Numerical groundwater modelling under changing water abstraction in Weyib watershed, Ethiopia

Abstract Groundwater is the primary source of water supply in Ethiopia. The study area was challenged due to increasing water demand, uneven water resource distribution, and noticeable changes in groundwater levels. The study focused on the examining of existing abstraction and future water demand scenarios on groundwater balance in the Weyib watershed using the WetSpass-M and MODFLOW-2005 models. The input datasets, such as aquifer properties, observed groundwater heads, hydrogeology, groundwater recharge, the Digital Elevation Model (DEM), and hydrological data were used. Datasets were prepared to better represent subsurface hydrology and its future demand effects evaluated using calibrated steady-state numerical groundwater modeling. The WetSpass-M and MODFLOW-2005 models depicted good performances during the simulations of groundwater recharge and groundwater budget under existing abstractions and estimated demand scenarios, respectively. The mean annual groundwater recharge estimated was 177.66 mm/year. The existing groundwater abstraction was 34,686.39, estimated short-term and long-term water demand scenarios were 72,113.61 and 93,795.57 m3/day, respectively. The upstream area has the highest groundwater head and recharge, while decreasing as it approaches the Weyib watershed outlet. During the outflow groundwater budget, the groundwater abstractions increased as expenses of river leakage and head dept. bounds increased. Moreover, the increasing groundwater withdrawal would reduce groundwater heads, and the estimated future water demand scenarios would substantially impact the groundwater budget, which would also have an impact on the watershed hydrology and ecosystem.


