Integrating Groundwater Modelling for Optimized Managed Aquifer Recharge Strategies

Abstract

Managed aquifer recharge (MAR) is a complex and hidden process of storing surplus water under the ground surface and extracting it as, when and where needed. Evaluation of the success of any MAR project is challenging due to uncertainty in estimating the hydrogeological characteristics of the subsurface media. This paper demonstrates the use of a groundwater model (MODFLOW) to evaluate a new, large-scale regional MAR project in the agricultural heartland in Punjab, Pakistan. In this MAR project, flood waters have been diverted to the bed of an abandoned canal, where 144 recharge wells (the wells for accelerating the recharge into the aquifer) have been constructed to accelerate the recharge to the aquifer. The model was calibrated for a period of five years from October 2015 to June 2020 on a monthly stress period and the resulting water levels were simulated till 2035. The water balance components and future response of the aquifer to different scenarios up to 2035 including with and without MAR situations are presented. The model simulations showed that MAR can contribute to the replenishment of the aquifer and its potential for the case study site to contribute significantly to the management of groundwater and to enhance supplies for intensive agriculture. It was further established that MODFLOW can help in the evaluation of effectiveness of a MAR scheme. This study is unique as it evaluates a significantly large MAR project in an area where this practice has not been developed for improving groundwater access for large scale irrigation. The model provides guidelines for decision makers in the region as well as for the global community and livelihood benefits for rural communities.

1. Introduction

Groundwater accounts for 99% of the earth’s freshwater reserves, providing vital potable supplies to a significant part of the rural population, particularly in semi-arid and arid regions of the world. In South Asia in particular, groundwater plays an increasingly important role in agricultural intensification thus ensuring greater food security for the burgeoning populations and also providing vital supplies for domestic and industrial needs. Yet, despite the vital contribution of groundwater to overall water security, the Sustainable Development Goals (SDGs) do not explicitly indicate the importance of groundwater to society [1], with the exception of SDG 6.6 which alludes to groundwater: By 2020, protect and restore water-related ecosystems, including mountains, forests, wetlands, rivers, aquifers and lakes. Despite this oversight, groundwater is linked to several SDG targets as it plays an indispensable role in food security, providing alternative supplies to manage during droughts and water scarcity, resilience to climate change and provides an important source of water in coastal areas where population centers are located. As a common pool resource, access to groundwater is the responsibility of individuals and its on-demand access is a blessing allowing farmers access to water for irrigation when surface water resources are not available; however, it also results in the overexploitation of the resource leading to unsustainable use of this vital natural resource. In [2], it was estimated that groundwater storage has been depleted by between 100 and 200 km³/yr. Of significant concern in South Asia and for that matter in many parts of the world where groundwater use is intensive, a significant challenge is the increasing contamination of fresh groundwater sources driven by anthropogenic activities mainly agricultural practices, increasing industrialization and domestic use. Additionally, geogenic sources such as arsenic and fluoride also pose a significant risk to human health particularly where groundwater is used without any knowledge of its quality aspects.

The primary water sources of Pakistan are generated from glacial melting, rainfall and groundwater [3]. Over 70% of Pakistan’s surface water resources are from the Indus River Basin (IRB), which accounts for water for 77% of the population of the country [4]. The IRB has an arid to sub-arid climate with an average annual rainfall of 240 mm. An extensive network of diversion structures, link canals, main canals, minor canals and watercourses is collectively called the Indus Basin Irrigation System (IBIS). It conveys water across Pakistan for the social and economic needs of the country, especially for irrigated agriculture which contributes 90% of Pakistan’s food production [5]. Surface water is now insufficient for Pakistan’s needs, and groundwater has become significantly important. The reduction in surface water supplies and the increased use of groundwater of marginal quality hints at a significant risk to Pakistan’s food security and its ability to provide food for a burgeoning population. Groundwater is the primary water source for many vulnerable communities, and its continued overexploitation is leading to a decline in groundwater levels and increasing risk of saline water intrusion into freshwater sources. This is particularly evident in densely populated cities (Lahore, Multan), and in the agricultural heartland of Punjab.

Amongst South Asian countries, Pakistan has become a large groundwater user with an annual extraction of 62 billion cubic meters (BCMs) from 1.4 million tubewells, supporting irrigation, domestic and industrial needs [6,7]. This exponential increase in groundwater use has occurred over the past 3 decades. The aquifer underlying the Indus River Basin (IRB), although large, has a negative water balance, implying that groundwater extractions exceed the annual recharging [8,9]. Floods in Pakistan, although often directly devastating to humans, can be a potential source of groundwater recharge either naturally or artificially [8,10]. Ref. [11] described a MAR case study in the Namoi Valley in Australia where groundwater is stored using flood water. They noted that not much attention has been paid to the use of flood water for groundwater recharge. Ref. [12] reported a MAR case study in Tamil Nadu, India, where recharge structures have been constructed including check dams and percolation ponds. They observed that water levels rose and water quality improved considerably, indicating the effectiveness of MAR. Ref. [13] reported a hydrological study of four streambed recharge structures, locally called check dams, in a monsoonal area of hard-rock terrain in Rajasthan, India, which was conducted over 3 years to evaluate their contribution to agricultural production. Net present value analysis for the net benefits of additional crop production attributable to recharge from check dams indicated that after accounting for average annual maintenance costs of 2.9% of capital costs, the benefit cost ratio of the four check dams averaged 4.1. Hence, this study shows that these check dams are economically attractive at the local level for securing irrigation water supplies through MAR. Ref. [14] experimented via a case study of MAR in Spain to evaluate the impact of clogging.

