**Managed Aquifer Recharge as a Strategic Storage and Urban Water Management Tool in Darwin, Northern Territory, Australia**

**Anthony Knapton 1, Declan Page 2,\*, Joanne Vanderzalm 2, Dennis Gonzalez 2, Karen Barry 2, Andrew Taylor 2, Nerida Horner 3, Chris Chilcott <sup>3</sup> and Cuan Petheram <sup>4</sup>**


Received: 24 June 2019; Accepted: 5 September 2019; Published: 9 September 2019

**Abstract:** Population growth and increased irrigation demand have caused a decline in groundwater levels that limit water supply in the Darwin rural area. Managed Aquifer Recharge (MAR) is a practical solution that can be adopted to augment stressed groundwater systems and subsequently increase the security of water supply. Aquifer storage capacity is considered to be the primary constraint to MAR where unconfined dolostone aquifers rapidly recharge during the tropical, wet season and drain again in the dry season. As a result, there is a general understanding that aquifers of this nature recharge to full capacity each wet season. However, the aquifer storage capacity and the potential for niche opportunities for MAR to alleviate declining groundwater levels has not previously been examined. This paper uses the Darwin rural area's Proterozoic Koolpinyah Dolostone aquifer and the existing Koolpinyah Groundwater System to evaluate the prospects of MAR using both infiltration and injection techniques. Direct injection wells in an aquifer storage transfer and recovery (ASTR) scheme were favoured in this area, as injection wells occupy a smaller surface footprint than infiltration basins. This assessment suggested MAR during the early to mid-dry season could alleviate the impact of the dry season decline in groundwater levels in the Darwin rural area. The use of a larger aquifer storage and recovery (ASR) system (5,000,000 m3/year) was also assessed as a potentially viable technical solution in the northern part of the aquifer where it is understood to be confined. The ASR scheme could potentially be scaleable to augment the urban water system and provide strategic long-term storage. Consideration must also be given not only to the strategic positioning of the ASR water bank, but also to the hydrogeology of the aquifers in which the systems would be developed. Not all locations or aquifer systems can successfully support a strategic storage ASR system. Scheme-scale feasibility assessment of an ASR water bank is required. The study reported here is an early phase of a series of investigations that would typically be required to demonstrate the viability of any proposal to apply MAR to increase the reliability of conjunctive groundwater and surface water supplies in stressed water resources systems. It focusses on assessing suitable storage areas in a lateritic aquifer.

**Keywords:** Managed Aquifer Recharge (MAR); aquifer storage and recovery (ASR); strategic storage; Northern Australia

#### **1. Introduction**

Urban potable water supply systems require a high degree of reliability and security. This can be challenging where rainfall is seasonal such as in the vicinity of the city of Darwin, Northern Territory, Australia. Darwin experiences distinct wet and dry seasons, with 95% of rain falling in the wet-season months (November to April). The annual average rainfall is 1423 mm.

Darwin's reticulated water network has traditionally relied upon surface water reservoirs with a minor component (~15%) from groundwater [1]. Borefields used for urban water supply are in the peri-urban Darwin rural area and target the Koolpinyah Dolostone aquifer [2]. Reticulated water demand by urban and industrial users has produced an immediate system yield shortfall of approximately 5,000,000 m3/year [1]. Locally, the water from this dolostone aquifer is also relied upon heavily for drinking and irrigation as the Darwin rural area is not connected to reticulated water supply. A consequence of this is that the residents of the Darwin rural area are particularly vulnerable to consecutive years of poor rainfall.

In 2016, the groundwater levels for the most part were low if not the lowest recorded for the past ten years for most of the Darwin rural area in the vicinity of the municipal borefield area (Figure 1). End of dry-season water levels can result in risk to the water supply for many groundwater users in this area and to nearby groundwater dependent ecosystems, such as Howard Springs. Figure 1b shows that the dry season groundwater levels approach the level at which Howard Springs reportedly ceases to flow [1,3]. MAR has been put forward as an option to help augment the stressed groundwater resource in the Darwin rural area. Previously, the potential for MAR in this area has been dismissed [4] without any technical assessment of viability using available data.

**Figure 1.** Hydrographs of bores in the Darwin rural area: (**a**) RN009266 at Middle Point and (**b**) RN009421 at Howard Springs.

This study investigates the potential for managed aquifer recharge (MAR) to: (i) reduce the risk of water stress for residents in the Darwin Rural Area; and (ii) provide addition storage capacity for reticulated supplies for urban and industrial use in Darwin. It examines how the options can interact to provide urban water supply security for the Darwin rural area and the potential for the development of a strategic storage for the City of Darwin. Specifically, it considers the potential for MAR into the most water stressed parts of the Koolpinyah Dolostone aquifer (e.g., MAR1-5 in Figure 2).

**Figure 2.** Location of modelling scenarios: Infiltration basin and well injection scenarios where aquifer is unconfined at MAR1-5; ASR Water bank scenarios where aquifer is confined in the northern section. White and black circles indicate a prospective location for a strategic ASR water bank (1.5 and 5 Mm3/year). MAR scenarios are described in detail in Table 1.

