Introduction

Climate change exacerbates global water security challenges particularly in arid to dry sub-tropical areas with high water demand (Stringer et al. 2021). Many towns in Australia suffer water shortages during droughts (Kiem and Austin 2013). A water industry survey in 2015 reported that only 46% of industry respondents and 36% of community respondents reported being ‘very confident’ or ‘somewhat confident’ that ‘Australia currently had sufficient water security to meet all social, environment and economic needs’, and reported an even more pessimistic view of water security in regional area towns (AWA 2015). Climate change is projected to worsen the outlook for water availability around the world (Rodell and Bailing 2023), including southern Australia (Cai et al. 2014) where groundwater resources in particular are likely to be affected (Walker et al. 2021). Historical underinvestment in water supply infrastructure and inadequate maintenance has heightened risks of water availability and water quality shortfalls (Armstrong and Gellatly, 2008).

Managed aquifer recharge (MAR) is the intentional, managed recharge of aquifers for later use or environmental benefit (Dillon 2005). MAR can be operated according to different objectives, e.g. balancing seasonal demand and supply or facilitating water recycling. MAR can also be used for increasing long-term water security and drought resilience as seen in various parts of the United States (Megdal and Dillon 2015; Scanlon et al. 2016) and Perth’s potable water supply (Gao et al. 2014). Studies have indicated significant potential for MAR in Australia’s Murray-Darling Basin (MDB) (Fuentes and Vervoort 2020; Gonzalez et al. 2020, 2021; Khan et al. 2008; Page et al. 2021), and northern Australia (Evans et al. 2017; Vanderzalm et al. 2022) and key irrigation regions around Australia (Page et al. 2021). Aquifer storage is effectively a reservoir that does not need to be built, the main infrastructure costs are limited to accessing the aquifer for recharge and recovery (Dillon et al. 2010). Limited understanding of the potential costs and benefits of MAR systems is cited as a barrier to wider adoption of MAR (Maliva 2014). International studies on the economics of MAR to date have focussed on retrospective analyses of existing sites, or specific feasibility assessments of proposed or conceptualised schemes. This has limited transferability to different locations, scales and operating conditions.

Transferability of results from case studies is especially important for proponents in regions where MAR projects do not exist and there is no local experience on which to draw from. Uncertainty over costs and economic performance has slowed adoption of MAR schemes globally (Maliva 2014). A framework for collecting and analysing financial costs of MAR systems was developed that proposed a classification system for MAR projects and costs and applies a cost effectiveness methodology based on recharge volume data, and capital and operating costs (Ross and Hasnain 2018). While based on a cost-benefit analysis framework, benefits were not included due to varying scheme objectives. A subsequent study provided an update to include several additional sites, estimate costs of recovered water, and quantified benefits and benefit-cost ratios based on the objectives of the different schemes (Ross 2022). Higher ratios were reported for schemes where the relative levelised cost of MAR supply was low compared to either increased agricultural production, the cost of alternative supply from a new water treatment plant, or the current average water tariff in the region.

Several factors potentially affect MAR scheme performance and costs including efficiencies of scale, operating rules, aquifer recharge and storage properties, and a range of social, legal and institutional factors (Ross and Hasnain 2018). Ross and Hasnain (2018) made several priority recommendations for assessing costs of MAR. Firstly, inclusion of time-varying operating costs, secondly, disaggregation of capital and operating costs to enable detailed examination of factors, and thirdly inclusion of a wider range of schemes scales, treatment trains, recharge rates, and storage efficiencies. Disaggregated costs for ten conceptual MAR schemes in Northern Australia were examined that revealed different scheme types and scales had differing proportions of capital and operating costs (Vanderzalm et al. 2022). An earlier study separated capital and operating expenses for estimates of conceptual infiltration basin and injection well schemes for five sites in the Murrumbidgee River catchment in the Murray-Darling Basin (Khan et al. 2008). While they used groundwater modelling to estimate aquifer recharge and storage potential to parameterise the cost assessments, uncertainty in the hydrogeological calculations and capital and operating cost assumptions were unaccounted. Both studies also assumed uniform operation across scheme lifetimes with full use of recharged water and no storage losses for calculating costs. For multi-year storage and drought supply, it is important to consider any losses as this directly influences the proportion of recoverable water (recovery efficiency), and consequently the unit cost of recovered water. Further disaggregation of capital and operating costs into separate items would provide more specific guidance on the origins of MAR scheme costs.

Testing the sensitivity of input parameters and assumptions is generally required according to economic assessment guidelines for infrastructure projects to ensure the process is robust to potential changes and ranges of scenarios. Probabilistic modelling is useful for high-risk projects and where there is greater uncertainty (European Commission 2014; Infrastructure Australia 2021). Economic assessments for a variety of MAR case study sites around Europe adopted this approach to quantify uncertainty through probability modelling and univariate sensitivity analyses (Damigos et al. 2016). Results captured the range of plausible economic net present values of the options and scenarios proposed for each site leading to more robust conclusions regarding social acceptance and justification for investment. Probabilistic modelling and sensitivity analyses were conducted for aggregated operating and maintenance costs, expected revenues and assumed benefits and externalities. This accounted for ranges in net present values but did not reveal sensitivities of individual parameters, assumptions or interactions. A variation on break-even economic analysis (Jovanović 1999) was applied to compare MAR options with equivalent surface storage to determine cross-over points where conditions determine that MAR is a financially better option (Arshad et al. 2014). The methodology represents another way of assessing economics under uncertainty to help decision making. Variability in hydrological and financial cost and benefit factors is captured and sets of conditions at which one option becomes worthwhile compared to another are identified. The method is applicable when there are multiple options for a given scenario to compare.