Research studies have shown storing of flood water as groundwater has a number of essential advantages when compared with surface water: as a rule, it is of higher quality, better protected from possible surface pollution including infection, not subject to evaporation, less subject to seasonal and perennial fluctuations and much more uniformly spread over large regions than surface water [15,16]. Ref. [11] provided an overview of the potential benefits and constraints for the execution of a MAR scheme in the Namoi Valley of Murray Darling Basin (MDB) in Australia using flood water and examined its social acceptability among the stakeholders at grassroot levels (farmers) by conducting field surveys. They concluded that MAR can be an opportunity for replenishment of the depleted aquifer, restoration of linkages between deep groundwater and surface flows and mitigation of drought impacts.

Two general approaches for groundwater management can address the overexploitation of groundwater: (i) demand-side management, i.e., reducing use and enhancing water use efficiency, and/or (ii) increasing supply through aquifer replenishment [17,18], including through managed aquifer recharge (MAR). MAR is “the purposeful recharge of water to aquifers for subsequent recovery or environmental benefit” [11,19]. Rural MAR projects can potentially increase water security for irrigated agriculture, reduce the impact of agricultural drought, with significant reduction in non-beneficial evaporation, improve groundwater quality in the vicinity of the MAR scheme and facilitate the conjunctive use of surface and groundwater resources. MAR schemes for recharging groundwater and its subsequent recovery in urban settings are widely used to manage water scarcity [9]. Agricultural MAR schemes designed to augment irrigation for intensive agriculture are less common, due to their complex nature and constraints such as the availability of land and the need to recharge significant quantities of water. The major prerequisite is availability of a suitable aquifer which depends upon many physical factors including size of aquifer, depth, permeability, stability and connections with other aquifers and ecosystems [11]. Constraints on MAR include concerns about impacts on groundwater quality, and the potential over-exploitation of the groundwater reservoir [11]. It is not surprising then that although the annual rate of MAR development worldwide is around 5%, this remains significantly behind the annual increase in extraction of groundwater [20]. Understanding how a complex MAR project is performing is important, and groundwater models can assist with this. A model is a scaled depiction of any physical phenomenon which elaborates the actual natural systems. Models can be used to study the behavior/response of a natural system under various field conditions. A groundwater model is a computer-based representation of the essential features of a natural geological system that uses the laws of science and mathematics [21,22,23].

This paper considers how MAR might be part of a toolkit to address the complex water challenges. It explores the case of an innovative technical research project on MAR undertaken by the Punjab Irrigation Department (PID) at the Old Mailsi Canal (OMC) in the Indus River Basin (IRB) in Pakistan. This study will be of particular interest as it uses a natural infrastructure, an abandoned canal (Old Mailsi Canal) and undertakes its rehabilitation with a series of recharge wells in the bed of the Old Mailsi Canal to improve the supply of groundwater for rural communities. The success of this project is crucial for improving the water security in Pakistan and its potential scaling out will be equally important as Pakistan faces a decrease in surface water from upstream diversions by India, as well as the increasing impacts of the changing climate. This rural MAR project has significant potential in improving supplies and water quality and thereby improving the food security situation in this region of south Punjab. This study is unique in the sense that (i) it evaluates a significant large rural MAR project in a country where MAR has not yet developed; (ii) the MAR project evaluated is the first of its nature in this region and (iii) MODFLOW has been used for the evaluation of a rural MAR project for the first time in Pakistan. The methodology developed can be applied to other similar conditions on regional and global level.

2. Materials and Methods

This study investigated the feasibility of a rural MAR site in the southern Punjab region of the IRB. A groundwater model (MODFLOW-GWVista 8) was developed to estimate the water balance components and future response of the aquifers to different scenarios up to 2035 with and without MAR situations. Previous groundwater model studies using MODFLOW are few in Pakistan and do not cover this area with reference to MAR investigation at a local scale [24,25]. The context and various inputs for the model are described in the following section. There is a possibility of mixed boundary conditions depending upon the hydrologic landscape of the area and models like MODFLOW are capable of handling such boundary conditions [26,27]. Refs. [28,29] provide an overview of successful cases studies of MAR in Australia.

2.1. Study Area

The case study includes the area around the MAR experimental site that is part of the Vehari district, Punjab province, Pakistan, as shown in Figure 1 The area is located between latitude 29.9719° N and longitude 72.4258° E, with an area of 1522 km². The elevation is roughly 140 m above sea level, and the area includes both rural and urban areas. The population of district Vehari is approximately 2.9 million, growing annually by just over 2.2%, with the majority of people relying on agriculture for their income [30,31]. The nearest large town is Multan, approximately 130 km from the study area.

The climate of the study area is generally hot and dry during the summer season that starts in April and continues until October. May, June and July are the hottest months, with mean maximum and minimum temperatures for these months of about 47 °C and 28 °C, respectively. The Vehari district is located within an agricultural region of the country, with limited industrial development. Vehari is an intensively cropped area, being the food basket for the province as well as being called the city of cotton [32,33]. As the average annual rainfall is less than 200 mm [34], water is supplied to the area by the Sutlej River and Pakpatan Canal. The river is subject to extreme variations in flow, with the mean monthly discharge during the summer about 15–20 times that of the winter [25]. Only 32% of irrigation requirements are fulfilled through non-perennial canal water and rainfall [35], so agriculture in the study area depends upon groundwater supply [36].

The study area was selected by the PID for a MAR project to augment the depleted aquifer [7,34,36,37]. Floodwater is to be diverted for MAR [25,36,38] using the existing Islam Headworks and newly constructed recharge wells in the bed of Old Mailsi Canal [39]. Due to the Indus Water Treaty (IWT) between India and Pakistan, according to which water of the Sutlej River was allocated to India, as a result, the Mailsi Canal system at Islam headworks has been abandoned since 1960. Groundwater flows from River Sutlej towards the center of the study area, as indicated by higher elevations of groundwater levels near the river. Extensive extraction of groundwater due to intensive agriculture has resulted in rapid depletion of groundwater levels in the area. There is adequate potential for the underground storage of flood water through MAR [40].