Both the augmentation of the Darwin rural peri-urban system and a strategic ASR scheme are favoured in an aquifer that has generally high hydraulic conductivity, low specific yields, a suitably large volume of unsaturated sediments, natural boundaries that limit vertical and horizontal losses of the stored water, and low salinity of the native groundwater. For the ASR scheme to be successful, the hydraulic conductivity of the storage aquifer must be high to allow high rates of infiltration or injection over a relatively short wet season period as well as enabling high rates of extraction of the stored water to meet urban requirements.

The general understanding is that unconfined aquifers in the Darwin catchments rapidly recharge to full capacity during the wet season and drain again in the dry season [4], a common occurrence in unconfined, shallow aquifers. Aquifer storage capacity in the wet season, when there is access to a source of water for recharge, is considered to be the primary constraint to replenishing these unconfined aquifers via MAR [5]. However, the aquifer storage capacity of the Koolpinyah Dolostone and the potential for MAR have not previously been examined. This study addresses this knowledge gap by considering niche opportunities for MAR, by assessing the additional volume of water that could be recharged to alleviate the impact of current pumping for urban water supply, rural residential use and horticultural water supply. This study can be considered as a pre-feasibility assessment using available information and has the following objectives:



**Table 1.** Howard East modelling scenarios.

Available surface water is assumed to be the source of water for recharge. The current Darwin Regional Water Supply Strategy does not address the vulnerability of water supply in the Darwin rural area [1]. Instead, it focuses on diversifying supply options to increase security and sustainability of supply to the Greater Darwin Region. Therefore, this paper is the first investigation of the potential to apply MAR to increase the reliability of conjunctive groundwater and surface water supplies in the rural area, a seasonally stressed water resource. However, the detailed technical, social and economic feasibility of specific MAR configurations is not addressed. The Australian MAR Guidelines [6] provide a comprehensive risk-based framework for assessment of scheme-scale technical feasibility, addressing human health and environmental risks along with operational issues such as clogging.

#### **2. Methodology**

#### *2.1. Koolpinyah Groundwater System Model*

The Northern Territory Government's Koolpinyah Groundwater System model was used to investigate the potential for MAR in the Koolpinyah Dolostone (i) to augment seasonally declining groundwater levels to prevent private water supply bores failing during the dry season and (ii) to provide a larger, potentially long-term strategic storage for urban water supply. The Koolpinyah Dolostone is largely unconfined in the stressed areas where augmentation is assessed, while to the north it is confined by mudstone.

The current Koolpinyah Groundwater System model was developed using the FEFLOW code. The model is fully described by CloudGMS [2]. The model domain covers approximately 1600 km2, including the Howard River catchment and the western part of the Adelaide River catchment and includes the extent of the Koolpinyah Dolomite based on the interpretation of airborne magnetics and electromagnetic data [2]. It comprises three layers: layer 1 (the upper layer) represents the laterite aquifer, layer 2 represents the Cretaceous sediments within the weathered dolomite and layer 3 represents the fractured zone of the Koolpinyah Dolomite. The Koolpinyah Groundwater System model was calibrated on groundwater levels and dry season discharge measurements, over the period from 1980 to 2014, using the automated inversion software PEST [2]. Extraction volumes are uncertain, as rural residential water use is not metered.

The model is reported to be Category 2, with the capacity to achieve Category 3, according the Australian Groundwater Modelling Guidelines [7], due to a reasonably long observation data set over the areas with greatest stress and since projected pumping scenarios have stresses similar in magnitude.

There are five types of boundary conditions in the model:


#### *2.2. Modelling Scenarios for Peri-Urban Groundwater Augmentation and a Strategic Storage*

The modelling scenarios to consider the potential for MAR in this rural area were simulated over the period 1996 to 2014 and focused on five stressed locations selected for recharge augmentation: MAR1 to MAR5 (Figure 2). Infiltration (SC1a) and injection (SC1b, SC2a) techniques were both assessed. At MAR1–MAR5, the Koolpinyah Dolostone is unconfined, and hydraulic heads were used to trigger and halt the enhanced recharge under these scenarios.

A key constraint for inter-seasonal recharge is securing a source of water during the dry season when the storage capacity becomes available in the aquifer. Therefore, the potential of a scheme to store water in the wet season, when it is likely to be available, was investigated. For this purpose, a strategic water bank formed with an ASR wellfield (SC1c, SC2b) was simulated to the north of the drawdown area (Figure 2) where the aquifer is understood to be confined (Figure 2). Each MAR scenario was compared to the base case (without MAR). Modelling scenarios are summarised in Table 1 and described in the following text.

In the stressed area, the first stage of modelling assessed the feasibility and type of MAR during the dry season (SC1a and SC1b, mid dry season). Hydraulic heads at a point in the centre of the MAR sites were used to trigger recharge augmentation, via using either infiltration basins or injection wells (Aquifer Storage Transfer Recovery (ASTR)). Trigger values were set equal to the observed end of dry season (30 November 2009) hydraulic head without MAR (MAR1, 2, 4 and 5~18 mAHD; MAR3~10 mAHD). A node spacing was allowed between recharge (MAR) and recovery locations (existing bores) to reduce interference.