This study aimed to address the gaps and limitations found in previous studies regarding transferability, transparency and uncertainty to provide estimates of the likely range of levelised costs of water supply for a range of MAR scheme types, scales and conditions to enable translation of results in various locations. Two common MAR scheme types, well injection and infiltration basin systems, that account for about 50% of schemes globally (Stefan and Ansems 2017) were assessed. Well injection is used to recharge deep and confined aquifers under pressure whereas infiltration basins percolate water under gravity to unconfined aquifers (Dillon 2005). Levelised cost distribution curves were calculated for a range of volumetric capacities relevant to town and small city drinking water supply for a range of typical operating conditions related to recharge and recovery rates (e.g. bore yields, permeability) and recovery efficiencies (or aquifer storage efficiencies). Disaggregated cost and global sensitivity analyses were conducted to identify the most important factors affecting costs and uncertainty for different systems. Several Australian regional area town water supply systems with a history of shortages were used as case studies to demonstrate the application of the methods and outputs for preliminary estimation of MAR scheme costs.

Materials and methods

Analysis framework and assessment model

The overarching analysis framework applied in this study is depicted in Fig. 1. Input variables are passed into an iterative model using water balance and financial assessment functions for a range of scheme scales and operating conditions. Outputs include levelised cost distributions, sensitivity analyses and disaggregated capital and operating costs. Case studies of regional area towns in the Murray-Darling Basin are oriented within the results envelope to indicate expected MAR scheme costs and sensitivities according to local scales and conditions.

Fig. 1
figure 1

Analysis framework showing input variables, iterative hydro-economic model functions, outputs, and case study orientation within the results envelope

Full size image

Levelised cost is a common metric for assessing water infrastructure projects that accounts for amortised capital and operating expenses over the life of the project per unit volume of water supplied (Ross and Hasnain 2018; Vanderzalm et al. 2022). While based on a cost-benefit analysis framework, levelised cost avoids assumptions about benefits, enables comparison with similarly costed MAR schemes, alternative water supply technologies, and allows comparison across system scales and end uses. Assessments assume full-cost recovery to ensure economic sustainability consistent with international pricing principles (e.g. Article 9 of the European Union Water Framework Directive, Australian National Water Initiative). Analyses are conducted over a 50-year project horizon in line with other assessments of MAR schemes and water infrastructure project lives reported internationally (OECD 2020). All monetary values presented in this study are reported in Australian dollars adjusted for inflation to the June 2021 reference base (2021 AUD).

Figure 2 shows the general structure of the assessment model. The model was driven by a 50-year time series of river discharge that, using variable thresholds, triggered recharge and recovery. Scheme capacity was set as an initial parameter, variable bore yield or infiltration rate determined the number of bores or area of basin required to achieve the target recharge capacity. Different ranges of recovery efficiency were variably input which determined stored and recovered volumes. The volumetric time series were used to calculate volume dependent costs. Operating costs in recharge and recovery years were discounted over the 50 years. Model outputs included total volumes recharged and recovered, percent recharge recovered, levelised costs of supply, total operating and capital costs, and further cost breakdowns for specific components of the modelled systems e.g. monitoring, pre-treatment, maintenance.

Fig. 2
figure 2

Hydro-economic model structure applied to well injection and infiltration basin simulations (superscripts link to Table 1)

Full size image

The models, scenarios and iterations used are summarised in Table 1. Two sub-models were created, one to simulate well injection and another to simulate infiltration basin systems. For each sub-model, four scenarios were used defining different ranges for injection/infiltration rates and aquifer recovery (or storage) efficiencies. 10,000 samples were drawn from within each range to capture uncertainty and enable sensitivity analyses (Section “Sensitivity analysis”). The analysis was repeated for each scenario for seven recharge capacities, 0.25, 0.5, 1, 2, 3, 4 and 5 million cubic metres (MCM) per year, based on town water demand shortfalls during drought (see Supp 1) to cover a plausible range of scales of operation for town water supply augmentation. This yielded a total of 70,000 model runs for each scenario. The input variables and ranges for well injection and infiltration basin sub-models are described in Table 2.