Basic steps in modelling include understanding the biophysical features of a system, identification of different sources and sinks, defining the boundaries of the system and demarcation of the hydrogeological/lithological conditions [22,41,42,43].

Three major factors which, in general, control the recharge to groundwater include the climate of the region, topography of the area and subsurface geologic formation. Depending upon the combination of these three major parameters, the controlling factors of recharge can be different from site to site and case to case and this leads to different boundary conditions of the groundwater models for a particular aquifer system where MAR is planned for implementation [44].

2.2. Rainfall Patterns

No weather station in the Pakistan Meteorological Department (PMD) is available in the study area; therefore, data for four surrounding stations Sahiwal, Multan, Bahawalpur and Bahawalnagar were obtained from the Pakistan Meteorological Department (PMD). The data was screened and filtered before analysis. For this study area, the rainfall data of the surrounding four stations (Multan, Bahawalpur, Bahawalnagar, Sahiwal) (Figure 2) indicates that there is very little rain during winter; however, droughts and floods occur due to climate variability.

2.3. Temperature Variations

Summer maximum temperatures are high, commonly exceeding 37 °C and at times exceeding 46 °C. Winter minimum temperatures seldom fall below freezing in the Indus Plain. The mean annual temperature is about 29 °C in the southwest part of the Punjab region. The temperature trends and fluctuations are shown in Figure 3 which indicates June-September are the hottest months of the year.

2.4. Lithological Description

Different organizations have drilled bore-logs in the study area from time to time (Ref. [45], Ref. [36]) and more recently by the Punjab Irrigation Department (PID) [25]. From these investigations, the subsurface lithology in the bed of the OMC is mostly sandy except the thin top layer, as shown in Figure 4, Figure 5 and Figure 6.

The surficial materials typically encountered along the bed of the OMC, through which the surface water flows, are of variable thickness ranging from 1 to 3 m, mostly alluvial, generally comprised of water-deposited silt, silty sand, interspersed with silty clay. These materials rest on the underlying sandy aquifer throughout the course of the canal. The silt particles in the top layer are well graded and generally not very permeable. By considering the extent and geological make-up of the surficial materials, the direct recharge of water from ground surface into the underlying sand aquifer through these materials is unlikely to be substantial due to the low hydraulic conductivity of the surficial materials.

2.5. Groundwater Model–Conceptualization and Development

The main objective of a groundwater model is to determine the behavior of an aquifer under different scenarios/conditions—natural or anthropogenic—in the future [46].

2.5.1. Conceptual Model

After identification of the physical parameters, the next step is to develop a conceptual model and translation of biophysical parameters into set of mathematical equations to represent the groundwater systems [42,47,48,49,50]. Once a model is calibrated, and sufficient level of confidence is gained that the model simulates the natural system very closely, it can be used for prediction [51], including the possible effects of a MAR system. Recharge (whether natural or managed) is an important component of groundwater systems, and it is either estimated and represented in the groundwater models or is estimated during the modelling exercise [27,52,53,54,55]. MODFLOW, developed by the USGS [26] with Groundwater Vista (GV 8) [56] was used in this study. GIS Software ArcGIS 10.6 was used to prepare the input files for the model and to simulate the aquifer behavior and predict the future trends in groundwater under different hydrological, agricultural and climatological situations. Figure 7 shows the groundwater modelling methodology. It depicts different hydrological processes showing various sources and sinks, aquifer geometry, subsurface lithological parameters and layers. The accuracy of the conceptual model is the baseline for development of numerical models which replicate field conditions [49,57,58,59].

PCG2 is the default solver in Groundwater Vistas, and it works well in most situations [56]. A major limitation to any modelling is the lack of information for some parameters. In this study, the limitations of different data sets are tabulated in

Table 1. Preliminary analysis of data availability and quality/accuracy of data sets.

Type of Data/Information	Remarks
Groundwater levels, elevations, locations	Poorly known, locations needed to be verified, some missing gaps, typo mistakes, only depth to water table data were available twice in a year. Some wells became dry. Data was collected from 2010 to 2020, but data only from 2015 to 2020 being reasonable could be used for model.
Aquifer layers/nature	No information was available. The aquifer was divided into two layers (first = 40 m, second = 260 m), keeping in view the depth of tubewells and piezometers. Based on the literature review, the aquifer was simulated as unconfined.
Aquifer parameters (Hydraulic conductivity, specific yield	No exact estimates were available from field tests; some initial estimates were obtained from bore-log analysis and literature review.
River stage, bed levels, thickness	Riverbed levels, width and thickness were obtained from the cross-sections and L-sections obtained from field offices during field-visits/surveys. But actual measurements at site may differ slightly due to meandering action of river.
Groundwater pumpage	Spatial distribution and direct estimates were not available. Pumpage was estimated based on available secondary data.
Climatic data (rainfall, temperature)	No weather station of PMD was available in the study area. Data of four surrounding stations, i.e., Multan, Bahawalpur, Bahawalnagar and Sahiwal were available to estimate the data for study area.
Water quality	No reliable and detailed data were available; therefore, sampling was performed to collect the samples of groundwater, canal/river and drain waters and analyzed in laboratory

.