Introduction
Groundwater is the primary source of freshwater demanded by the world's population (Eslamian et al., 2023).Groundwater is the most precious natural resource because it is more safeguarded from pollution and less impacted by seasonal variations than surface water comprised of rivers, swamps, lakes, etc. (Daniel et al., 2022;Sisay et al., 2023;Zammouri et al., 2014).Globally, groundwater heads were dropping, and the magnitude of this decline was not well understood (Akter & Ahmed, 2021).Due to increasing population growth, urbanization, climate change, and human activities, which pose a significant impact on hydrology and sustainable water resource management (López-Valencia, 2019;Mengistu et al., 2021;Pathak et al., 2019).Groundwater withdrawal lowers storage and baseflow, likewise lowering streamflow and affecting environmental low flow rates in dry periods (Gleeson & Richter, 2018).
Eventually, Africa is predicted to face water stress before 2025, owing primarily to rising water demand due to population growth (Mamo et al., 2021).Groundwater supplies approximately 80% of the domestic water supply for rural and urban residents throughout all climatic zones, which contributes to the socioeconomic advancement of Ethiopia (Gebere et al., 2020;Mengistu et al., 2021;Tolera & Chung, 2021).However, groundwater has been extensively abstracted without indepth study in Ethiopia (Azeref & Bushira, 2020).Even though Ethiopian development relies on agricultural-led industrialization, this was challenged due to uneven spatiotemporal rainfall distributions and hydrological extremes like drought and flood (Mamo et al., 2021;Taye et al., 2021;Tefera, 2017).To satisfy household, commercial, and livestock water demands, the community is dependent on groundwater abstraction, which has a significant impact on the aquifer system (Mengistu et al., 2021).Ethiopia is dealing with rising water demand and water use disputes caused by declining groundwater recharge and baseflow (Fantaye et al., 2023;Goshime et al., 2021;Mamo et al., 2021;Tefera, 2017).Although understanding groundwater resources contributes to the country's development, this has not been adequately studied due to data limitations.Consequently, numerous groundwater development projects fail, leading to economic losses since the role of subsurface water in water-related developments was not previously envisaged on a national scale (Mengistu et al., 2021).
To guarantee sustainable groundwater availability in the region, quantification of spatiotemporal recharge variation and assessing the effect of abstractions on hydrology had to be tackled (Ayenew et al., 2013;Chunn et al., 2019;Morsy, 2023).Estimating groundwater recharge was required to preserve the equilibrium between recharge and abstractions and to comprehend the controlling factors affecting spatiotemporal groundwater recharge (Yifru et al., 2020).The soil, land use, rainfall, and climatic variables together had an impact on groundwater recharge.In humid areas, a noticeable increase in rainfall intensity would lower recharge because substantial amounts of rainfall intensity may exceed the infiltration capacity; nevertheless, it may increase recharge in arid regions (Bates et al., 2008;Mamo et al., 2021).Furthermore, unlike climate or hydrological data, there is no straightforward method for measuring groundwater recharge, and instead, indirect estimation methods were commonly used (Tolera & Chung, 2021).
Nowadays, hydrological models are crucial for efficiently identifying, estimating, and understanding hydrological dynamics (Aredo et al., 2021b;Bizhanimanzar et al., 2018;Sutanudjaja et al., 2011).These models are implemented for the analysis and quantification of natural and anthropogenic activities' effects on hydrology and to recommend suitable insight for sustainable water resource management practices and policies (Jafari et al., 2021).For instance, the WetSpass-M physical-based distributed hydrological model had promising performance in estimating groundwater recharge and water balance in Ethiopian watersheds (Demissie et al., 2023;Gelebo et al., 2022;Molla et al., 2019).Also, the MODFLOW model depicted outperformed performances in numerical groundwater modeling and groundwater hydrology response to changing human and anthropogenic activities (Aliyari et al., 2019;Chunn et al., 2019;Daniel et al., 2022;Gobezie et al., 2023;Gropius et al., 2022;Molina-Navarro et al., 2019;Qin, 2021).MODFLOW-2005 is a 3D finite-difference groundwater model developed by the U.S. Geological Survey Department (Harbaugh, 2005).The MODFLOW-2005 model simulates groundwater budget, pumping, and interactions among surface and groundwater based on defining packages and classifying sub-surface aquifer systems into finite-difference cells (Gao et al., 2019;Harbaugh, 2005;Kirubakaran et al., 2018).Due to a lack of time-series groundwater level observation data, most studies selected a steady-state groundwater modeling instead of a transient-state (Aghlmand & Abbasi, 2019).Understanding the hydrogeology was a vital step in portraying subsurface hydrology in the modeling procedures (Daniel et al., 2022;Sisay et al., 2023).
In general, the study area was challenged by erratic precipitation and declining agricultural production due to increasing water demand and population growth.As a result of this, noticeable groundwater level fluctuations have been observed in the downstream areas, which are susceptible to drought (Aredo et al., 2021a;Awulachew et al., 2007;Kassahun & Mohamed, 2018;Serur & Sarma, 2018a, 2018b;Shawul et al., 2016).
Earlier studies in the Weyib watershed focused mostly on surface hydrology.However, a few research has concentrated on the estimation of potential groundwater recharge at the basin scale.The impact of current abstraction and forecasted water demand on the groundwater budget has not been studied in the Weyib Watershed.To perform transient state groundwater modeling, there is a lack of time-series groundwater level observation data in the Weyib Watershed.Moreover, this study selected the steady-state numerical groundwater MODFLOW-2005 model to precisely represent aquifer and subsurface hydrology.However, no prior study has practiced numerical groundwater modeling under changing groundwater abstractions in the Weyib Watershed.The main objectives of this study include: (1) estimation of groundwater recharge using the WetSpass-M model; (2) MODFLOW-2005 model performance evaluation using collected groundwater observation heads data; and (3) numerical groundwater modeling with existing abstractions and estimated short-term and long-term water demand scenarios.