Table 1 Sub-models and scenarios of variable ranges for calculating MAR scheme cost distributions across a range of scales and operating conditions
Full size table
Table 2 Input variable ranges and assumptions for MAR scheme cost distribution modelling (*superscripts link to Fig. 2)
Full size table

The sub-models were written in Python 3 (Van Rossum and Drake 2009) using a water balance function to account for recharge, storage and recovery triggered by a hydrological time series, and a net present value function for financial assessment. Equations are given in full in Supp 1. The approach made several general assumptions:

  • A 50-year asset life for MAR schemes modelled using an annual time step; infrastructure with shorter service lives (e.g. pumps, instrumentation) were replaced at discounted future costs;

  • Recharge and recovery for every 50-year model run was triggered using a historical time series of average annual river discharge;

  • Recharge rates and volumes not limited by aquifer extent or hydraulic properties;

  • Water sourced for MAR had an associated opportunity cost to reflect the scarcity of the resource;

  • The number of injection bores or infiltration basin area required to achieve each target recharge capacity (0.25-5 MCM/y) was determined by injection or infiltration rate and annual days of operation variables;

  • Annual days of operation variable represents non-continuous scheme operation e.g. to allow time for maintenance;

  • Variable annual operating costs depending on volumes recharged and recovered amortised over scheme life;

  • Initial capital costs incurred, and construction completed in first year, operations allowed to commence in second year;

  • No residual value of infrastructure included at the end of the evaluation period;

  • Land costs were not included as schemes were not location specific.

Water sourced for MAR was assumed to have an associated opportunity cost reflecting the scarcity of the resource. Market prices and volume weighted average trade prices are commonly used water valuation methods (Seidl et al. 2020). The water market system in Australia’s Murray-Darling Basin is recognised as one of the most advanced in the world (Grafton et al. 2011). Across the MDB, volume weighted average water trade price records begin in 2008. As a 50-year assessment period was used a proxy dataset was needed. Volume weighted annual average trade prices across the MDB from 2008 to 2022 (BOM 2023) were significantly lower (Welch’s t-test p < 0.01) when annual average river discharge exceeded 100 m3/s at Lock 1 Blanchetown (BOM 2022). In years when river discharge exceeded 100 m3/s, average water prices were $0.02–0.09/m3, below this discharge rate average prices were $0.11–0.52/m3 (see Supp 1). The analysis therefore used the price for the former, with annual average river discharge at Lock 1 for the 50-year period 1972–2021, used to trigger recharge and recovery based on an uncertain threshold of between 100–120 m3/s to represent variance in operating rules. MAR can use other water sources such as urban stormwater and recycled water for recharge; levelised costs of these sources are typically one or two orders of magnitude more expensive than the surface water opportunity costs used here (WSAA 2020).

Sensitivity analysis

A global sensitivity analysis approach was adopted to evaluate the influence of each variable on the levelised cost of recovered water while exploring the full parameter space and including interactions between simultaneously varying factors. This approach is preferred where there are numerous interdependent variables as local sensitivity analyses are generally limited to highly linear or additive models typically varying one parameter at a time relative to a baseline scenario (Saltelli and Annoni 2010). The sensitivity analysis was coupled to the hydro-economic model described in Section “Analysis framework and assessment model” (summarised in Fig. 2) written in Python 3 (Van Rossum and Drake 2009) using the SALib package for Monte-Carlo sampling and independence measure calculations (Herman and Usher 2017). Sensitivities were calculated using the Delta-moment independent measure. This measure was selected as it considers the entire distributions of inputs and outputs, accounts for correlations among parameters, and does not require a structured parameter sampling scheme to determine the influence of inputs on output uncertainty (Borgonovo 2007). Latin hypercube sampling was chosen as it efficiently generates random combinations of values across their ranges with a minimal number of samples (Helton and Davis 2003). With about 20 parameters for the well injection and infiltration basin models, 10,000 iterations were used to fully cover the input space. As multiple scenarios were run, sensitivity analyses were performed on ‘best’ and ‘worst’ cases for the lowest and highest scheme design capacities to compare differences in parameter sensitivities. Best cases assumed injection rates of 25–33 L/s and basin infiltration rates of 0.8–1.0 m/d, and recovery efficiencies of 95–100%. Worst cases assumed injection rates of 5–10 L/s or infiltration rates of 0.1–0.2 m/d, and recovery efficiencies of 80–85%. Results were benchmarked against a random variable (0–1) to assess their statistical significance.

Case study areas

To illustrate the application of the estimated cost distributions for preliminary, indicative estimates of MAR scheme costs, six towns in central and western New South Wales (NSW) regions of the Murray Darling Basin were used as case studies. These towns were selected based on their histories of water restrictions during droughts and recent investment in investigations or infrastructure to improve water security (NSW Government 2017). The conceptual MAR scheme components for the case study towns of Tamworth, Bathurst, Dubbo, Moree, Narromine and Gunnedah are described in Supp 1. Estimated storage potential for each conceptualised scheme was made based on depth to water leaving a 4 m buffer to the surface to avoid water logging and salinisation (NRMMC-EPHC-NHMRC 2009), assuming radius of influence of 1 km, and an effective porosity of 0.2–0.25 typical of alluvial aquifers (Heath 1983). The annual recharge capacity was set based on the rate required to store the potential volume over 4 years. A summary of critical differences influencing costs is discussed with respect to the type of information required to efficiently reduce uncertainty in entry-level scheme viability analysis.