2.5.2. Model Design and Extent/Geometry

Setting up the groundwater model requires that the project area be divided into sub-areas called nodes and cells. Both time constant and time variant data are required for each cell. For this case, keeping in view the hydrogeological boundaries, the model was extended beyond the PID experimental site of the OMC, with an area of 56 km by 47 km (2632 km²) selected for modelling as shown in Figure 1. A grid of 1000 m by 1000 m was sketched, with 56 rows and 47 columns to cover the whole area to be modelled. The aquifer was divided into two layers. Keeping in view the pumping depth and depth to water table in the study area, the first layer was kept at 40 m, and the second layer was 260 m, with a total aquifer depth of 300 m. As no evidence of bedrock in the model area was available from the bore-logs, the aquifer depth was assumed to be sufficient, such that the effects of groundwater extraction were away from the aquifer bottom [25]. After identification of all hydrological stresses, sinks, sources and aquifer geometry, pumping depth, depth to water table, thickness of aquifer, heterogeneity and boundary conditions, a conceptual model of the study area was developed and is shown in Figure 8. Groundwater Vistas [56] software package was used to convert a conceptual model into a numerical model. The numerical model was created in a series of processes/tasks. Model design and extent with grids/ boundaries and other basic information are shown in Figure 9 and model structure is shown in

Table 2. Model structure, extent and grids.

Sr.	Parameter	Inputs
1	Grid Area/Model Extent	No of Cols: 46 No of Rows: 57 Total area: 2632 km2
2	Grid cell size	1000 m × 1000 m
3	Basic Input Units	Meters for length, square meter for area, cubic meter for volume, days for time and then m3/day for flow rate
4	Model boundaries	East–South: Sutlej River North–East: Pakpattan-Islam Link Canal North–West: Pakpattan canal upper South–East: S-M-B Link Canal
5	Total aquifer thickness	300 m
6	No. of Layers	2
7	Type of aquifer	Unconfined
8	Specific yield (Sy)	0.15 to 0.30
9	Observation Wells/targets	21
10	Pumping tubewells	6000
	Diesel tube wells	3400
	Electric tube wells	2600
11	Simulation/calibration period	October 2015 to June 2020 (57 stress periods)
12	Initial heads/conditions/steady state	Post-monsoon 2015 (October 2015)
13	Time step	1 month (30.44 days)
14	Groundwater extraction	Agriculture, domestic, industry, aquaculture
15	River package	Main canals, Sutlej River, major distributaries simulated under river package
16	Net recharge	Variable over the space and time
17	Future prediction period	July 2020 to September 2035 (240 stress periods)

.

Locations (x, y coordinates) of all piezometers used for monitoring were physically verified using the GPS to ensure the authenticity and reliability of the data to be used in groundwater modelling. The depth to water table in all observation points/piezometers in the study area was physically checked/verified to ascertain the accuracy and reliability of data.

2.5.3. Groundwater Levels Data for Model Calibration and Initial Conditions

Considering the availability of data for the maximum number of piezometers, the hydraulic head data of 21 piezometers for the period from October 2015 (post-monsoon) to June 2020 (pre-monsoon) was selected for model. The hydraulic head in Oct-15 was taken as the initial conditions for the model which varied from 115 m to 133 m, as shown in Figure 10.

2.5.4. Digital Elevation Model (DEM)

MODFLOW requires initial head values in the model in all cells before it can start simulating the flow conditions. In this study, IDW [60] was used to interpolate the groundwater levels in ArcGIS software and to prepare the shape files of natural surface level (NSL) and the initial conditions to be uploaded in the model. Field surveys were conducted to obtain the NSL values and additional observation points were obtained from Google Earth Pro to generate the DEM of model area, as shown in Figure 11.

2.5.5. Model Boundaries

In this study, Sutlej River formed the Southern model boundary, Pakpattan-Islam Link (PI-Link) Canal was used as river boundary in the Northeast, in the North–West of the study area Pakpattan canal upper was used as a river boundary in the model and in the South–East of the study area, Sindhnai-Mailsi-Bahawal (SMB) Link Canal was used as a river boundary. Brach canal and main distributaries were also simulated under the River Package of the model. The hydrological components simulated in the model under the River Package are shown in Figure 12.

2.5.6. Groundwater Recharge Components

Two components (recharge from rainfall and recharge from irrigation) were estimated; the latter was further divided into two subcomponents: recharge from canal irrigation (minors, water courses and irrigation fields); and recharge from return flows of tubewell irrigation. Rainfall for the study area was estimated using the available data of four surrounding stations viz Multan, Bahalwalpur, Bahawalnagar and Sahiwal. A map of the average annual rainfall over the study area is shown in Figure 13.

Recharge to the aquifer receives contributions from rainfall and irrigation at the field level which is supplied by canal water and supplemented by groundwater to meet increased cropping intensities. Recharge can be estimated as a function of rainfall and irrigation recharge using Equation (1).

$${\mathrm{R}}_{\mathrm{i}}={\mathrm{f}}_1{\mathrm{P}}_{\mathrm{e}}+{\mathrm{f}}_2{\mathrm{I}}_{\mathrm{c}}+{\mathrm{f}}_3{\mathrm{I}}_{\mathrm{t}}$$

(1)

where R_i is total recharge to the aquifer from irrigation water and rainfall; P_e is effective rainfall; I_c is irrigation supply from canals; I_t is irrigation supply from tubewells; and f₁, f₂ and f₃ are factors which were used to adjust recharge rates during model calibration. A typical range for f₁ is from 0.15 to 0.25, f₂ ranges from 0.25 to 0.35 and f₃ from 0.18 to 0.2 for the Punjab. The model area was divided into different recharge zones (Figure 14) based on the hydraulic conductivity distribution representing low, medium and high recharge zones as obtained from the bore-logs. The recharge rate was further adjusted during calibration based on the hydraulic conductivity.

2.6. Groundwater Extraction

Groundwater in the study area is a major source of irrigation, drinking and industrial uses. Many farmers have installed their own tubewells which are unregulated. Groundwater levels have fallen to 15 to 25 m below the NSL. Tube-wells installed by the farmers in district Vehari during the years 1994–2020 are shown graphically in Figure 15.