Description of the study area
Weyib River drains roughly 3611 km 2 in Ethiopia's Genale Dawa River Basin lies between 6.83° to 7.46°N latitude and 39.53° to 40.50°E longitude (Figure 1).The topography ranges from 1739 m in the lower reach to 4346 m above mean sea level in the highland of Bale Mountains National Park.Annual rainfall in the watershed ranged from 851.04 mm to 1341.23 mm, with bimodal rainfall patterns.The mean annual maximum and lowest temperatures are 22.23°C and 7.28°C, respectively.The major tributaries of the watershed such as the Shaya, Tegona, and Tebel Rivers.Hydrogeological formations of the Genale-Dawa River Basin, especially in the Weyib watershed were dominated by volcanic rocks and extensive aquifers with fracture permeability (Kassahun & Mohamed, 2018).The geological classes were broadly classified as quaternary volcanic and sediments, and tertiary volcanic successions.The quaternary volcanic and sediment classes were Scoraceous vesicular olivine phyric basalt formation with estimated aquifer productivity ranges from high to moderate in the central and downstream areas of the watershed.Tertiary volcanic successions were divided into five groups: Alkali trachyte and basalt flows with moderately productive aquifer; Alkali trachyte flows with low aquifer productivity; Teltele basalt flows and lower flood basalts range from high to moderate productive aquifer; and Nazeret Group (Stratoid silica-ignimbrites, tuffs, ash, rhyolites, trachyte, and minor basalt) ranges from moderate to low productive aquifer (Figures 2 and 3).

Data collection
This study collected both primary and secondary data as input for modelling.Primary groundwater head data was measured at wells using a deep-water level meter in the Weyib watershed.Historical climate data for weather stations such as Dinsho, Gasera, Ginir, Goba, Goro, Hunte, Sinana and Agarfa were collected from the National Meteorology Agency, Ethiopia.Existing groundwater abstraction data for domestic demand was collected from the Bale Zone Water and Energy office.The baseline population of the study area was collected from the Ethiopian Statistics Service.The Weyib River streamflow, Genale-Dawa River basin hydrogeological and geological maps were collected from the Ministry of Water, Irrigation and Energy.The Landsat 8 OLI/TIRS C1 Level-1 satellite images from Path 167 and 168, and Row 55 and 56 with no clouds were downloaded and used to create the watershed's land-use map.The maximum likelihood pixel-based trained technique using supervised image classification by ERDAS Imagine software.The soil types were downloaded from the FAO database and converted to a soil texture map depending topsoil percentage of particle size fractions.The digital elevation model (DEM) was downloaded from the U.S. Geological Survey Department's official website (https://earthexplorer. usgs.gov/).The slope map was prepared using the DEM map of the study area using a spatial analysis tool.

Groundwater recharge estimation
WetSpass-M model is a physical-based distributed hydrological model for monthly, seasonal, and annual time steps in watershed or basin scale was applied to estimate groundwater recharge (Batelaan & De Smedt, 2007;Gelebo et al., 2022;Molla et al., 2019).The input data used were soil, DEM, slope, land use, temperature, streamflow, baseflow, wind speed, rainfall, evapotranspiration, and groundwater heads.Stations-based meteorological data points were interpolated using the IDW technique to have spatial scale.The Weyib River streamflow was separated into baseflow and direct runoff using the IHACRES algorithm embedded in the HydroOffice tool.Separated baseflow and direct runoff were used for model performance evaluations during the estimation of groundwater recharge.The WetSpass-M model performances was evaluated using statistical indices such as coefficient of determination (R 2 ) and Nash-Sutcliff Efficiency (NSE).The values of NSE range from -∞ to 1; the higher the value the better agreement between the observed and modelsimulated values depicted in Equation 1 (Moriasi et al., 2007;Nash & Sutcliffe, 1970).Where; O i is the i th observed value, P i is the i th model-simulated value, Ois the mean of observed data, Pis the mean of the model-simulated value and n is the number of observations.

Groundwater modelling techniques
Groundwater process conceptualizing was required for handling modelling procedures, which commenced by characterizing data gathered to figure out the subsurface water budget (Kirubakaran et al., 2018).MODFLOW-2005 model requires inputs such as a spatial recharge, DEM, aquifer properties, geological, hydrogeological, hydrological, and groundwater heads data.The MODFLOW-2005 model was defined with 128 rows, 206 columns, 500 m by 500 m for grid size, and the model extent depicted with 64,125 and 103,010 for row and column sections, respectively.The Weyib watershed shapefile was assigned as one value for active cells, and outside areas as zero called inactive cells.The watershed's top layer elevation was defined depending on the extracted from DEM, and these were assumed as the model top elevation.