The second stage of modelling considered the potential for MAR via injection (ASTR) with injection commencing at the start of the dry season for approximately three months until the start of July at MAR1 and MAR3 (SC2a, early dry season). A constant head was maintained from 1 April (or when heads declined below the trigger) to 1 July (MAR1 constant head of 36 mAHD applied, MAR3 constant head of 20 mAHD applied). Importantly, this scenario removed pumping from town water supply bores to assess the potential benefits to the stressed areas of using the strategic ASR water bank for town water supply.

In the assessment of the ASR water bank, the first stage of modelling considered 1,500,000 m3/year of injection and recovery (SC1c) (5 ASR bores, 27.4 m3/day per bore for 120 days). The ASR water bank scenario was developed further in the second stage of modelling (SC2b), with a 5,000,000 m3/year water bank (16 SR bores, 27.4 m3/day per bore for 120 days) replacing extraction for town water supply in the stressed area.

#### **3. Results and Discussion**

This first stage of modelling was used to determine if MAR recharge via infiltration or injection was more feasible for this area. The opportunity to create a strategic ASR water bank further to the north in the confined part of the aquifer was also investigated.

Both infiltration and injection modelling scenarios in the stressed areas were limited by the established model construction and scale. In areas where the cone of depression was not well represented, the node spacing between recharge and recovery locations resulted in limited instances where recharge was triggered in areas MAR2, 4 and 5. Recharge was triggered at MAR1 and MAR3 (Figure 3) and both techniques achieved a comparable gain in groundwater level when recharge was triggered (Figure 4).

The end of dry-season hydraulic head within a currently stressed area can clearly be increased by MAR (Figure 5, e.g., for injection). At MAR1, the impact on end of dry-season water table was up to 8 m when recharge was applied midway through the dry season and ~2 m when applied at the start of the wet season (Figure 4). The results for MAR3 indicate that infiltration basins and injection wells applying recharge midway through the dry season could cause the water table to rise by ~10 to 15 m at the end of the dry season (Figure 4). Injection at the start of the dry season resulted in an approximately 6 m increase over the base case in end of dry-season water table (Figure 4). However, further assessment is required of the potential for any rise in hydraulic head at the end of the dry season to protect against bores running dry remains.

Overall, the injection (ASTR) method was considered more prospective than infiltration due to fewer land use constraints. In the infiltration scenarios, an infiltration rate of 0.015 m/d was adopted to prevent excessing mounding, which resulted in heads approaching the surface and the invoked seepage face boundary conditions. The properties assigned to layer 1, representing the surficial laterite layer, could allow higher infiltrations fluxes; however, the lower permeability of layer 2 impedes vertical groundwater movement and results in excessive mounding. With a low infiltration rate (<0.02 m/d) a large area of land (>100 ha) would be required for infiltration basins. Land availability for infiltration basins is influenced by land use, which may be achievable in areas with larger block sizes, but areas to the west may not be as prospective due to basin area relative to block sizes.

**Figure 3.** Comparison of monthly modelled recharge volume under infiltration (SC1a) or injection (SC1b, SC2a) type MAR scenarios: SC1a and SC1 b recharge occurs during mid-dry season, recharge triggered by dry season minimum hydraulic head of MAR1~18 mAHD, MAR3~10 mAHD. SC2a recharge occurs early dry season, constant head maintained from 1 April to 1 July, MAR1 constant head of 36 mAHD, MAR3 constant head of 20 m AHD.

In the mid-dry season MAR scenarios, most of the augmented recharge was typically triggered between August and December each year. This interval coincides with the period of highest rural residential groundwater use (June to December) when surface water is in high demand for reticulated supply. Despite this competition for water, the volume of recharge required for the inter-seasonal MAR is small (i.e., 1–5,000,000 m3/year), and therefore is unlikely to have a significant impact on the volume of water in surface storage.

The potential for recharge at the start of the dry season was assessed in the second stage of modelling, through application of a constant head from the end of the wet season to the start of July (Table 1). Recharge was typically triggered for 3 months and the median recharge over the simulation period was 1,200,000 m3/year. The magnitude of the increase in hydraulic head at the end of the dry season was less than for the mid-dry season recharge scenario due to the time interval that recharge was triggered in, but presumably having an impact over a larger area. End of dry season hydraulic head declines suggest that sufficient storage remains available for wet-season recharge.

**Figure 4.** Example cross-section of hydraulic head response under infiltration (SC1a) or injection (SC1b, SC2a) type MAR scenarios at (**a**) MAR1 and (**b**) MAR3 for end of dry season groundwater level (30 November 2009). SC1a and SC1b Recharge occurs during the mid-dry season, recharge triggered by dry season minimum groundwater level of MAR1~18 mAHD, MAR3~10 mAHD. SC2a Recharge occurs in the early dry season, with constant groundwater level maintained from 1 April to 1 July, MAR1 constant head of 36 mAHD, MAR3 constant head of 20 mAHD. Location of the cross-section is shown on Figure 5.

**Figure 5.** Hydraulic head response to an ASTR (injection well) MAR scheme (SC1b) triggered by declining hydraulic heads. Recharge is typically triggered between August and December. (**a**) The base case without MAR and (**b**) with the effect with MAR.