Initial estimates of the groundwater extraction were calculated by Equation (2).

P_tw = Q_tw × OF × N

(2)

where P_tw is the pumpage by tubewells in the study area in ft³/s, Q_tw is the discharge of one tubewell in ft³/s, OF is the operational factor of a tubewell (working hours per 24 h), and N is the total number of tubewells. In addition to irrigation use, groundwater extraction for domestic and industrial uses was also estimated and included in total extraction. Extensive pumpage in the study area has caused depletion of groundwater levels creating adequate underground storage potential of floodwater [40].

2.7. Aquifer Parameters

The vertical conductivity (Kv) and other parameters of layer 1 and layer 2 are shown in

Table 3. Horizontal and vertical hydraulic conductivity.

Layer	Horizontal Hydraulic Conductivity (Kh) in m/Day	Vertical Hydraulic Conductivity (Kv) in m/Day	Specific Yield
Layer I	2–73	0.2–7.3	0.05
Layer II	2–73	0.2–7.3	0.06

. The horizontal hydraulic conductivity was calculated from bore-log (the logs prepare from the subsurface lithology) data from district Vehari, which varied from 2 to 73 m/day. Vertical hydraulic conductivity was calculated as 0.2 to 7.3 m/day [61]. It is an unconfined aquifer and the values of hydraulic conductivities for both the layers are the same [62].

3. Results

3.1. Model Calibration

Initially, the model was calibrated for steady-state conditions, which means no change in the temporal domain. The data required for steady-state model development is a boundary of the study area in the form of a shape file, which was created in ArcGIS 10.6. Model parameters were adjusted within their ranges, as provided in Table 3, to obtain the best fit of observed and measured heads. The model was run and calibrated for transient simulation which means non-steady-state or the simulation which changes with respect to time. Time-variant data series required for the calibration period of the model (Oct-2015–2020) was collected and processed. Transient calibration is the process through which the simulated heads are matched with observed heads. In this study, the calibration of the model was performed through trial and error in which the recharge and hydraulic conductivity parameters were adjusted to minimize the head difference between the calculated head and observed head at steady-state conditions. The groundwater simulation is a good fit when the computed head values are in close agreement with the observed head values.

The transient model was simulated from Oct-2015 to June-2020. The input files of the River, Recharge, and Well Package varied over the temporal and spatial domains. Different parameters including hydraulic conductivity, recharge and riverbed conductance were adjusted during model calibration within likely ranges to reduce discrepancies between measured and simulated groundwater levels. The model area was divided into different recharge and hydraulic conductivity zones. The initial heads for the transient model were based on the steady-state model. The transient model was calibrated using head data collected on a seasonal basis for the calibration period. The absolute residual mean (ARM) parameter was used as the main evaluation criteria for the model calibration. Different parameters in the model were adjusted to obtain a good fit between the observed and simulated heads. Piezometer P3 is located in a barren belt so in this area, less pumpage and less recharge is anticipated. Piezometers P4, P19, P20, P21, P2, P26, P6, P8, P15 and P16 lie in the non-perennial area, so in this zone, pumpage is increased. The initial stage of model calibration resulted in an ARM of 0.98 m.

3.2. Model Fine-Tuning of Model

Initially the model was calibrated up to the level of 0.98 ARM. Later, the calibration of the model at individual observation points throughout the model area was examined minutely and the model was fine-tuned by further calibration. The northern boundary, i.e., the Pakpatan-Islam Link canal was improved as the canal is non-perennial, i.e., runs only for six months of the year which improved the results of model calibration further. The conductance of the canal bed was adjusted to obtain the best fit in the model. During calibration, it was found that the aquifer in the vicinity of piezometer No 23 was highly connected with the Pakpatan Main Canal, i.e., the canal and aquifer were hydraulically connected. Aquifer parameters in the model area at different locations were adjusted to obtain the best fit results throughout the model area which improved further the calibration of model. From this, the model calibration improved and a value of 0.70 of ARM was achieved, which was considered satisfactory. The results of different runs for model fine tuning are depicted in Figure 16.

Verification of the calibrated model was performed by comparison of the observed and simulated groundwater levels at different locations in the study area, to see if these two values were in good comparison. Scatter plots and hydrographs for different observation wells were prepared for model verification. Statistical parameters were also calculated to quantify the accuracy and reliability of the calibrated model. The scatter plot shows a good agreement (R² = 0.93) between the observed and computed heads, as shown in Figure 17. Hydrographs were plotted, as shown in Figure 18. Statistical parameters obtained from the model after fine-tuning are given in

Table 4. Statistics of calibrated model for transient conditions (2015–2020).

Name of Parameter	Value
Residual Mean	−0.08
Residual Standard Deviation	0.87
Absolute Residual Mean	0.70
Residual Sum of Squares	154
RMS Error	0.88
Minimum Residual	−2.77
Maximum Residual	2.00
Range of Observations	20.70
Scaled Residual Standard Deviation	0.042
Scaled Absolute Mean	0.034
Scaled RMS	0.042
No. of observations	200

.

Further details of the calibrated model regarding water balance, x-sectional analysis and other aspects have been well explained in [16].

3.3. Future Scenarios Formulation

Five scenarios were simulated to test the effectiveness of the OMC as a MAR site which included the following: (i) Business as Usual scenario (the conditions at the time of June 2020 prevail up to 2035); (ii) Business as Usual (BAU) (pumpage increases by 10% at the end of 2035); (iii) A MAR project with 2.83 m³/s (100 cfs) is in place in the bed of Old Mailsi Canal; (iv) A MAR project of 14.16 m³/s (500 cfs) is in place; and (v) A MAR project with 28.32 m³/s (1000 cfs) is operational. The scenarios were designed as per the future anticipation of MAR project execution at site by PID. The transient calibrated model was run for all five scenarios and the average annual changes in groundwater table were calculated along with the water balance components. The results of three MAR options were compared based on area of influence, changes in storage and rise/fall in the water table.