Boundary conditions
Based on few studies characterized hydrological, hydrogeology and geological maps of the Genale Dawa River basin (Kassahun & Mohamed, 2018).This aquifer productivity and geological formations were used to set boundary conditions in the MODFLOW-2005 model.The watershed's western areas were covered by Alkali trachyte flows geological formations, the southwest parts, such as the Dodecha and Selka areas in the Sinana district, and the northern areas like Gasera and Agarfa, have high to moderate aquifer productivity, which was defined as General Head Boundary (GHB) package.The WetSpass-M model simulated the spatial groundwater recharge and applied in the top-active layer as a recharge package in the MODFLOW-2005 model.The well package was defined depending on collected well data such as aquifer properties, borehole locations and pumping rates.The pumping rates of the wells range from 26 to 2084 m 3 /day.The river package was used to simulate major rivers such as Weyib, Tegona, and Shaya.In Ethiopia's Abaya Chamo lakes basin, the study assumed the river head in the cell was 0.5 to 2 m as goes to the downstream area below the mean elevation in the cell (Daniel et al., 2022).The study followed similar assumptions stated above with consideration of elevations for each cell and elevation data extracted from the DEM for the Weyib watershed.Additionally, the study used unpublished documents from the Genale Dawa River basin (GDRB) master plan such as aquifer characteristics, hydrogeological, hydrological and geological maps, to define boundary conditions for groundwater modelling (Figure 4).

Model calibration and simulation
Initial hydraulic conductivity values were assigned for each geological class defined in the GDRB master plan document and literature review (Kassahun & Mohamed, 2018), and then hydraulic conductivity was reconfigured during the calibration process.The model's performance and accuracy rely upon the input datasets and modelling approach (Daniel et al., 2022).The MODFLOW-2005 model was calibrated by adjusting hydraulic conductivity parameters till it achieved adequate convergences among measured and simulated groundwater heads using the statistical indices.Measured groundwater heads in the 53 boreholes throughout the Weyib watershed were defined as head observation packages.The calibrated MODFLOW-2005 model was quantified to assess the impact of existing withdrawal, short-term and long-term water demand scenarios on the groundwater budget.

Model performance evaluation
Model performances were evaluated by the coefficient of determination (R 2 ), mean error (ME), and mean absolute error (MAE) with a comparison of measured and simulated heads (Equations 2, 3 and 4).The R 2 values range from -∞ to 1; If the results close to 1 depict the highest goodness of fit among measured and model-simulated (Moriasi et al., 2007).The ME states discrepancy among, and the MAE depicts the mean absolute values (Azeref & Bushira, 2020;Chai & Draxler, 2014;Tolera & Chung, 2021).
Where; h o and O i are observed heads at the i th locations, h s and P i are the model simulated heads at the i th locations, O is the mean of observed, P is the mean of the model-simulated and n is the sum of observations wells.

Existing abstraction scenario
The existing abstraction scenario considered 2022 as a baseline period scenario to compare forecasted water demands in the Weyib watershed.The actual water supplied for the existing demand was 34, 686.39 m 3 /day and collected from Bale Zone water and energy offices.

Estimated water demand scenarios
The estimated water demand scenarios were prepared depending on the forecasted population for short-term and long-term scenarios.The study area's population growth rate was 2.13% from 2020 to 2050 (Serur & Sarma, 2018b).The baseline population was 1.2 million in the 2021 year (Ethiopian Statistics Service, 2021).Ethiopian Statistics Service proposed a population projection method for Ethiopia case based on an exponential growth rate using Equation 5 (Haregeweyn et al., 2012).
Where P t is the population forecasted at a specified period, P o is the baseline population, e is the natural logarithm base, r is the annual growth rate, and t is the base and forecasted period time difference.

Short-term water demand scenario.
In the estimated short-term water demand scenario, ranging from 2026 to 2035, the estimated population and water demand will be 1.65 million and 72,113.61m 3 /day, respectively.

