri 2014) is the process of entering components into an integrated analysis based on the pairwise comparison matrix of AHP using the same scale of values. Each input layer was reclassified using a single ratio scale with very high, high, moderate, low, and very low recharge potential. The relative weight of each groundwater recharge has been calculated using a pairwise comparison matrix of AHP that takes into account the reclassification criteria. The weighting values had been examined for consistency before being utilized to produce a map of the spatial distribution of recharge inside the Shinile catchment, as demonstrated in Section 2.3.1.

A map of the spatial distribution of groundwater recharge within the Shinile catchment was produced using the weighted overlay tool of ArcGIS 10.4 software by combining the reclassified layers of lithology, slope, lineament density, soil type, LULC, and drainage density with their corresponding percent effects on recharge. With the help of the Weighted Overlay analysis tool, values in the input raster layers are reclassified into the following categories: very high, high, medium, low, and very low. To do this, a map of possible recharging zones is created by multiplying the cell values of each factor magnificence by the factor weight and then combining the resulting cell values collectively (Esri 2014; Raviraj et al. 2017).

The southern portion of the study area is made up of Precambrian (undivided), which is uncomfortably overlain by Triassic and Permian Adigrate sandstone (lower sandstone), Antalo limestone (Jurassic), and a cretaceous upper sandstone. The northern portion of the study area is primarily made up of Tertiary intrusive and Holocene alluvial deposits. Depending on the likelihood that a lithological unit may contain groundwater, different weights have been assigned to each one. Alluvium was given a higher score than sandstone.

The slope is the angle formed at a certain place on the topo surface by the tangent plane and the horizontal plane (Maidment 1993). The slope is connected to the local and regional topography, which has a big impact on how groundwater moves and gets recharged into aquifers (Gupta & Srivastava 2010). Different slopes with imperfections ranging from very steep to soft define relief. The amount of infiltration and runoff is determined by the slope (Hewison & Kuras 2005).

Five drainage density categories in the research region were established based on the findings: extremely low, low, medium, high, and extremely high (Figure 5). Prioritization was given to the categories with the lowest drainage densities, which were low, medium, high, and very high, in that order.

According to O'Leary et al. (1976), lineaments are the surface depiction of fundamental geological and structural characteristics like fractures, faults, and joints. Lineaments were derived automatically from ALASKA-DEM (Jarvis et al. 2008) and Landsat 8 OLI July 2020 images and manually from aeromagnetic data (Botswana Geoscience Institute (BGI)). The USGS Earth Explorer website (https://earthexplorer.usgs.gov) provided January 2020 Georeferenced Landsat-8 OLI and thermal infrared sensor (TIRS) Geotiff data for download. Useful for extracting geological formations and rock qualities is band 8 (0.50–0.68 m) (Dyke et al. 2018).

Soil plays a major role in controlling surface water infiltration and percolation into the aquifer (Anbazhagan et al. 2005). A soil map of the region was made using the HWSD (FAO/IIASA/ISRIC/ISS-CAS/JRC 2012). Four main soil textural classes: heavy clay, clay loam, loam, and sandy clay loam have been established.

Agricultural, Grassland, Shrubland, Human Settlements, Acacia, Bare Land plus Desert Acacia and Disperse Shrub, Desert Sand, and Bare Land habitats have all received lower ratings than Forestland, which is followed by each of these categories.

Using the groundwater recharge potential zones map weighted overlay approach, the study region was categorized into five classes: very high, high, moderate, low, and very low (Figure 9). There are very high to moderate groundwater recharge zones in the northern and northwest region, which is covered in woodland, farming, and desert sand. The eastern and outer ridge of the research region, which is largely covered by bare land and desert sand, Precambrian (undivided), Antalo limestone (Jurassic), and cretaceous upper sandstone with a very high slope, has low to very low groundwater recharge zones.

The overlay analysis-created map of prospective groundwater recharge zones was compared with the deep borehole data already in existence to ensure that the AHP-guided technique was successful in identifying potential groundwater recharge zones (Figure 9). It has been observed that current well fields correspond to high recharge zones, as indicated in Figure 9. A total of 41 deep borehole data were used for validation, with the majority of them located in the sub-northern basin's region. Due to geography and little population settlement, data availability in the southern part of the sub-basin is constrained. Furthermore, the uneven distribution of point data made it impossible to plot the groundwater level. As a result, the weighted index overlay analysis was used to solely overlay the point data from the drill well to test the validity of the groundwater potential map.

The overlay analysis has shown that each borehole well has layers of high and extremely high groundwater recharge potential zones. The borehole is critical to determining, managing, and conserving groundwater recharge. According to several studies (Lentswe & Molwalefhe 2020; Tolche 2021), possible zones for groundwater boreholes can serve as a proxy for potential zones for groundwater recharge.

Different research indicated that integrating GIS and RS with the analytical hierarchical process is an efficient technique for the identification of groundwater recharge and potential zones. Factors like geology, slope, lineament density, soil texture, LULC, and drainage density were used to prepare thematic layers using AHP. The analysis indicated that the groundwater recharge potential zone is mapped into five classes: regions with very high, high, medium, low, and very low recharge potentials. The northern and northwest region of the study area, covered in forest, farmland, and desert sand, has very high to moderate groundwater recharge zones. The eastern and outer ridge of the study area, largely covered by bare land and desert sand, with a very high slope, has low to very low groundwater recharge zones.

It can be seen that the groundwater recharge potential maps predicted in this study overlaid with the current well locations which indicates a very high level of accuracy. The study's findings can be used as a point of intervention by water planners and policymakers to build soil and water conservation structures that will increase the catchment's capacity for recharging. This study can be used for the implementation of watershed management practices to improve groundwater recharge and thus reduce water scarcity in the area. The methodology used in this study can be useful in other areas where groundwater recharge point identification is required for sustainable water management.

No specific grant was given to this research by funding organizations in the public, private, or not-for-profit sectors.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.