Marginal Abatement Cost Curves for Water Scarcity Mitigation under Uncertainty

Water scarcity and drought are considered to be among the most critical global risks to society (WEF 2019) with short and long-term effects on citizens, ecosystem services, biodiversity and the economic sectors that depend on a reliable water availability. In Sweden, which is considered to have good access to natural water resources, only 1 % of the renewable water is extracted for use in households, agriculture and industry (Eurostat 2017). The industry is the main water user in Sweden, accounting for about 61% of water use. Around 23% of the water is used by households and 3 % by the agricultural sector. The remaining 13% is used within other user categories, such as construction, retailing, hotel and restaurant, transport, and public administration. Thirty-five percent of total water, and 88% of household water, is provided via the public water supply system for which the municipalities are responsible (Statistics Sweden 2017). However, both water usage and water availability vary largely across the country and do not always overlap geographically. Hence, local water imbalances are not uncommon, and the southern, central and coastal areas along the Baltic Sea in particular may experience water shortages during the summer (Statistics Sweden 2017). Nevertheless, the low precipitation and high summer temperatures that hit Sweden for the third consecutive year in 2018 acted as a wake-up call. The vastly reduced access to water had major consequences for the Swedish society, with failing crop yields and comprehensive emergency measures to maintain the water supply to households, industry, and critical societal functions. In the summer of 2018, around 30% of the municipalities prohibited urban irrigation and called for careful use of drinking water. Farmers experienced their worst harvest since the 1950s, and the lack of grazing and feed led to emergency slaughter of livestock and six-month queues to the slaughterhouses (Sjökvist et al. 2019).

Planning to avoid and manage water scarcity is an expanding area within integrated water resource management (IWRM), see e.g. Dukhovny et al. (2013). In 2007, the European Commission presented an initial proposal on how to address and mitigate the challenges posed by water scarcity and drought within the European Union (EC 2007). In line with the IWRM literature, the ensuing European Commission reports emphasized that the most appropriate way to address water scarcity and drought is to use an integrated approach based on a combination of demand and supply measures. However, additional water supply infrastructures, such as water transfer and desalination plants, are to be carried out only after all demand-side measures have been considered. Demand and supply management measures can be broadly categorized as: 1) measures that increase the water supply by converting natural resources into available and reliable water; 2) measures that reduce demand by encouraging efficient and controlled water use; and 3) measures that increase productivity, e.g. via improved practices and water reuse (Addams et al. 2009; Pinto and Marques 2017). Measures to prevent the water-related impacts of climate change are expected to be extensive and costly (Westling et al. 2019). Furthermore, the quality of data relating to such measures can be highly variable and uncertain (Addams et al. 2009). Hence, it is necessary that policymakers and practitioners are provided with relevant decision support that would allow them to compare the range of possible mitigation measures and make economically rational investment decisions.

In climate policy-making, marginal abatement cost curves (MACC) are frequently used to provide guidance on greenhouse gas (GHG) mitigation measures. MACCs have gained in popularity since they succeed in illustrating and comparing a range of complex measures from various sectors in an easily understandable format. The curves specify the marginal cost of abatement for each analyzed measure while enabling assessments to be made of the total abatement costs through the integral of the curve (Kesicki 2012). MACCs have been used for GHG assessments in a variety of sectors, e.g. the cement industry (Worrell et al. 2000), the transport sector (Peng et al. 2018), and forestry and agriculture (Eory et al. 2018a; Moran et al. 2009). MACCs have also been used in other policy areas but only a few studies have applied them to water challenges (Addams et al. 2009; Chukalla et al. 2017). Most MACC studies do not take the influence of uncertainty into account, but often highlight the difficulty of including uncertainties in the cost curve format (Kesicki 2012). The limited representation of uncertainty has led to concerns that the cost curves give an impression of robustness without showing the underlying uncertainties (Eory et al. 2018a). By ignoring the influence of uncertainties, the usefulness of the MACCs is reduced. The MACC studies that have quantified uncertainties have most often been limited to analyzing one or two specific parameters, such as the discount rate, through scenario or sensitivity analysis (Peng et al. 2018). Recently, Peng et al. (2018) and Eory et al. (2018b) presented statistical uncertainty analyses based on Monte Carlo simulation for MACCs in the transport and agricultural sectors. However, there is no commonly accepted state-of-the-art approach to performing uncertainty analysis of water scarcity mitigation measures in a MACC context. Addams et al. (2009) acknowledged this and welcomed contributions that could improve their MACC framework by addressing uncertainties associated with water challenges. This paper attempts to meet that need.