Long-term water demand scenario.
In the long-term water demand scenario, the period ranges from 2036 to 2045, 2.05 million for the forecasted population and the estimated water demand will be 93,795.57m 3 /day.Generally, Figure 5 shows the methodological conceptual framework followed to estimate the groundwater budget under changing existing abstractions and forecasted water demand scenarios using WetSpass-M and MODFLOW-2005 models.

Groundwater recharge
Mean annual groundwater recharge estimated using the WetSpass-M model in the Weyib watershed.In a comparison of observed and model-simulated mean monthly baseflow and runoff, the model performance was 0.90 and 0.85 for R 2 , and 0.95 and 0.89 for NSE, respectively.Comparable WetSpass-M model performances were reported (Molla et al., 2019).The annual groundwater recharge ranges from zero to 560 mm with a mean value of mm/year (Figure 6).Likewise, the annual average recharge varied significantly throughout Ethiopia (Demissie et al., 2023;Gelebo et al., 2022;Mengistu et al., 2021;Molla et al., 2019).Spatial groundwater recharge was applied in the top-active layer of the MODFLOW-2005 model as a recharge package.

MODFLOW-2005 model performance assessment
The model parameters were altered to achieve a good match between measured and modelsimulated heads using the trial and error method (Azeref & Bushira, 2020).The MODFLOW-2005 model performance evaluation was 0.995 for R 2 , −1.204 for ME and 10.503 for MAE in the Weyib watershed.The hydraulic conductivity parameter's sensitivity was analyzed by raising and reducing it by 10, 25, and 50% in comparison to the difference between measured and simulated groundwater heads.The model sensitivity analysis depicted that it was highly sensitive to 50%, moderately sensitive to 25% and less sensitive to 10% during the drop in hydraulic conductivity values.Rising hydraulic conductivity values of 10 and 25% were less sensitive, whereas 50% were moderately sensitive.The comparison between the observed head and the model-simulated heads depicted a good match (Figure 7).Furthermore, the model performance assessments applied for measured and model-simulated groundwater heads in the Weyib watershed were depicted as statistically sound.Comparable model performances observed in Ethiopia's Modjo River catchment (Gebere et al., 2020), Kombolcha catchment (Azeref & Bushira, 2020), Abaya Chamo lakes basin (Daniel et al., 2022) and Borkena catchment (Gobezie et al., 2023).

Simulated contours groundwater head variations
The highest simulated groundwater head was depicted in the Weyib watershed upstream areas, while the simulated head dropped toward the watershed outlet and reached the minimum close to the Alemkerem gauging stations of the Weyib River (Figure 8).

Water balance of Shaya and Tegona catchments
As per the modeling result, Shaya and Tegona catchments contribute a significant percentage to the Weyib watershed groundwater budget.Shaya Catchment had a greater weightage of the total groundwater budget and recharge than Tegona Catchment.The horizontal exchange of the Shaya catchment was higher compared to the Tegona catchment.In the Shaya catchment, the findings of the groundwater budget were mainly dominated by a horizontal exchange toward the Weyib watershed (Table 1).Tegona catchment's groundwater budget such as groundwater recharge, head dep bounds, and horizontal exchanges shares the highest percentages (Table 2).