4. Discussion

4.1. Water Balance from Transient Model

The transient water balance (2015 to 2020) for the composite model (both layers) is given in

Table 5. Summary of water balance (MCM/year) transient conditions—whole model.

Component	Inflow	Outflow	Net
Recharge	111.24	0	111.24
River	166.38	−4.22	162.15
Well	0.00	−387.07	−387.07
GHB	4.74	−3.60	1.15
Drain	0	−1.22	−1.22
Net	282.39	−396.12	−113.73
Average annual depletion of aquifer:			75 mm/year

.

The resultant summary of water balance by the model indicated that the main canals, link canal, branch canals and river are the major source of inflow to the model area followed by recharge (from rainfall, irrigation return flows and tubewell pumpage return flows). Significant outflow from the model area is from groundwater extractions, i.e., the pumpage of groundwater by tubewell, which includes extraction of groundwater by agricultural, domestic and industrial tubewells in the study area. Model results have yielded the information that the groundwater reservoir is depleting at an average overall annual rate of 75 mm over the period of 2015 to 2020. Net loss of aquifer storage indicates the depletion of the water table which is increasing the cost of pumping.

4.2. Layer-Wise Summary of Water Balance for Transient Model

Layer-wise, the water balance was calculated by model to analyze the interlinkages and behavior of recharge between the layers. Water balance outputs for layer 1 and layer 2 are shown in

Table 6. Summary of water balance for Layer 1 (MCM/year).

Component	Inflow	Outflow	Net
Recharge	111.24	0	111.24
River	166.38	−4.22	162.15
Bot	11.85	−395.61	−383.75
GHB	2.58	−1.82	0.76
Drain	0	−1.22	−1.22
Net	292.04	−402.87	−110.83
Average annual depletion of aquifer:			−73 mm/year

and

Table 7. Summary of water balance for Layer 2 (MCM/year).

Component	Inflow	Outflow	Net
Top	395.61	−11.85	383.75
Well	0.03	−387.07	−387.04
GHB	2.17	−1.78	0.39
Net	397.80	−400.71	−2.90
Average annual depletion of aquifer			−2 mm/year

, respectively.

The water balance analysis of the first layer indicated that the water table is being depleted at an average annual rate of 73 mm. The major outflow from Layer 1 is from the bottom of the layer, which flows downward to Layer 2, while the major inflow component is from the rivers and canals and from recharge. The major outflow component from Layer 2 is the extraction of groundwater by tubewells, the source for which is the inflow from the top layer. The pumping from Layer 2 is 387 MCM which results in a net inflow of 383.8 MCM from Layer 1. Average depletion of the groundwater table in the study area for calibration period is 75 mm per year, out of which a major portion, i.e., 73 mm, is occurring in Layer 1 while a small depletion of 2 mm per annum takes place in Layer 2. The layer-wise analysis of model outputs indicates that recharge is into Layer 1, groundwater flows from Layer 1 to Layer 2, and groundwater is being pumped from Layer 2. This indicates that if we have this level of recharge in Layer 1, water levels in Layer 2 will not decline very much but there will be a gradual decline in Layer 1. So, MAR will be an important contribution to reducing the decline in groundwater in Layer 1.

4.3. Future Scenarios/Anticipated Use of Groundwater

The study area is an intensively cropped region where progressive farming is also taking place. Many farmers are turning from electricity/diesel to solar power as a source of energy for running the tubewells to extract groundwater. Growing population, changing climatic conditions and changing land-use patterns are the major drivers putting continuous pressure on the groundwater reserves. An increasing population demands more food and fiber for which extra water is required. Under the scenario of diminishing or constant surface water supplies, the extraction of groundwater is anticipated to rise. Changes in land use are also a cause of reduction in groundwater recharge as paved/constructed areas are increasing into traditional agricultural lands. Similarly, droughts owing to climatic changes also warrant more extraction of groundwater. Data have indicated that groundwater use in the study area is increasing continuously, and the same trend is expected for the coming decades. Therefore, three major possibilities regarding future groundwater use in the study area are as follows:- (i) extraction of groundwater will increase; (ii) it will remain the same at the current rate; and/or (iii) it will reduce with the passage of time. This may happen if farmers are encouraged to change cropping practices towards low water use crops.

Option one is clearer, and its chances are more as compared to the other two scenarios. However, if some water management interventions are introduced which may increase surface water availability, under this situation we may anticipate option ii and iii for future uses of groundwater in the study area.

The calibration period for the model is from 2015 to 2020; the future prediction has been planned for fifteen years, i.e., up to 2035. Population and cropping intensity are increasing to meet food requirements while the surface water supply is almost constant. Extraction of groundwater is anticipated to rise continuously as dependence on groundwater is likely to increase in future. Keeping in view the climatic conditions, anticipated surface water availability, expected use of groundwater and the MAR project of PID at the OMC, different future scenarios were designed, as outlined in

Table 8. Future scenarios for the groundwater use in OMC.

Scenario	Scenarios/Options	Remarks/Description
S1	Baseline scenario (BASE)	Pumpage, River Package, and Recharge Package in October 2019 to September 2020 are repeated till 2035. Other stresses are the same as 2019–2020, i.e., the last year of calibration.
S2	Business as usual (BAU)	Pumpage is increased by 10% from 2020 to 2035. River and Recharge components of last 5 year repeated.
S3	MAR pilot project is in place with 100 cfs. (MAR1)	2.83 m3/s (100 cfs) flow is released into the bed of Old Mailsi Canal (first 45 RDs of the canal) for three months (July–August) every year till 2035
S4	MAR project is in place with 500 cfs (MAR2)	14.12 m3/s (500 cfs) flow is released into the bed of Old Mailsi Canal (first 45 RDs of the canal) for three months (July–August) every year till 2035
S5	MAR project is in place with 1000 cfs (MAR3)	28.32 m3/s (1000 cfs) flow is released into the bed of Old Mailsi Canal (first 45 RDs of the canal) for three months (July–August) every year till 2035

.