Results and discussion

This study uses three different visualisations to support interpretation of results; (1) levelised cost distribution curves for different scales and operating conditions; (2) heat maps of disaggregated capital and operating expenses; and (3) plots of first-order global sensitivity indices at either ends of scheme scales and operating conditions. This is done for well injection and infiltration basin systems and results are compared with published costs of MAR systems in the international literature. Six case study towns in regional Australia are then used to demonstrate the application of results for indicating the range of potential costs of MAR that could be expected.

Levelised cost distributions of MAR systems

Levelised costs of recharge for well injection systems ranged from around $0.20–0.30/m3 for large volume systems to $0.75–1.05/m3 for small systems (Fig. 3a). Recharge costs do not account for storage losses so differences between curves in Fig. 3a were driven by bore yield, which determined the number of bores required to achieve the target recharge rate. Lower yield requires more bores to achieve the same recharge rate. Estimated costs are comparable to $0.10–0.13/m3 of water recharged for similar capacity schemes (3–7 MCM/y) reported by Ross and Hasnain (2018) using well injection with natural water (Table 3). Recharge costs are also comparable to estimates of $0.15–0.37 for conceptual injection systems of 1–5 MCM/y recharge capacity reported by Vanderzalm et al. (2022). A small scheme in Ovezande, Netherlands (0.006 MCM/y) reports recharged water costs of >$2/m3. Large systems in Texas and California (Orange) in the order of 80–300 MCM/y have reported recharge costs around $0.38/m3. A feasibility assessment of a 0.2 MCM/y injection scheme in the Namoi region of Australia was estimated at $0.30/m3 (Dillon and Arshad 2016). This is around a third of the cost of a similar capacity scheme indicated in Fig. 3. Dillon and Arshad (2016) estimated capital costs were 10% of total levelised costs and operating 90% while this study estimated that for a scheme of 0.25 MCM/y, total capital costs were around 25% and operating costs were 75%. Initial capital costs have a relatively strong influence as they are not discounted over the life of the project, especially for small schemes with higher proportional capital costs.

Fig. 3
figure 3

Distribution of levelised cost estimates (mean ± SD) of recharge (a) and recovery (b) for different well injection scheme conditions and scales; points indicate costs of conceptualized schemes for case study areas

Full size image
Table 3 Levelised costs reported internationally for existing MAR sites and feasibility assessments
Full size table

Estimated levelised costs of recovered water were 30% higher than recharged water costs for schemes with high recovery efficiencies (> 95%), and up to double the costs for schemes with low recovery efficiency (RE) (80–85%) (Fig. 3b). Recovered water costs included pumping and accounted for a range of recovery (or storage) efficiencies. Visual comparison of the spread of curves in Fig. 3a with 3b indicates that RE was a markedly stronger influence on cost than bore yield. This is also borne out by the sensitivity analysis (later discussed in 3.2 and shown in Fig. 4). Levelised costs of recovered water ranged from $0.60-$2.20/m3 for schemes < 1 MCM/y capacity, and $0.30-$0.80/m3 for schemes > 4 MCM/y capacity.

Fig. 4
figure 4

Injection well scheme first-order input variable sensitivities (S1) for scenarios representing smallest and largest design capacities under ‘best’ and ‘worst’ case recharge rates and recovery efficiencies

Full size image

In comparison, estimates for small capacity schemes of 0.34 and 0.68 MCM/y supplying drinking water in Florida were $0.71 and $0.35/m3 respectively (Table 3), however, these figures were based on approximate annual costs (Maliva 2014). For another 6.7 MCM/y scheme supplying drinking water in Florida, recovered water costs are reported as $0.13/m3. Recovered water cost for a 0.125 MCM/y scheme in Dinteloord, Netherlands was estimated at $0.98/m3 (Ross 2022). That scheme uses recycled water with high treatment costs although it is not clear how treatment costs were incorporated in the levelised cost estimate. For a scheme of that scale, it is surprising that the cost of advanced wastewater treatment would result in a levelised cost much smaller than reported ($1.80/m3) for a 14 MCM/y recycled water well injection scheme in Perth, Australia (Table 3).

A feasibility study of a 6 MCM/y injection scheme in Albacete, Spain estimated a recovered water levelised cost of $0.17/m3 (Rupérez-Moreno et al. 2017). Visually extrapolating Fig. 3b, it is estimated that this would be about half of what could be expected based on the current study using a central curve. Rupérez-Moreno et al. (2017) do not mention water treatment and refer to infiltration wells, and not injection under pressure, both of which would add capital and operating costs.

Khan et al. (2008) estimated a levelised cost $0.22/m3 for large injection systems of 60–90 MCM/y capacity using surface water in the Murrumbidgee region in Australia. They included a representative capital cost of ~$500k for a filtration pre-treatment system but specified no disaggregated operational costs associated with treatment or maintenance. The current study did not assess schemes of this scale, but it is likely that for high efficiency, high yielding operating conditions, larger scale injection systems could approach a similar cost. International studies indicate that efficiencies of scale eventually plateau as seen for the infiltration basin and injection schemes of 148 and 294 MCM/y in Orange County, California with levelised costs of $0.66 and $0.37/m3 respectively (Table 3).