Maintaining hydraulic heads at the start of the dry season may be more practical and economical than waiting until later in the dry season when surface water resources are likely to be stressed and may not be available for MAR. Proximity to groundwater discharge locations would need to be assessed if this option was to be considered in more detail, to ensure the additional recharge is maintained within the aquifer. The scenarios considered indicate a reduction in the dry-season hydraulic head declines; however, the effectiveness of MAR in providing protection to individual groundwater users would need to take bore construction details into account. Infrastructure from the recharge source to recharge locations would also be required.

The ASR water bank scenario recharged a confined portion of the aquifer in the wet season and the model results suggest that the injected water remained localised (<2 km from the ASR bore) to the ASR borefield. With 1,500,000 m3/year of injection and recovery, hydraulic head increased by ~10 m at the end of the wet season and declined by ~5 m at the end of the dry season, when compared to the base case without MAR. Increasing this to 5,000,000 m3/year of injection and recovery, hydraulic head increased by ~20 m at the end of the wet season and declined by ~10 m at the end of the dry season, in relation to the base case.

The potential for development of a 5,000,000 m3/year water bank to replace extraction for municipal supply in the stressed area of the Koolpinyah Dolostone was considered. The cessation of pumping from the municipal supply bores alone provided only localised benefits that did not alleviate groundwater declines in all of the five MAR locations assessed, due to multiple uses (irrigation, rural residential). The MAR water banking approach could serve as a longer-term option for urban water cycle resilience against reduced recharge during poor wet seasons. While the end user has not been defined, it is possible that the ASR water bank could contribute to Darwin urban water supply. Infrastructure from the recharge source to the ASR site and from the ASR site to the end user would be required. This approach is of broad international interest as it demonstrates that there is potential niche application of MAR in lateritic aquifers that have been previously dismissed as unsuitable for MAR. The study demonstrates that there may be opportunities to develop strategic urban water supplies in aquifers, though further investigations would be required to develop a specific scheme.

MAR scheme numerical modelling is typically undertaken at local (scheme) scale of a few kilometres, rather than the regional scale (10s to 100s of kilometres) of the Koolpinyah Groundwater System model. Due to the scale of the existing model, its discretization and the proximity of the stressed areas to the no-flow boundary, it is not possible to evaluate the hydraulic impact of individual MAR schemes. Nonetheless, this regional scale model was sufficient to undertake an entry-level assessment of MAR feasibility.

This pre-feasibility assessment suggests MAR may be beneficial in the Darwin rural area; however, scheme-scale feasibility assessment of an ASR water bank is required. The Australian MAR Guidelines [6] provide a comprehensive risk-based framework for assessment of scheme-scale technical feasibility, addressing human health and environmental risks along with operational issues such as clogging. Regardless of MAR technique, it is necessary to assess the risk of clogging due to recharge water quality and how this can be managed to ensure a sustainable operation.

This study is innovative, as the general area proposed for an ASR water bank has very little existing groundwater use or hydrogeological data. This is a common theme to many Mar investigations and at times may lead to pre-mature dismissal of MAR opportunities in aquifer such as these. Nevertheless, it will be essential to confirm the degree of confinement and the hydraulic response to MAR, the water quality within the storage zone and its potential impact on the utility of recovered water for urban supply and measures to minimize the impact of well clogging. In addition to MAR technical feasibility, it is essential to consider solutions which are supported by the community and are economically viable. Considerable investment in infrastructure would be required for MAR to alleviate stress across multiple areas in the Darwin rural area. However, a combined approach, with actions such as demand management and well maintenance, may prove to be lower cost than alternative water supply strategies.

#### **4. Conclusions**

Through this MAR modelling assessment, it was concluded that MAR (by a variety of means) is technically feasible to augment recharge and reduce the magnitude of groundwater decline in the Darwin rural area. Recharge via direct injection may be favoured due to having a smaller surface footprint, particularly in the more urbanised areas. The availability of a source of water for recharge during the dry season was considered a potential constraint for MAR. However, the volume of recharge required for the inter-seasonal MAR is small (i.e., 1-5,000,000 m3/year), and therefore is unlikely to have

a significant impact on the volume of water in surface storage. Of particular interest internationally is that this study identified an opportunity for a strategic water bank of 5,000,000 m3/year using ASR wells to the north of the stressed area is another option to augment groundwater resources, storing wet-season excess surface water in the confined part of the aquifer for use when needed.

This pre-feasibility assessment suggests MAR may be beneficial in lateritic aquifers such as the Darwin rural area and that this assessment could be of broader interest in evaluating lateritic aquifers that have typically been deemed unsuitable for MAR.

**Author Contributions:** A.K. developed the groundwater model, applied the groundwater model to this investigation and contributed to paper writing. D.P. and J.V. conceived the MAR investigation and contributed to paper writing. D.G., K.B., N.H. contributed to data analysis and paper writing. A.T., C.C., C.P. provided oversight of the project and contributed to paper writing.

**Funding:** This research was funded by Australian Department of Agriculture grant number [22668] and is part of the CSIRO Northern Australia Water Resource Assessment, part of the National Water Infrastructure Development Fund: Water Resource Assessments.