4.4. Forecast Water Balance by Model Under Different Scenarios

Under scenario S1, the groundwater pumping, recharge from river and canals, and recharge from minor canals, water courses, and irrigation fields, return flow from groundwater and rainfall during October 2019 to September 2020 were repeated till the end of the simulation period 2035 to understand the impacts of stresses during the last year of calibration on aquifer response. Water balances for S1 and S2 are given in

Table 9. Water balance under different scenarios.

Components	S1	S2	S3	S4	S5
Recharge	143	113.26	143	143	143
River	164	220.54	195	202	213
Well	−386	−674.84	−419	−419	−419
GHB	16	78.93	16	15	13
Drain	0	0.00	0	−1	−1
Net	−63	−262.11		−60	−51
Avg. Annual Fall	−41	−172		−39	−34

. Under Scenario 2 (S2), it was assumed that all other hydrological components will remain the same but pumpage will increase by 10% from 2020 to 2035. Recharge pumpage is increased by 10% from 2020 to 2035.

4.5. Simulation of MAR Project at OMC

Three scenarios, S3, S4 and S5, were designed for the MAR project of PID at OMC which is under implementation. For the future forecast of the impact of the MAR project on groundwater, three scenarios were designed viz., S3, S4 and S5 with 2.83 m³/s, 14.16 m³/s and 28.32 m³/s (100 cfs, 500 cfs and 1000 cfs) of flows to be diverted into the OMC, respectively, during July, August and September every year. The experimental reach of 14 km of OMC was divided into three segments/ponds. A detailed experimental layout showing major features at each pond with a monitoring setup of batteries of piezometers perpendicular to each pond is shown in Figure 19.

The purpose of these different ponds is experimenting with different options and to test which intervention/setup performs well with respect to the recharge of the aquifer. Descriptive parameters of the three ponds are given in

Table 10. Description of three ponds for the MAR project at OMC.

Pond No	Length (m)	Width (m)	Depth of Pond (m)	Bed Level (m-amsl)	Remarks
1	5793	47	3.14	134.84	The pond has been set at the original canal bed level as per design parameters of the OMC
2	3354	46	3.75	133.5	The upper crust layer of comparatively harder strata of around 0.5 to 1 m has been removed and consequently the bed of canal became lower. This was treated as an intervention for pond 2.
3	4573	47	3.14	133.91	The bed of the canal has been set at design level. A total of 144 recharge wells have been constructed to accelerate the infiltration rate for enhanced recharge. The diameter of each recharge well is 183 cm (6 ft) and depth is 305 cm (10 ft). This is an intervention for pond 3.

.

The historical data and clarification sought from PID indicated flood water will be available for MAR during the flood/rainy season for a period of about three months (July–September) every year. While running the future simulations for S3, S4 and S5 (with MAR options), three months (July–September) were simulated as wet stress periods from July 2020 onward, while other months of the year were treated as dry stress periods in the model. OMC was simulated with three different reaches consisting of Pond1, Pond2 and Pond3. The design of the recharge wells constructed on site in Pond3 of OMC is shown in Figure 20.

4.6. Water Balance for MAR Scenarios

The calibrated model was run for scenarios S3, S4 and S5 (three MAR options) to determine the water balance at the end of simulation period (2035).

Hydrographs at target P8 (close to OMC) for scenario S4 and S5 are shown in Figure 21, which reflect that from 2015 to 2020, the water table drops continuously. When in 2020, MAR is placed in action, the water table rises about 2 m (for S4) and about 3.5 m (for S5), and then drops continuously with cyclic effects. The cyclic rise and fall is due to wet and dry stress periods typical of monsoon floods. MAR has been placed in action during three stress period which are July, August and September (flood season) every year; the other months of the year are dry, i.e., there is no MAR.

These results indicate that MAR is contributing towards the improvement in aquifer storage and is facilitating the retardation of aquifer depletion rates, but it could not balance the water budget. At the end of the simulation period (2035), the overall water balance is still negative (−34 MCM/year) and the water table is falling at the rate of 34 mm/year. It can be noticed that MAR continuously reduced the water table rate of decline from 43, 39 to 34 mm per year under different MAR scenarios (S3, S4, S5). This means an increasing flow rate for MAR gradually from 2.83 to 14.16 to 28.32 m³/s has resulted in a reduction in the water-table falling rate. These increasing doses of MAR have also reduced the storage depletion gradually and have reduced the depletion rate of the water table.

The calibrated model is a representation of the actual field conditions at this case study site, as shown in

Table 11. Comparison of outputs of different future scenarios.

Sr No	Model	Duration/ Description	Net Change in Reservoir (MCM/Year)	Annual Rise (+) or Fall (-) of Water Table (mm/year)
1	Steady State	October 2015	0.0049	0.003
2	Transient calibration	October 15–June 20	−114	−75
3	S1-Scenario 1 (Business as usual)	October 15 to September 20 October 20–35, last 5-year cycle repeated)	−60	−39
4	S2-Scenario 2 (2020–35) pumping increased gradually to 10% at the end of 2035)	October 15 to June 20 historical data and the July 20 to September 35 last year data repeated with increase in pumpage)	−262	−172
5	S3-Scenario 3 (MAR1) (As in S2 plus 2.83 m3/s (100 cfs) diversion from Islam Barrage during July–September form 2020–35)	October 15 to September 20	−66	−43
6	S4-Scenario 4 (MAR2) (As in S2 plus 14.12 m3/s (500 cfs) diversion from Islam Barrage during July–September)	October 2015 to September 2035	−60	−39
7	S5-Scenario 5 (28.32 m3/s (1000 cfs) MAR) (MAR3)	October 2015 to September 2035	−51	−34

(Sr No 1, 2). It gives confidence in predicting the future response of the aquifer under various hydrological conditions without (Sr No 3,4) and with MAR (Sr Nos 5,6,7).