Levelised costs of recharge for infiltration basins ranged from $0.04–0.08/m3 for schemes ≥ 2 MCM/y, and $0.05–0.14/m3 for schemes ≤ 1 MCM/y (Fig. 5a). This range is comparable to the recharge costs reported internationally for similar scale infiltration basin schemes ranging from $0.09–0.38/m3 (Table 3). The differences between curves in Fig. 5a are driven by infiltration rate with slower rates requiring a greater basin area to achieve the same rate with additional construction and maintenance costs. Variability was driven by initial capital costs that were proportionally higher for small volume schemes. Levelised costs of recharge levelled out at capacities ≥ 3 MCM/y indicating a limit to efficiencies of scale.

Fig. 5
figure 5

Distribution of levelised cost estimates (mean ± SD) of recharge (a) and recovery (b) for different infiltration basin scheme conditions and scales; points indicate costs of conceptualized schemes for case study areas

Full size image

Levelised costs of recovered water from infiltration basin systems were about 30% higher than recharge costs for schemes with high RE (> 0.95) and up to double the cost of recharge for schemes with lower RE (0.80–0.85) (Fig. 5b). Infiltration basin recovered water levelised costs for schemes of ≤ 1 MCM/y ranged from $0.09–0.24/m3 and for schemes of ≥ 2 MCM/y ranged from $0.06–0.14/m3 (Fig. 5b). Comparable recovered water costs for surface water infiltration basins within the range of capacities estimated here were between $0.08/m3 for a 6.45 MCM/y system in Finland (Tanttu and Jokela 2017) and $0.11/m3 for a 1.87 MCM/y basin in Iran based on a feasibility study (Kalantari et al. 2009), see Table 3. Levelised costs of recovered water for infiltration basins reported in the literature did not show a clear relationship between cost and scale however, this is only based on six examples with different end uses and assessment periods. For example, the recovered water cost of $1.19/m3 for a 23 MCM/y infiltration basin system in Finland seems very high compared to the other Finnish scheme and the Iranian example. This is mainly attributed to expensive treatment requirements for potable use for the Finnish scheme (Ross 2022). The current study however, estimated costs of supply of recovered water for supply to a drinking water treatment plant and did not include post-treatment costs for direct human consumption.

MAR costs reported in this study generally compare favourably against alternative storage and supply options. An Australian water industry study reviewed levelised costs for using recycled water (treated wastewater) to augment reservoirs and groundwater finding an average around $2.00/m3 from 31 projects assessed, while precinct scale stormwater harvesting (> 1 MCM/y) was around $0.60–2.50/m3 (WSAA 2020). Desalination can provide another source of water from brackish or saline groundwater or seawater. A review of costs per unit capacity found capital costs of $665–3900/m3 and operating costs of $0.52–0.88/m3 for schemes with a capacity in the range of 4–26 MCM/y assuming year-round operation (Pearson et al. 2021). Scale was an important determinant of costs as were brine disposal costs for some schemes, e.g. inland sites using brackish or saline groundwater. Capital costs for 5 MCM/y MAR schemes estimated here were about $2.40/m3 for well injection, and around $0.50/m3 for infiltration basin systems (Fig. 6a and b). Costs of dam storage were reviewed across 98 sites and found to vary from $0.43–2.21/m3 with a median of $1.07/m3 (Petheram and McMahon 2019). Dam costs per unit yield (89 sites) accounting for evaporation ranged from $0.72-$2.78/m3, with a median of $1.42.

Fig. 6
figure 6

Distribution of total capital costs (mean ± SD) for injection well systems (a) and infiltration basin systems (b) for different scheme scales and conditions

Full size image

Disaggregated costs and sensitivities

Total capital costs for injection well systems ranged from <$2 million to around $16 million (Fig. 6a), and from around $0.5 million to $3 million for infiltration basins (Fig. 6b) across the scales and conditions assessed. Capital costs were generally linearly related to scheme capacity with the recharge rate influencing the cost required to achieve the same volumes. Injection scheme capital costs were driven mainly by pre-treatment infrastructure (coagulation, multimedia filtration and granular activated carbon), and injection well construction (Fig. 7a). Capital costs of infiltration basins were largely determined by costs of feasibility studies (independent of capacity), purchase of pumps and pipes, and to a lesser extent, infiltration basin construction (Fig. 7b).

Fig. 7
figure 7

Disaggregated capital and operating average costs across well injection scheme sizes as a proportion of levelised cost of recovered water

Full size image
Fig. 8
figure 8

Disaggregated capital and operating average costs across infiltration basin scheme sizes as a proportion of levelised cost of recovered water

Full size image

Disaggregated capital and operating costs were dominated by pre-treatment operating expenses for injection well schemes of all sizes accounting for 40–70% of levelised costs of recovered water (Fig. 7). The distribution of costs for infiltration basin schemes, while more evenly distributed across scheme sizes, showed that capital expenses mainly associated with feasibility studies were a much higher proportion (20–50%) of levelised costs for smaller schemes (≤ 1 MCM/y) than for larger schemes (5–13%) (Fig. 8). Larger infiltration basin schemes had higher proportional operational costs related to opportunity costs of water (up to 40% of levelised costs of recovered water) and maintenance (up to 16%) and attracted higher capital costs for construction (mainly pumps and pipes, up to 25%).