**Acknowledgments:** This research was undertaken within the Northern Australia Water Resource Assessment as part of the Australian Government's Agricultural Competitiveness White Paper, the government's plan for stronger farmers and a stronger economy and was supported by the Department of Agriculture and Water Resources. The authors acknowledge the contributions of Des Yin Foo, Dale Cobban and Mardi Miles (Department of Environment and Natural Resources, NT); and David George, Shane Papworth, Trevor Durling (Power and Water Corporation). The authors gratefully acknowledge the thoughtful review comments of three anonymous reviewers and Peter Dillon, academic editor, in improving the quality of this paper.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Mapping Economic Feasibility of Managed Aquifer Recharge**

**Jean-Christophe Maréchal 1,2,\*, Madjid Bouzit 1, Jean-Daniel Rinaudo 1,2, Fanny Moiroux 1,2, Jean-François Desprats 1,2 and Yvan Caballero 1,2**


Received: 17 November 2019; Accepted: 18 February 2020; Published: 2 March 2020

**Abstract:** Managed aquifer recharge (MAR) constitutes a potential and promising solution to deal with several water management issues: water shortage, water level depletion, groundwater pollution, and saline water intrusion. Among others, the proper siting and cost evaluation of such a solution constitutes sources of uncertainty for the implementation of MAR schemes. In this study, we proposed a methodology for the assessment of the levelised cost of recharged water through an infiltration basin, including investment and operating costs. The method was implemented in a GIS-tool in order to build maps of levelised costs at the aquifer scale. The sensitivity analysis allows for the identification of the main natural characteristics (water quality and availability, etc.), technical (system life duration, recharge volume objective, etc.), and economic parameters (energy price, discount rate, etc.) that dominate the final cost estimate. The method was applied to a specific case study on an alluvial aquifer in Southern France. This new information on the economic feasibility of MAR scheme should be incorporated with more classical GIS-MCDA (relying on soil characteristics, aquifer storage capacity, land use, etc.) in order to properly site the system. Further information on financial and economic feedback from MAR implementation and research on the fate of recharged water are needed for a better benefits evaluation of this solution.

**Keywords:** infiltration basin; cost function; suitability map; groundwater; MAR; GIS-MCDA

#### **1. Introduction**

Groundwater is the world's largest freshwater resource. It provides an increasing quantity of water for irrigated agriculture and hence for global food security. If groundwater abstraction exceeds the natural recharge for a long period of time, intensive use and induced groundwater depletion occur. Depletion is widespread in large groundwater systems of the world [1]. Excessive extraction for irrigation where groundwater is slowly renewed is the main cause of the depletion, and climate change should exacerbate the problem in some regions where natural recharge is expected to decrease (the Mediterranean area for example). The effects of groundwater depletion are complex and dependent on the aquifer, and include (i) lowering of water tables leading to increased cost of pumping or drying up of wells [2]; (ii) reduced groundwater baseflow to streams, springs, and wetlands affecting ecosystems [2]; and (iii) land subsidence potentially damaging buildings and infrastructure [2]. Lowered water tables can lead to salinization by saltwater intrusion in coastal regions. Similarly, groundwater depletion can promote the spread of other types of pollution [2].

Managed aquifer recharge (MAR) constitutes very promising solutions for dealing with water shortage, water level depletion, groundwater pollution, and saline water intrusion [3]. It consists of infiltrating water abstracted from surface water resources (rivers, streams, lakes, etc.) through infiltration basins (indirect recharge) or injection wells (direct recharge [4]) in order to increase the natural groundwater recharge.

It is important to properly locate MAR systems according to the infiltration characteristics of the soil, the aquifer capacity to store water, the water resource location, and land use constraints. The site suitability assessment for MAR can be achieved by combining multi-criteria decision analysis (MCDA) for solving spatial problems with geographical information systems (GIS). For that purpose, many studies have been published in the literature describing methods for mapping the technical suitability of MAR solutions, most of them using GIS tools [5–7]. They combined the spatial analysis capacity of GIS with MCDA methodology that guides the decision making process—the resulting approach is called GIS-MCDA and has been recently implemented on web tools [8].

Despite many benefits and demonstrated advantages of the MAR, the growth of this solution has been much lower than expected due to the lack of a sound economic feasibility analysis. The performance and cost–benefit analysis of MAR scheme are key factors for the sustainability of this solution [9]. The costs of MAR schemes are influenced by a wide variety of hydrogeological, socio-economic, and legal and institutional factors [10].

A short review of GIS-MCDA for suitability mapping of MAR schemes shows that, except for a few studies [11], the cost is rarely included in such an analysis, maybe because of a lack of feedback on costs and financial data on MAR devices [10]. Despite the fact that several environmental variables (such as soil infiltration rate [12]) constitute a surrogate for economic evaluation, a specific economic analysis of MAR can bring about substantial information into a GIS-MCDA for MAR location [11].