The model has calculated that without any interventions, the depletion of aquifer storage is 114 MCM annually, with an average annual decline of water levels by 39 mm. This indicates that the aquifer is under stress and water levels are falling continuously as net storage is showing negative trends. The potential response in the aquifer under five scenarios is shown in Figure 22.

This shows that without MAR S2 (500 cfs) is the worst scenario, when there is a 10% increase in pumpage at the end of 2035, canal flows are the same, and no MAR scheme is in place. It has resulted in the maximum depletion of the groundwater reservoir, i.e., 262 MCM per year storage loss with 172 mm/year depletion rate.

From Table 9, it is evident that the aquifer behaves almost the same under scenario S1 and S4, which means that if the aquifer is recharged with a flow rate of 14.12 m³/s (500 cfs) for three months (July-August) during the year, it can almost balance the increasing pumping pressure, but the trend remains negative. For the scenario S4—a flow rate of 14.12 m³/s (500 cfs) is diverted for MAR in the bed of OMC and recharge wells are in place in Pond3—the overall annual change in storage reduces to −60 MCM, but still a negative trend prevails in the model area, as the water table falls at the rate of −39 mm/year; however, the rate of decline has been reduced. For S5 (1000 cfs), the annual reduction in storage reduces further to −51 MCM and the depletion rate is −34 mm/year.

It is noted that under all scenarios, the overall trend of storage is negative, and the water table falls. However, it can be found that MAR, when placed in action, reduces the rate of depletion. This concludes that MAR can be an option for the sustainable management of groundwater; it can reverse the pressure of aquifer exploitation. For the sustainable use of groundwater in the study area, either the groundwater extraction needs to be moderated or the managed recharge of aquifer is required to be enhanced. The calibrated model now can be run/used for different scenarios of MAR to predict the aquifer response under different stresses.

Volume of flow into the bed of OMC has been increased gradually in the order of 100 cfs, 500 cfs and 1000 cfs to compare the impact of MAR on the groundwater reservoir. It has been found that increasing the MAR volume has a positive impact on the aquifer depletion. Figure 23 indicates that increasing the flow rate for MAR has positive impacts as S4 and S5 have yielded a 9% and 23% improvement in aquifer replenishment. Similar trends have been observed in the falling water table. Under S3, the water table depletion rate is −43 mm/year which reduces to 39 (9%) and 34 (21%) mm/year under S4 and S5, respectively; indicating a 9% and 21% improvement in the depletion rates of the aquifer. Increasing the release of flows into OMC improves the aquifer storage and reduces the depletion rate. However, it is still noteworthy that the increasing flow rates could not eliminate the overall depletion trends.

The area of influence under different MAR scenarios is also of importance for decision making and investment in MAR schemes. The increasing trend of the area of influence of MAR under three options (S3, S4 and S5) predicted by the model is shown in Figure 24. It was found that the area where the water table rise is more than 4 m has increased by 8%, 20% and 20% under scenario S3, S4 and S5, respectively, as compared with Scenario S2. This indicates that S4 and S5 have an almost similar effect on the area of influence. The predicted areas of influence under different scenarios are shown in Figure 25.

This only shows the area’s impact. More research is needed to understand what this means for agricultural output, and benefit cost analysis. Using the model leads to the conclusion that MAR can improve the aquifer storage and contributes towards the sustainable use of groundwater, but it cannot do it all alone. Other supply- or demand-side management and improved conjunctive management will also be important, for example, improved land and water management practices by farmers. This is the same in other regions [19,63].

5. Conclusions

A groundwater model for the area has been developed, calibrated and applied for the future prediction of the groundwater situation in 2035 in the study area. The model now can be used for predicting the impact of any future stresses on the aquifer in the study area. The model can be used/replicated in areas in Punjab and to evaluate the impact of MAR schemes. The model developed is specific for this area and can be used by PID and others in future work for other locations as well. The groundwater modelling tool has been used for the first time for the evaluation of a MAR project in Pakistan, which can forecast the impacts of any volume of water to be diverted for MAR.

MAR improved the underground water availability but could not completely mitigate the adverse impacts on the overall stresses on groundwater. It reduces the depletion rates and increases the aquifer storage. The cyclic behavior predicted by the model indicates that the aquifer is responsive to recharge from the OMC and MAR which can be implemented at the case study site. Nominal improvements in water table rise between S4 and S5 can be attributed to the fact that water mounding under the bed of OMC goes up to about 4 m and then water flows toward the surrounding areas of aquifer. Secondly, the hydraulic conductivity (Kv) of the canal bed is a controlling and limiting factor for MAR for aquifer storage.

The above assessment is positive, but to support decisions, the monitoring of the aquifer from diversion of flood events over a period of 3 to 5 years will be required for policy makers to support further investment in larger MAR projects. The floods in the Sutlej River can offer an ideal opportunity to divert excessive flood waters into the OMC. The comparison of floods in normal years and large floods taking place in August 2023 offer a unique opportunity to capture and recharge flood waters and to monitor impacts. Moreover, diversion of some of the flood water into OMC will also provide some flood mitigation in the Sutlej.

Groundwater modelling can provide more detailed information, and under different possible potential scenarios. The MODFLOW groundwater model developed is the first time it has been applied to examine the validity of MAR project in this region.