The primary aims of this paper are therefore to: 1) provide systematic handling of uncertainties in MACCs of water scarcity mitigation efforts; and 2) exemplify the MACC approach by evaluating and comparing municipal, industrial, agricultural and household mitigation efforts on the island of Gotland, one of the most water-stressed parts of Sweden. The MACC is based on financial analysis that demonstrates the financial trade-offs implied by different mitigation measures, thus providing critical economic insights for all parties concerned and an important basis for reaching decisions. Making this approach available in the literature is a valuable contribution to the economics of water scarcity mitigation through which the uncertainties behind narrowing the water deficit become apparent.

Gotland is Sweden’s largest island, located in the Baltic Sea about 100 km from the mainland. The annual precipitation on the island is approximately 550 mm. However, thin soil layers, extensive drainage, and a lack of coherent reservoirs in the sedimentary limestone bedrock lead to high precipitation run-off. The number of people who live on Gotland all year round is about 58,000, just a fraction of the vast number of visitors. In 2016 over 2 million people traveled to Gotland, and the number of guest nights at hotels and other commercial accommodation facilities exceeded 1 million (Region Gotland 2017). The peak season for tourism is during the summer, resulting in a large seasonal variation in water demand and with the highest demand occurring when water supplies are at their lowest. Compared to other parts of Sweden, the water supply system of Gotland is built on a large proportion of private solutions, with only 67% of the households connected to the public water supply system. Around 65% of the public water supply system relies on groundwater. The remaining 35% of the water comes from lakes and one desalination plant (a further desalination plant is currently under construction).

Gotland is a single municipality and due to its geographical location, there is no possibility of augmenting the water supply from neighboring municipalities. For short-term water supply disruptions, an emergency plan will assure water distribution to supply important societal activities. For longer disruptions however, e.g. in connection with extended periods of low water availability, there is no plan for how to continue to provide water to the community. In addition to an already constrained water supply situation, it is estimated that the total water demand on the island will increase by more than 40% through to 2045 (Table 1). This will require water that is currently not available on the island (County Administrative Board of Gotland 2018).

Figure 1 demonstrates the MACC for the analyzed agricultural, household, industrial and municipal measures. The MACC is based on the mean values of cost per cubic meter and water availability for each measure. The results presented in Fig. 1 demonstrate that on Gotland, the household measures are the least cost-effective when it comes to providing additional water availability. Agricultural measures, such as increased sub-irrigation and irrigation dams, together with traditional municipal measures, such as increased groundwater extraction, are among the most cost-effective. One measure, i.e. retrofitting showerheads and faucets at hotels, is financially beneficial due to the hot water savings, and subsequent energy savings, that ensue from this measure (Fig. 2).

The largest contributor to water availability potential is the irrigation dams measure, which has the potential to meet the entire forecast irrigation requirements. Of the municipal measures, increased groundwater extraction and desalination offer the greatest water availability potential. However, on Gotland desalination is almost 25 times more costly than increased groundwater extraction, due in part to the long pipelines needed to reach demand centers.

Figure 2 shows the range of uncertainties associated with the cost per unit calculation of each measure. This is illustrated by means of error bars based on the 5th, 50th and 95th percentiles. The water availability uncertainties are included in the cost per unit calculations shown on the y-axis, whereas the x-axis is limited to showing the annual mean value only. See Online Resource 3 for water availability uncertainties.