In this study, we proposed to elaborate the cost function of recharge devices, taking into account capital and operating costs in order to compute the levelised cost of MAR. This cost function was spatially distributed in order to build a map of levelised cost in the study area. This method allowed for the identification of the part of a territory where the cost of MAR is expected to be lower compared with other regions, subject to the hydrogeological characteristics that affect aquifer storage capacity, the ability to recover water for high valued uses, and the environmental impacts of imposed changes to groundwater conditions.

#### **2. Methodology**

#### *2.1. MAR Project Design*

One MAR scheme can be divided into several engineering components (Figure 1): water abstraction system from the surface stream, water transfer pipe, pretreatment system, infiltration basin, and surface and groundwater monitoring.

**Figure 1.** Main engineering components of a managed aquifer recharge (MAR) scheme with infiltration basin and associated investment (IC) and operating costs (OC).

Several parameters or natural characteristics constitute a set of values that characterize the MAR scheme project (Table 1). The size of an infiltration basin depends on the natural characteristics of the soil and the target water volume (*Q*) to be recharged. The soil infiltration rate (*i*) is a key parameter as the surface area (*SB*) of the required infiltration basin is inversely proportional to this soil characteristic. The basin size is also linearly proportional to the target volume of recharge. The rate and the yearly duration of infiltration (*N*) are dependent on the water availability into the surface reservoir and provide the instantaneous flow rate (*q*) that are taken into account for calculating the diameter *di* of the pipes necessary for transferring the water from the abstraction place to the infiltration basin. Distance *D* and elevation difference *Z* between the surface water resource and the infiltration basin constitute the main characteristics to design the transfer infrastructure. The main parameters of such a project are listed in Table 1. These parameters are used for estimating the cost of the MAR scheme. Other site-specific parameters may also be incorporated in the analysis.


**Table 1.** Main parameters of an MAR scheme with infiltration basin project design.

#### *2.2. Economical Approach-Cost Function*

On the basis of the parameters/characteristics from Table 1, a cost function was built in order to compute the capital and operating costs of a MAR scheme using infiltration basins. These costs are described below according to the engineering component to which they are associated (Figure 1) and summarized in Table 2.

#### 2.2.1. Investment Costs

The investment costs of a MAR project such as an infiltration basin cover seven main expenditure items:



#### 2.2.2. Operating Costs

Operating costs cover the operating and maintenance costs of the MAR device. These are annual and recurring costs, expressed in €/year. These expenses can also be grouped into seven main items:


#### 2.2.3. Levelised Cost

For a MAR scheme, the levelised costs can be defined as the constant level of cost each year to cover all the capital, operating, and maintenance expenses over the life of a MAR project divided by the annual volume of recharge (or infiltration). Levelised costs provide an effective means to compare the costs of water from alternative projects [13]. The levelised cost takes into account the duration life (*T*) of the MAR scheme and the depreciation (or discount) rate *r*, which is the rate at which the value of an asset is reduced each year.

The levelised cost is computed using equations described in Table 2, some of them from the cost Observatory of Rhône Mediterranean Corsica Water Agency (AERMC). The latest provided the levelised cost of pretreatment; therefore, for this cost only, we used assumptions on levelised cost. Finally, the levelised cost is computed using Equation No. 2 (Table 2), considering life duration of the MAR scheme and the discount rate.


**Table 2.** Summary of costs for a MAR scheme (recharge through infiltration basin option). α and β are fractional parameters in order to define specific costs as a fraction of other costs.

#### *2.3. Costs Mapping Method*

In the mapping approach, the study area is gridded according to the DEM resolution (50 × 50 m cell size). In each cell, a land value is determined according to local databases and the infiltration rate is deduced from permeability maps built for numerical modelling studies of the aquifer. For each cell (column *i*, raw *j*), shortest distances *D<sup>n</sup> i,j* to surface resources (stream or lake) *n* are computed along with the corresponding elevation difference *Z*<sup>n</sup> *i,j* between the abstraction cell and the infiltration cell (Figure 2). Using these data, one levelised cost is computed for each cell and *n* associated surface water resources. Then, for building the cost map, the minimum levelised cost is calculated among the *n* available resources according to:

$$LC\_{i,j} = \min\left(LC\_{i,j'}^1LC\_{i,j}^2 \dots LC\_{i,j}^n\right) \tag{1}$$

The minimum cost is therefore considered for each cell of the analysis domain.

**Figure 2.** Cells gridding of the study area (example for a case with *n* = 3 available water resources: two streams and one lake). *D*<sup>n</sup> *<sup>i</sup>*,*<sup>j</sup>* and *H*<sup>n</sup> *<sup>i</sup>*,*<sup>j</sup>* are, respectively, minimum distance and elevation difference between the cell (*i,j*) and resource n.

#### **3. Reference Case Study**

#### *3.1. Case Study Description*

The reference case used was the Vistrenque and Costières Plain case study (VCP). This aquifer is located in Southern France between the Gardon River to the east and the Vidourle River to the west (Figure 3). The VCP area consists of a plain (Vistrenque) and a plateau at very low altitude (Costières) bordered to the north by the Nîmes garrigues and to the south by the Rhône plain and the Petite Camargue. Several types of unconsolidated rocks, among which alluviums largely dominate, constitute the aquifer. The VCP aquifer is unconfined on 84% of its surface area and confined on the rest.