The global sensitivity analysis showed that one of the most sensitive variables at either end of well injection scheme sizes or operating conditions on levelised cost of recovered water was the annual discount rate (Fig. 4). In contrast, discount rate was a less sensitive variable in the infiltration basin scenarios, especially low volume schemes (Fig. 9). This is due to the proportionally higher pre-treatment operating costs for all injection scheme sizes that were discounted across the 50-year assessment period (Fig. 7) compared to the infiltration basin schemes where operating costs were a lower proportion of total costs (Fig. 8). The ratio of total operating to total capital costs for injection schemes was 1.3–3.2 compared to 0.5–1.9 for infiltration basin schemes. Discount rate was also more sensitive for high volume injection and infiltration basin schemes and those with lower recharge rates and recovery efficiencies (Figs. 4 and 9). Similarly, this is attributed to the discounting of operational costs most of which were dependent on either recharge or recovery volumes.

Fig. 9
figure 9

Infiltration basin scheme first-order input variable sensitivities (S1) for scenarios representing smallest and largest design capacities under ‘best’ and ‘worst’ case recharge rates and recovery efficiencies

Full size image

The selection of the social discount rate to apply in financial and economic assessment is dependent on the nature of the investment being made, the industry sector and the funding body itself (e.g. public or private), among other potential variables. The Infrastructure Australia assessment framework recommends a central value of 7%, a minimum of 4% and a maximum of 10% (Infrastructure Australia 2021). This aligns with most Australian state guidelines and would be the range required for presenting evaluations for projects seeking public funding at these levels. MAR project proposals for town water supply would likely seek some level of public funding so this study evaluated the full recommended range. The appropriate rate to apply is generally a function of regional economic growth scenarios and project horizons and is a critical input in economic analysis particularly where public funding is concerned. Given its sensitivity, the main principle is that a range of rates should be tested that is contextually relevant.

Well injection pre-treatment operating cost was a highly sensitive variable across scenarios (Fig. 4). In comparison, basin maintenance was a relatively insensitive variable particularly for low volume infiltration basin schemes (Fig. 9). Infiltration basins assumed no treatment was required prior to recharge and that clogging risks could be managed through scraping of basin floors after every year of recharge for which the proportion of levelised cost was low (Fig. 8). While injection pre-treatment capital and operating costs were volume dependent (capital based on design capacity, operating based on recharged volumes), sensitivities were higher for larger schemes (Fig. 4). This is because pre-treatment costs were non-linearly related to volumes (see Supp 1) and were randomly varied by 20% to represent uncertainty. Consequently, the variation in costs for large schemes was higher and therefore more sensitive.

This study shares similar findings to others that have shown that the higher costs of well injection compared to infiltration systems are generally driven by costs of treatment to manage clogging risks (Ross and Hasnain 2018; Vanderzalm et al. 2022; Zheng et al. 2021). While clogging is a risk with any type of MAR system, well injection is susceptible to clogging due to physical, chemical and biological processes that are exacerbated by high hydraulic loads at the well face during injection. Aside from treatment to reduce suspended solids, nutrients and organic carbon, regular backflushing of injection wells, among other preventative measures, can minimise clogging and maintain adequate injection rates (Martin 2013). Results highlight the importance of understanding source water quality, treatment requirements and aquifer properties for managing clogging risks and to reduce uncertainty in viability assessments of well injection systems.

Provisions were made for costs of water quality investigations including clogging experiments in feasibility assessments for both well injection and infiltration basin schemes in this study in accordance with guidelines for MAR projects (Dillon et al. 2016; NRMMC-EPHC-NHMRC 2009). Costs of feasibility assessments were a larger proportion of levelised costs for infiltration basin schemes (20% for high capacity and 40% for low capacity) compared to well injection schemes (1–6%) due to the lower relative costs of other components of infiltration basin systems (Figs. 7 and 8). Feasibility study costs were assumed independent of scheme capacity so variation in this cost variable translated to much higher sensitivity for low-volume infiltration basin schemes (Fig. 9). Constraining feasibility assessment cost estimates and requirements commensurate with risks could reduce overall costs and uncertainty for smaller volume systems with lower maximal risk.