Impact of water abstractions on the groundwater budget
The groundwater budgets were simulated under changing existing abstractions, short-term and longterm water demand scenarios in the Weyib watershed.The existing groundwater abstractions estimated was 34, 686.39 m 3 /day.Depending on population projections, the estimated short-term and long-term water demand scenarios were 72,113.61and 93,795.57m 3 /day, respectively.The simulation of the groundwater budget under existing abstractions depicted that the total inflow and outflow to the groundwater system was 2.36 Mm 3 /day (Table 3).The groundwater balance differences between inflow and outflow were −1.98E +03 m 3 /day and percentage discrepancies close to zero.During existing abstractions were 2.11E +04 m 3 /day and 2.06 Mm 3 /day for river leakage to inflow and groundwater leakages to river flow, respectively.The short-term groundwater budget differences increased relative to existing abstractions with −4.62E +03 m 3 /day with a −0.20% discrepancy.During short-term water demand scenarios, the outflow of river leakage and head dep bounds were decreased in comparison to existing abstractions (Table 4).The impact of long-term water demand scenarios on the groundwater budget significantly increased wells withdrawal and head dep bounds, and the inflow-outflow difference was −5.70E +03 m 3 /day with a −0.24% discrepancy (Table 5).Moreover, as groundwater pumping increased the river leakage outflow were decreasing with 2.06, 2.03 and 2.01 Mm 3 /day for existing, short-term, and long-term scenarios.The groundwater budget showed that head dep bounds inflow was greater than outflow and eventually, the river leakage inflow less than the outflow during existing, short-term, and longterm scenarios.The groundwater abstractions increase with the expenses of river flow and head dep bounds during existing, short-term, and long-term scenarios.The groundwater modelling depicted that groundwater heads and recharge were high in upstream areas, whereas low in the downstream and central areas in the Weyib watershed.The study provides insights for sustainable groundwater resource development and management to address a drastically increasing water demand.The groundwater budget differences were −1.069E +05 m 3 /day during the baseline scenario using the MODFLOW model in Rib catchment, Ethiopia (Mamo et al., 2021).Future population projections could end up drastic increase in water demand, thereby requiring to  decrease in water supply distribution system losses (Tefera et al., 2023).The studies highlighted that groundwater abstraction was drastically increasing, and substantial water supply distribution losses in urban areas, and recommended to decrease distribution losses.Similarly, High river leakage and head dep bounds outflow were depicted in Ethiopia's upper Awash sub-basin (Birhanu et al., 2021), Little Akaki watershed (Tolera & Chung, 2021), and Modjo River (Gebere et al., 2020;Sisay et al., 2023).In Ethiopia's watersheds groundwater abstractions cause considerable effects on the groundwater budget (Birhanu et al., 2021;Fantaye et al., 2023;Tefera et al., 2023;Tolera & Chung, 2021).Simulation of the groundwater budget under existing and estimated groundwater abstractions would be useful as a reference for experts, researchers, policy, and decision-makers in the region.Further research outlook is needed to integrate surface and groundwater modelling under consideration of factors such as natural and human activities using the MODFLOW-2005 model.

Conclusions
The study aimed to estimate the groundwater budget under existing abstraction and estimated water demand scenarios in the Weyib Watershed.The WetSpass-M and MODFLOW-2005 models outperformed when evaluated by statistical indices.The annual groundwater recharge ranges from zero to 560 mm with a mean of 177.66 mm/year.The existing abstractions were 34, 686.39 m 3 /day, and future short-term and long-term scenarios were 72,113.61and 93,795.57m 3 /day, respectively.The groundwater budget differences were −1.98E + 03, −4.62E + 03, and −5.70E +03 m 3 /day for existing abstractions, short-term and long-term scenarios, respectively.The simulated groundwater budget inflow was 2.36 Mm 3 /day nearly equal to the outflow under existing abstraction.The river leakage outflow was 2.06, 2.03, and 2.01 Mm 3 /day during existing, short-term and long-term scenarios, respectively.The head dep bounds outflow declined in the long-term scenario (0.262 Mm 3 /day) when compared to existing abstractions (0.267 Mm 3 /day).The highest groundwater heads and recharge were observed in the upstream areas, and these values dropped as approached the Weyib watershed outlet.Due to increasing short-term and long-term water abstraction scenarios a cause noticeable decrease in the river leakage and head dep bounds outflow.Moreover, increasing groundwater abstractions would decrease in water budget, which would also have an impact on the sustainable groundwater resources availability in the Weyib watershed.

Figure 2 .
Figure 2. Geological map of the Weyib watershed.

Figure
Figure 4. Boundary conditions definitions for groundwater modelling.

Figure 5 .
Figure 5. Methodological conceptual framework of the study.

Figure
Figure 7. Measured and modelsimulated groundwater head graphs.