Up until now, the aquifer is considered as being in a fragile hydraulic equilibrium, but the expected climate change impact on natural recharge and the increase of water abstraction induced by population concentration should lead to a potential water table decline, as already observed in the past during dry periods. Apart from water saving measures, MAR through infiltration basin using available surface water constitutes a possible alternative solution.

There are several surface water resources in the study area:



**Figure 3.** Map of the Vistrenque and Costières Plain (VCP) case study. Available surface water for MAR purposes is identified (in blue: rivers and streams; in green: Bas-Rhône and Languedoc regional water company (BRL) canal network).

#### *3.2. MAR Design and Characteristics*

The characteristics of the MAR scheme project on the VCP case study are summarized at Table 3. In that case study, the recharge rate objective was fixed at 1 Mm3/year for 8 months (*N* = 243 days/year) because surface water (canals or rivers) was not available during 4 months/year in low stages. For the reference case, we assumed a distance and an elevation difference between the surface resource and the infiltration basin, respectively, of 1000 m and −10 m. We considered that the solution of free water (from the river) was preferable to the canal water which is costly. We assumed that a primary pretreatment was necessary to remove the silt and fine material present in the river water. Its levelised cost was *LC*<sup>4</sup> = 0.10 €/m3. Other parameters (regarding basin dragging and maintenance) are listed in Table 3.


**Table 3.** Main characteristics of the MAR project on the VCP case study.

#### **4. Results**

#### *4.1. Reference Case*

The total investment cost obtained for the VCP case study was €1.7 million. The predominant cost items were the cost of implementing the pretreatment system (*IC*4), as well as the cost (*IC*2) of installing the water abstraction system (here, the reference example considered a river intake), corresponding to 53% and 23% of the total investment cost, respectively (Figure 4). The costs related to land purchase (*IC*5) and monitoring costs (*IC*7) represented a negligible part of the total investment costs. The investment cost of transferring water (*IC*3) was low compared to *IC*<sup>4</sup> and *IC*2. In cases where the distance *D* (1000 m in the reference example) between the surface water resource and the recharge area is higher, this cost item can take a significant part of the total investment cost.

The graph below summarizes the operating costs for the same reference case (Figure 4). The cleaning was fixed every *Nc* = 5 years, on a removal and replacement of *Hc* = 30 cm of gravel pack. The total operating cost obtained was nearly 130,000 €/year. The predominant cost item was the cost related to the pre-treatment of water (*OC*4), corresponding to 54% of the total operating cost (Figure 4). The water transfer cost (energy, *OC*3) was reasonably high in our reference example because

an average altitude difference (−10 m) was chosen. If the water supply to the device was gravitational, *OC*<sup>3</sup> would have been negligible. In the case where water was drawn from canals or water networks, the water purchasing cost *OC*<sup>1</sup> became one of the main cost items. Although it was generally perceived as high by operators, the cost of maintaining the basin and its surroundings (*OC*5) was found to be low compared to other cost items, such as *OC*4.

In this reference case, both investment and operating costs were dominated by pretreatment cost (i.e., in this case an assumption of a pretreatment levelised cost of 0.10 €/m3).

**Figure 4.** Partition of the various cost items for the reference case: (**a**) investment costs, *IC* (in €) and (**b**) operating costs, *OC* (in €/year).

#### *4.2. Levelised Costs Mapping*

The levelised cost map was obtained for the VCP case study applying Equation No. 2 (Table 2) using parameters from Table 3 and spatial variables distributed on the VCP maps (Figure 5a–d). The map of levelised cost is presented below (Figure 5e). The black areas correspond to the sectors excluded from the cost analysis (mask created from the unfavourable sectors from land use analysis), in order to isolate urbanized, artificialized areas, where it would be impractical and less interesting to install an MAR device (in addition to the difficulty of estimating land costs on these sectors).

The levelised cost LC ranged from 0.13 €/m<sup>3</sup> (along the Vidourle) to nearly 0.55 €/m<sup>3</sup> (on the contours of the entity; Figure 5). The average *LC* (average over the whole entity) was 0.29 €/m3. The most prominent criteria were:


**Figure 5.** VCP case study: (**a**) map of soil infiltration rates *i* (m/day), (**b**) land market value *LMV* (€/m2), (**c**) distance to one river *D* (Vidourle in this case), (**d**) elevation difference *Z* with closest river/stream, (**e**) minimum levelised cost *LC* map (in black: areas identified as unfavourable to infiltration basins due to land use constraints, most of the time corresponding to urban areas).

#### **5. Discussion**

#### *5.1. Sensitivity Analysis*

A systematic sensitivity analysis was performed to determine the effect of various parameters on the levelised cost of the MAR scheme. Sensitivity is defined as the rate of variation in one factor with respect to a variation in another factor. The normalized sensitivity is used to compare parameters and is defined as [14]:

$$S\_{i,t} = \frac{\partial O}{\partial P\_i / P\_i} \tag{2}$$

where *Si,t* is the normalized sensitivity of *i*th input parameter at time *t*, O is the output function of the system (i.e., the levelised cost in this case), and *Pi* is the ith input parameter of the system (in our case: T, Q, r, D, etc.). The partial derivative of this equation can be approximated by a forward differencing formula as [15]:

$$\frac{\partial O}{\partial P\_i} = \frac{O(P\_i + \Delta P\_i) - O(P\_i)}{\Delta P\_i} \tag{3}$$

The latest equation measures the influence that the fractional variation in a parameter, or its relative error, has on the output [15].