Recovery efficiency (RE) was a more sensitive parameter in best-case scenarios with high RE (0.95-1.0) for both well injection and infiltration basin schemes of any size, than worst-case scenarios with low RE (0.80–0.85) (Figs. 4 and 9). This was due to the higher variability in the volume of recovered water at high RE (coefficient of variation of 10% compared to 3% at low RE) determining the levelised cost of recovered water. RE represents losses in aquifers such as through mixing with saline groundwater or seepage from the target aquifer that reduces recoverable volumes. Annual average RE in the range 0.80 to 0.85 is representative of those reported in studies of freshwater mixing losses in brackish aquifers (Clark et al. 2015; Miotlinski et al. 2014; Pavelic et al. 2006). This study highlights the importance of understanding RE when considering long-term, multi-year storage for increasing drought resilience as the effects of RE are compounded the longer water is stored. Sites reporting only the cost of recharge or assuming all water is recoverable (see Table 3) miss this potentially critical factor. Assuming uniform operation of recharge and recovery, e.g. average annual recharge and recovery across the assessment period, simplifies assessments and for seasonal storage and supply scenarios this may be of lesser consequence with respect to storage losses. Even at high recovery efficiencies, variability in losses due to mixing with saline groundwater or through seepage into other stratigraphic units or to the surface can introduce substantial uncertainty in levelised costs of recovered water. Understanding potential storage losses to constrain this uncertainty is therefore critical in early planning and design stages for assessing scheme viability.

Opportunity costs of water were more sensitive for higher volume schemes and particularly infiltration basins (Figs. 4 and 9) where this was 20–40% of levelised costs of recovered water (Fig. 8). The opportunity cost of water for recharge was based on a water allocation trade price signal as an indicator of scarcity in a contested water resource area such as Australia’s Murray-Darling Basin. In the MDB, water allocations are set as proportions of a total water entitlement (permanent right) based on water availability in the system over time and are tradeable within defined zones. Trade price volatility is a feature of the Australian water market (Zuo et al. 2019). While volume weighted average trade price is a recognised valuation method used by banks, evaluators and water brokers (Seidl et al. 2020), it along with other signals remains subject to the “paradox of water pricing” where “the price of water almost never equals its value and rarely covers its cost” (Grafton et al. 2020). This study used the range of volume weighted average prices across the MDB over 15 years of historical trade when flows were above a threshold (see Supp 1). This was a simplification required to determine timing and associated costs of recharge whereas in practice, operating rules for recharge and recovery are more likely to be developed around fair market value rather than historical cost accounting (Seidl et al. 2020).

Case studies of regional area towns in the murray-darling basin

Two of the six case study areas, Tamworth and Bathurst, were conceptualised as well injection systems to penetrate surficial and interbedded clay layers indicated in these areas. The aquifers in both areas are narrow and shallow deposits of unconsolidated sediments in comparison to the alluvium lower in some of the other river valleys (e.g. Moree, Gunnedah, Narromine). This combined with relatively shallow water tables, meant the volumetric storage potential for MAR is likely to be limited to around 1 and 3 MCM for Tamworth and Bathurst respectively (see Supplementary 1). These volumes could be achieved over four years of recharge at 0.25 and 0.75 MCM/y. Tamworth and Bathurst have annual average potable water demands of 8.8 and 6 MCM respectively (NSW DPE 2022). MAR schemes recovering water at these capacities would meet 9% and 42% for up to 4 years of annual demand shortfalls due to 30% reductions in town water allocations as occurred during 2018-20 drought conditions at Tamworth and Bathurst respectively (WaterNSW 2022).

Groundwater salinity in both areas is fresh so storage losses due to water quality are likely to be low. Bore yields along with lithological information suggests that injection and recovery rates in these systems could be limited. Indicative levelised costs for these scheme capacities would be around the middle of the cost curves in Fig. 3b at around $1.50/m3 for Tamworth, and $1.10/m3 for Bathurst. Capital costs would be around $2 million for Tamworth and $3 million for Bathurst (Fig. 6a). Disaggregated costs for small volume well injection schemes were dominated by pre-treatment operating costs, and to a lesser degree, pre-treatment capital costs (Fig. 7). This was also reflected in the sensitivity analysis where the discount rate range had a strong influence on the variability of treatment operating costs (Fig. 4).

The treatment chain for well injection costed here used engineered chemical and mechanical devices following recommendations made by Page et al. (2014) for treating turbid river water to reduce clogging risks for injection. These could be substituted for passive treatment devices like sedimentation basins, constructed wetlands or biofiltration for water quality control prior to injection (Gonzalez et al. 2015; Pavelic et al. 2006). Water quality investigations and clogging studies recommended during feasibility assessments (Dillon et al. 2016; NRMMC-EPHC-NHMRC 2009) would help define treatment requirements and reduce uncertainty in treatment cost estimates.

MAR schemes for four case study areas, Gunnedah, Moree, Dubbo, and Narromine, were conceptualised as infiltration basin systems. These areas featured generally unconfined aquifers located on wider sections of river valleys. Storage potential in the Gunnedah area is likely to be limited by water table heights around 10 m below ground level, resulting in an estimated storage potential of around 4 MCM with the assumptions of porosity and radius of influence made (see Supp 1). At Moree, depth to water varies around 15–20 m and additional aquifer storage is approximated at 7 MCM, while at Dubbo, depth to water is around 20 m in some areas and storage is estimated at around 10 MCM. The vertically and laterally extensive aquifers near Narromine and deeper water table reaching 35 m or more resulted in a storage estimate of 20 MCM. These volumes could be stored in four to five years with scheme capacities of 1, 2, 2.5 and 5 MCM/y respectively. Gunnedah, Moree, Dubbo, and Narromine, have annual average potable water demands of 2.9, 3.1, 9.1, and 1.3 MCM respectively (NSW DPE 2022). MAR scheme with these capacities would meet 73% to well over 100% of annual demand shortfalls for 4–5 years if town water allocations were reduced by 30% during drought, as occurred in Dubbo during 2018-20 (WaterNSW 2022). In the case of Narromine, a large-scale scheme could be used to distribute benefits to other users and the environment or augment potable water supplies for longer periods.

Gunnedah has variable yields (6–47 L/s) and fresh to brackish groundwater so indicative levelised costs of recovered water could be around $0.10–0.17/m3 for a 1 MCM/y infiltration basin scheme depending on actual infiltration rates and RE (Fig. 5b) with capital costs of around $700,000-900,000 (Fig. 6b). Moree has both moderate to high yielding bores (15–30 L/s) and generally fresh groundwater so indicative costs could be expected to be at the middle to lower end for a 2 MCM/y infiltration basin scheme around $0.09/m3 with a capital cost of around $1.2 million. Dubbo has moderate yields (~ 15 L/s) and fresh groundwater placing levelised costs around $0.10/m3 in the middle of distributions for a 2.5 MCM/y infiltration basin scheme (Fig. 5b) with a capital cost of about $1.3. million (Fig. 6b). Narromine has very high yields (~ 200 L/s) and fresh to brackish groundwater placing levelised costs at around $0.09/m3 (Fig. 5b) with capital costs of about $2 million (Fig. 6b) which is in the mid-low end of cost distributions for a 5 MCM/y infiltration basin scheme.

Disaggregated costs for infiltration basin systems of 1 MCM/y or less were driven by initial costs of feasibility studies and capital costs of pumps and pipes (Fig. 8). As scheme capacity increased beyond 1 MCM/y, opportunity costs of water and capital costs of pumps and pipes become more dominant. The sensitivity analyses for high volume infiltration basin schemes also reflected that uncertainty of these costs drive variability in levelised costs of recovered water (Fig. 9). For MAR within water market environments, understanding market dynamics for developing operating rules is critical for reducing uncertainty in economic viability assessments of MAR schemes. Understanding market behaviour in response to external factors e.g. water availability, climate patterns, weather forecasts, commodity price fluctuations, and demand patterns would inform operational rules and plans for recharge and recovery timing for maximum cost efficiency and benefit. Within a water market system such as in Australia’s Murray-Darling Basin, there is a question around how water suppliers would participate in the market using different strategies for sourcing, storing and distributing water through MAR. For example, whether it is better to buy permanent water rights and sell temporary rights, rely solely on temporary trades at the market spot price, or use a futures market strategy.

Conclusions

Managed aquifer recharge (MAR) is proven to increase drought resilience however, uncertainty regarding costs has been a barrier to implementation. This study aimed to address current gaps by presenting a set of transferable distributions of financial costs for well injection and infiltration basin schemes covering a range of operating scales and conditions. Three key advancements were made. Firstly, a time-varying financial cost assessment framework based on water balance time series of recharge, storage and recovery was developed. Secondly, the effects of aquifer storage losses on costs of recovered water were evaluated. Thirdly, a probabilistic approach to account for uncertainty in input parameters was applied analysing disaggregated capital and operating expenses and conducting global sensitivity analyses to highlight the most important factors for reducing uncertainty. Six case study examples of town water supplies in Australia’s Murray-Darling Basin were used to demonstrate application of the methods and interpretation of results.

Levelised costs of recovered water (in 2021 AUD) ranged from an average of $0.07-$0.24/m3 for infiltration basin schemes, and $0.32-$1.88/m3 for well injection systems depending on scheme capacities and operating conditions. Levelised costs of recovered water were between 30% and 100% more than costs of recharge for systems of equivalent capacities and operating conditions. Understanding potential storage losses to constrain this uncertainty is therefore critical in early planning and design stages for assessing scheme viability.

Disaggregated capital and operating expenses varied according to scheme scale and operating conditions. As the dominance of different cost components changed, the sensitivities of variable input ranges that determined these cost factors also shifted. This non-stationarity in global sensitivities suggest where non-linearities and tipping points exist and where the value of information in reducing uncertainty and potentially cost is highest and indicates that these can be different depending on site characteristics.

Based on cost distributions, levelised costs of recovered water were estimated for conceptual well injection and infiltration basin schemes across six towns within the Murray-Darling Basin. These schemes represented a range of operating scales and conditions and demonstrated the application of results based on basic information about the groundwater system in each location. Insights were given on the factors to focus on to reduce uncertainty in each case.

Costs and confidence will likely change over time with increased deployment of MAR. One of the key determinants of MAR project success is the presence of similar projects in the same aquifer or region as lessons can be shared with future projects. An assessment of MAR scheme viability for a specific location is not complete until potential economic benefits have been assessed. These were not covered in the current study as the aim was to provide generic, indicative cost distributions and sensitivities for a range of scales and conditions. Quantification of benefits relies on project-specific analyses of economic returns and externalities. This study focussed on developing a time-varying, probabilistic approach to estimating project costs and potential benefits for assessment of specific sites could be treated in a similar way.