The CVs for the analyzed measures are between 0.10 and 0.74 (Online Resource 4). Of the municipal measures, metered leak detection with a CV of 0.49 was associated with the largest uncertainties. In the household, industry and agriculture sectors, small-scale desalination (0.74), reuse of mining drainage water (0.32), and small-scale sub-irrigation (0.34) respectively were the measures associated with the greatest uncertainties.

Figure 3 shows the degree to which the input variables of small-scale sub-irrigation co-vary with the marginal cost, expressed using Spearman rank correlation coefficients between −1 and 1. Input variables related to estimates of the sub-irrigated area contributed most to the outcome uncertainty. Correlation coefficients for all measures are provided in Online Resource 5.

The impact of discount rates was modelled using scenario analysis. Based on mean values, the ranking order does not change substantially between the three analyzed discount rates, and not at all for the six highest-ranked measures, indicating a robust cost curve order (Online Resource 6).

A MACC is unique to every region or country for which it is performed. Both the prioritized set of measures and the final ranking order depend on site-specific conditions. This paper has established a projected MACC of water scarcity mitigation measures for the island of Gotland. However, some limitations should be noted. Firstly, experts were used to estimate the uncertain quantities of input variables. While expert elicitation can be performed in a formal setting, e.g. using the Sheffield Elicitation Framework (Oakley and O'Hagan 2016), the elicitation performed in this study was more informal and was based on group and individual discussions. By using an informal elicitation setting, the risk of biased assessments increases as e.g. more outspoken experts may be given more room in the discussions or experts speaking outside their area of expertise may remain unchallenged. Secondly, local expertise was prioritized as both costs and potentials depend on site-specific conditions. However, by using local expertise the number of previously implemented measures on which estimates can be based becomes limited. This may further impact on the estimated quantities. Thirdly, the scope of this paper was limited to estimations of financial costs and did not allow for inclusion of any ancillary effects, such as environmental improvements or other externalities. Some of these additional effects may be substantial and could change the marginal costs and ranking order if included. It is important to bear these limitations in mind when interpreting the results provided.

As previously mentioned, uncertainties are commonly not considered in MACCs and there is no commonly applied approach for uncertainty and sensitivity analysis. The probabilistic approach proposed here enables a thorough uncertainty analysis where the variation in estimated water availability and cost can be assessed and thus the robustness of the measures evaluated. Furthermore, a comparison of the relative uncertainty of the measures is made possible, and detailed sensitivity analyses could highlight the input variables that contribute most to outcome uncertainties. Such information provides a more informed picture of the financial risks and water potentials associated with each measure and could thus affect the perception of which measure is most beneficial. It also facilitates decisions regarding which variables ought to be investigated further in order to reduce uncertainties.

The results show that the agricultural measure of increased irrigation dams may be a cost-effective path to improved water availability through overall enhanced storage capacity. The use of on-farm water storage enables farmers to extract water from watercourses at times when the flow is high and store it for use when the flow is low. The large potential of the agricultural sector to contribute to increased water availability through water savings is well known (Dworak et al. 2007; EC 2018). On-farm storage, however, does not typically save water but instead reduces the pressure on water bodies at critical times while securing an irrigation supply and creating recreational and aesthetic benefits (Berman et al. 2012). On an island such as Gotland, with high precipitation runoff, the measure will, in addition to the above-mentioned benefits, also help keep the water on the island, improving overall water availability and the potential for multipurpose water use.

The most cost-effective measure analyzed for Gotland was retrofitting showerheads and faucets in the hotel industry. While the theoretical annual volume conserved was rather low due to the limited number of hotels on the island, the MACC confirmed that there may be significant cost savings to be made when focusing on measures designed to save hot water. This is in line with previous findings, as reduced hot water usage leads to reduced energy consumption (Fidar et al. 2016). In general, many water-saving measures and technical devices available to the tourism sector, such as efficient faucets and dishwashers, are similar to those available to households. The measures are often easy to implement and have short payback periods (Dworak et al. 2007). The household measures analyzed in this study were, however, the least cost-effective measures. The high unit costs were due in part to treatment technologies needed to meet the quality requirements. While other studies have found rainwater harvesting economically beneficial for non-potable outdoor use, e.g. Dallman et al. (2016), the potable use requirement in this study made it uneconomical. In addition, it was assumed that all household measures were implemented in each household even though for some of the measures costs could be lowered by households sharing the systems.