It must be specified that it is not a GIS-MCDA sensitivity analysis but a sensitivity analysis of the levelised cost of MAR scheme in the VCP case that has been computed for all the input parameters. They can be classified into three groups: (i) group 1 contains the highly sensitive parameters: life duration (*T*), recharge volume (*Q*), distance (*D*), and elevation difference (*Z*) between water uptake and infiltration basin, and water purchase and pretreatment costs; (ii) group 2 contains moderately sensitive parameters: yearly availability of water (*N*), soil infiltration rate (*i*), and discount rate (*r*); (iii) group 3 contains the lowly sensitive parameters: land market value (*LMV*), infiltration basin depth (*p*), duration between basin desilting (*Nc*), and thickness (*Hc*) of sands to scrab. The results for four of the most sensitive parameters are illustrated at Figure 6 with the following range of variation being explored: system life duration between 5 and 35 years, distance *D* between 0 and 5000 m, annual recharge volume between 0.01 and 5 Mm3/year, and water treatment cost between 0 and 0.30 €/m3. The cost sensitivity is positively linearly dependent on distance between abstraction and infiltration locations, and on water treatment cost, which tend to dominate the levelised cost beyond median values. The levelised cost appears to be highly dependent on annual recharge volume and life duration of the system, especially at low values of these parameters. This means, for example, that the levelised cost can be highly reduced by increasing the annual recharge volume up to 1 Mm3/year (≈0.2 of parameter range, *<sup>x</sup>* axis of graph at Figure 6). Regarding the system life duration, the minimum life duration should be above 20 years (>0.5 of parameter range).

**Figure 6.** Levelised cost sensitivity analysis to water treatment cost, recharge water volume, MAR system life duration, and distance between the abstraction and recharge locations.

#### *5.2. Approach Limitations and Outlines*

This methodology relies on several assumptions regarding various costs and technical characteristics of the MAR scheme, which result in uncertainties in the final computed levelised cost screening. Therefore, the method should not be used as an accurate tool in a prefeasibility analysis but as a tool to compare several options concerning, for example, (i) the location of the MAR scheme, (ii) the water resource which will be used, or (iii) the required pretreatment processes. The tool can also be improved during the pre-project analysis as new data and information are collected in engineering studies.

The various criteria considered in the cost function are derived from various sources; economic and financial feedback; and, in some cases, are based exclusively on expert opinion, in the absence of information from the literature. It is therefore important to consider that the costs obtained at the end of this analysis are orders of magnitude, based on a certain number of assumptions. For this reason, the method should be used in a relative way, with the objective to site projects at locations with relatively low costs.

The cost assessment carried out in this study is similar to a cost–efficiency analysis (CEA), considering that all the volumes of water brought to the infiltration basin are recharged and stored in groundwater. No consideration was given to the capacity of the aquifer to store infiltrated water, nor the ability to retain it so that it can be recovered for high-valued uses. Part of the recharged water may enhance the discharge of groundwater to watercourses or other aquifers. Additional hydrogeological investigations are needed in order to evaluate the contribution of recharged water to intended economic and environmental benefits. The analysis carried out did not include the assessment of such benefits of MAR projects. This may require aquifer characterization and hydrodynamic modeling of the site.

#### **6. Conclusions**

This methodology was provided in order to approximate the levelised cost of an MAR scheme using an infiltration basin. Uncertainty in several input parameters and the lack of economic and financial feedback on MAR system costs introduce uncertainty into the calculated levelised cost.

The developed tool should be used as a way to identify the sensitivity of the cost for several input parameters, as well as guiding the sitting of the MAR device in a relative way. It should help in making decisions on the design of the MAR system. The method is general and can be applied in other contexts and other countries where information on MAR costs is available thanks to economic feedback.

Finally, this information on economical feasibility should be followed by more classical suitability analysis such as those relying on soil characteristics, aquifer storage capacity, and land use in order to properly site the MAR scheme. The levelised cost provides an effective means to compare the costs of MAR with alternative water projects.

**Author Contributions:** J.-C.M. and J.-D.R. had the original idea; M.B. and J.-C.M. conceived the methodology; F.M. and M.B. developed the tool; J.-F.D. developed the GIS application; F.M. applied the methods/tools and analysed the data; Y.C. contributed to materials and analysis tools; J.-C.M. wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Bureau de Recherches Géologiques et Minières (BRGM, French Geological Survey) and Agence de l'Eau Rhône-Méditerranée-Corse (Rhone Mediterranean Corsica Water Agency).

**Acknowledgments:** This study results from a scientific collaboration between Bureau de Recherches Géologiques et Minières (French Geological Survey) and Agence de l'Eau Rhône-Méditerranée-Corse (Rhone Mediterranean Corsica Water Agency).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**

The following abbreviations are used in this manuscript:


#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