Of the municipal supply measures, conventional groundwater and surface water extraction were shown to be more cost-effective than non-conventional desalination. The high energy requirement of desalination is highlighted in the literature as a major limitation of this measure and a reason behind the often high unit costs (Pinto and Marques 2017). On Gotland, the unit cost of desalination was further affected by the long pipelines needed to reach demand centers. The high cost associated with long-distance piping has previously been shown to alter the order of cost-effectiveness of alternative water supply measures (Lamei et al. 2008). The unit cost of desalination on Gotland was about 25 times higher than that of increased groundwater extraction.

The second most cost-effective municipal measure was improved wastewater treatment for irrigation. As agricultural irrigation has been estimated to have high potential to increase uptake of water reuse, encouragement of this measure was one of the main objectives in the recent European Commission proposal on minimum requirements for water reuse (EC 2018). In addition to the low unit costs on Gotland, this measure may also provide several co-benefits that are not considered in this study, such as improved water quality in nearby watercourses, increased groundwater recharge, and reduced eutrophication of the Baltic Sea.

Even though some measures are highly cost-effective, existing barriers could hinder implementation and hence the possibility of improving the situation. These barriers may be financial, e.g. high initial capital costs, legal, e.g. environmental permits, or social, e.g. lack of awareness of measure efficiency (Addams et al. 2009). To overcome these barriers, some form of policy intervention is usually needed. The MACC provided does not include information on policy instruments to enhance measure implementation. MACCs can, however, be used to guide policy development by helping policymakers identify which measures need an extra push (Vogt-Schilb et al. 2014). To further facilitate fact-based guidance in policy design, more detailed information on drivers and barriers will be gathered and included in future research.

When comparing investments of unequal useful lives, either the equivalent annual cost approach or the replacement chain method is normally used (Brigham and Ehrhardt 2008). The equivalent annual cost approach involves estimating the (net) present value for each measure over its lifetime and dividing it by the lifetime annuity discount factor. The replacement chain method involves repeating the measure investment over the least common useful lifetime. Most MACC studies use the former approach, i.e. life-cycle based costs, to compare individual measures, see Ibrahim and Kennedy (2016). In this paper, the replacement chain method is applied to calculate present values for each measure over a common time horizon. The present values are then calculated as equivalent annual costs distributed evenly over the time horizon. Although this arrangement is not the typical MACC approach, it is used here as it allows fair comparisons to be made between measures over long time horizons (IMWG 2013).

Traditional MACC studies normally assume that analyzed measures will be implemented in sequence, from the least to the highest cost per unit (Dunant et al. 2019). Furthermore, the individual measures are normally totaled to reveal the overall potential of all the measures. This means that account must be taken of possible interactions between measures, e.g. when the implementation of one measure changes the water availability potential of another measure. Such interactions can and should be considered when calculating a cumulative potential, by e.g. adjusting the potential of the interacting measures (Addams et al. 2009). However, as actual implementation does not necessarily follow the order of cost-effectiveness, and as the aim in this paper was to visualize the water availability potential of each measure and not the cumulative potential, the MACC was constructed without taking possible interactions into account.

Finally, it should be mentioned that the process of deriving a MACC is as valuable as the final result (Eory et al. 2018a). The process necessitates the involvement of many stakeholders and encourages communication between water providers, farmers, industry and public authorities. It thus improves overall awareness of inter-sectorial challenges and possible ways to address and overcome them. The presented approach provides a quantitative visualization of the range of possible measures and a common framework for the discussions surrounding water scarcity mitigation. The MACC approach presented here will be further developed to include externalities, seasonal and spatial variability, interactions between measures, barriers to implementation, and feasible policy instruments.

The main conclusions of this paper are: