The general approach of the evaluation process is shown in Fig. 1. The applied approach is almost similar to the DAI approach (the decision analysis approach; Marttunen and Hämäläinen, 1995; Karjalainen et al., 2013) that provides strong interactivity with the stakeholders through facilitated meetings and personal computer-aided interviews of stakeholders.

In a very first step, the problem has to be defined clearly. The second step involves a complete identification of the stakeholders who are involved in the problem. The snowball approach is used for that purpose. A crucial step (step 3) is the identification of all stakeholder objectives. Objectives are identified with stakeholders, representing their values and interests. They are ordered in a hierarchy tree with different levels of detail, considering the different scope, inconsistency and explicitness (Keeney and Raiffa, 1976). We arranged lower-level objectives into ecological, economic and social objectives (Bogardi et al., 1982). Afterwards (step 4), potential measures are defined and combined into alternatives (step 5).

The description of impacts is carried out through a consequence table (step 6) showing the consequence that a given alternative will have for a given objective. A consequence matrix is built up through various sub-steps (Keeney and Raiffa, 1976). First, the physical impact of an alternative to the hydrological system is analysed and quantified using measurable units. Then, its socioeconomic and economic impacts are quantified. To include the possible impacts of climate change on preferences, two impact matrices are created, where one depicts the status quo of available water resources and the other one includes changes in future water resource availability. In step 7, interviews are conducted to elicit stakeholders' preferences which are described in detail below. The results from steps 6 and 7 are combined with Eq. (1) to rank the alternatives (step 8) among the stakeholders. Afterwards, they are presented in a workshop (step 9), where stakeholders are also asked to rank the measures independently of any prior objectives. The results (8 and 9) are assessed and interpreted in step 10, and differences in rankings of future scenarios are compared (step 11). Finally, a sensitivity analysis is conducted to test the robustness of the obtained results towards uncertainties in the attribute levels of the alternatives (consequence matrix) and in the attribute valuation of the stakeholders.

Stakeholders are asked in the interviews to evaluate different levels of attributes for every objective on a standardised scale. The attribute level range must cover all the attribute levels of the alternatives, limited by the highest and lowest levels of all of them (Keeney and Raiffa, 1976).

The single attribute value function vi (ai) is assessed with the direct rating method, which is a numerical estimation method (von Winterfeldt and Edwards, 1986). The evaluation is done with Microsoft Excel^© on an interactive evaluation platform, so that the decider is able to see a visualisation of answers (Fig. 2). The range of attribute levels is limited by the highest and lowest levels of the attribute in all alternatives of both scenarios. The respondent is asked to estimate the strength of preferences of every attribute level on a numeric scale between 1 and 0, with 1 as the most preferred level and 0 as the least preferred level. The remaining levels have to be rated between 1 and 0, considering the space between two attribute levels as the strength of preference between them. The value function is a transformation of the attribute levels of the objective on a comparable scale between 1 and 0.

In addition to the value functions, the weights of the different objectives have to be elicited for every representative. To avoid the risk of stakeholder behavioural biases, it is of high importance to understand the possible influences and impacts on the results of different methods (Roberts and Goodwin, 2002). Different weighting methods lead to different results, although they are based on the same theoretical assumptions (Pöyhönen and Hämäläinen, 2001). Therefore, it is important to use a method which on the one hand, is easily understandable for the interviewed persons and on the other hand, statistically applicable and traceable. In this study, we have applied the SWING method (von Winterfeldt and Edwards, 1986). In this approach, the attribute ranges are explicitly incorporated into the elicitation questions, being proven to be a successful method in convergence tests (Pöyhönen and Hämäläinen, 2001).

The calculation of the results was realised with Web-HIPRE^© version 1.22, a web version of the HIPRE 3+ software for decision analytic problem structuring, multi-criteria evaluation and prioritisation (Mustajoki and Hämäläinen, 2000). The further processing of the data was realised with Microsoft Excel^©.

The preferences have been ranked and the different fundamental actions have been analysed. Additionally, a sensitivity analysis has been conducted to deal with uncertainty in the valuations and also with uncertainty in the attribute levels. The conflict potential has been interpreted and the ranking discrepancy expressed by the mean standard deviation of the average ranking of all stakeholders. The influence of the dynamic variables has been analysed by comparing the two scenarios, observing the changes in the rankings. Finally, the aptitude of the MAVT method has been evaluated, comparing the results of the model with the results of the holistic alternative ranking, which have been evaluated by some catch questions in the interviews.

The Mancha Oriental (MO) aquifer is located in south-eastern Spain in the eastern part of the Mancha plain, mainly in the provinces of Albacete and Cuenca, in the Castilla–La Mancha region (central Spain), with small areas in the Valencian Community and Murcia (see Fig. 3). It is part of the Júcar River basin district as one of its 52 groundwater bodies, with a total extension of approximately 8000 km2 (CHJ, 2009), and is consequently the biggest aquifer of this system and one of the largest carbonate aquifer systems in Spain. The major part of the system belongs to the catchment of the Júcar River, which is strongly connected to the aquifer.

The region is characterised by a semiarid continental climate, with an effective average rainfall of about 350 mm year-1 varying between 150 mm in dry years and 750 mm in humid years (López-Fuster, 1999). The average net precipitation of the system from 1945 to 1975 was about 338 Mm3 year-1 (CHJ, 2009). However, in the last decade, it decreased to 292 Mm3 (86 %).

With its high agricultural activity and semiarid climate, about 90 % of water in the eastern Mancha region is demanded by agriculture (CHJ, 2009), of which the major part is used for irrigation and a small part for livestock breeding. Urban water demand accounts only for about 10 % of the total demand and plays a secondary role. The intensive expansion of irrigation since the early 1970s and the cultivation of high water consumption crops led to a significant increase in water demand and groundwater pumping. In the last years, the gross extractions have stabilised between 300 and 450 Mm3. However, the renewable resources are assessed between 280 and 330 Mm3 year-1 (CHJ, 2009) and, consequently, the aquifer balance is still negative.

The intensive groundwater pumping has led to a significant drop in the groundwater level, and the piezometric level is locally about 35 m or more under the level of the natural regime (Sanz et al., 2009). As a consequence, the springs, whether they are permanent, temporal or ephemeral, are suffering a notable change in the discharge, and 13 % of them have dried up (López-Sanz, 2010). Also, the discharge of the Júcar River is heavily influenced, as stream aquifer interaction in the upper reach has changed from a gaining river to a losing or even non-connected river.

The Júcar River is the main surface watercourse in the Mancha region. In the upstream limit of the aquifer region, the Alarcón reservoir (with a storage capacity of 1112 Mm3) serves the main streamflow regulator, providing water supply for urban demand in the region and partially for agriculture. El Molinar reservoir represents the downstream geographic limit of the Júcar River reach within the MO region.

The MO aquifer system has suffered since the early 1970s from a continuous drop in groundwater levels due to intense groundwater pumping. This pumping has been provoked by an important transformation from dry land to irrigated land and consequently an increasing demand for irrigation water for agriculture. Promoted by economic incentives, the development of an intensive agriculture has led to a total irrigated area of about 100 000 ha, whereof the predominant part is supplied by groundwater. Agriculture is the most important economic factor in the MO region. However, the overexploitation of the aquifer produces important ecological impacts, ranging from the drying of springs and wetlands to the disappearance and regime alteration of the rivers and the pollution of groundwater from nitrogen leaching due to the intense fertiliser use. Nitrate concentrations in groundwater of up to 125 mg L-1 have been measured at certain locations (Moratalla et al., 2009), and the aquifer has been designated a nitrate vulnerable zone by the Castilla–La Mancha regional government.

Yet, water use from the aquifer is limited by the River Basin Authority (Confederación Hídrográfica del Júcar) and the Jucar River Basin Management Plan. The regulation and control of water abstractions is managed by the Junta de Regantes de la Mancha Oriental, with about 800 members and an irrigated area of 90 000 ha. Groundwater abstractions and water use are controlled by remote sensing and personal inspections (Castaño et al., 2010). The total irrigated area fluctuates between 100 000 and 110 000 ha.

Various measures have already been proposed to halt overexploitation and achieve the goal of sustainable aquifer management, by controlling the quantitative overexploitation of the aquifer through collective actions (López-Gunn, 2003) and achieving a good chemical status of the aquifer through the use of fertiliser standards and fertiliser taxes (Peña-Haro et al., 2010, 2014).

This study intends to identify the points of conflict between stakeholders to create a basis for further planning, and also to sensitise the stakeholders to possible impacts caused by different management. A general approach to find well-accepted measures should be found by testing the MAVT method, incorporating the stakeholders. And finally, the method's aptitude for conflict resolution in water management has been evaluated. The impact assessment of the measures and the evaluation of alternatives have been realised for the year 2027, which is homogeneous with the provided deadline of fulfilling the goals set in the WFD in the second instance. To evaluate the robustness to external changes, two future scenarios were considered: a static and a dynamic one, including possible changes in the available water resources in future. The application of the MAVT in the case study of the MO aquifer should give an idea of the possible impacts of various alternatives, its fulfilment of the objectives for the different stakeholders, and the conflict potential of the alternatives between the stakeholders. Furthermore, the aptitude of the MAVT as a tool for conflict resolution in water management has been tested.

Management alternatives were not defined by experts, but were created from already proposed measures. Given that the overexploitation of the MO aquifer has been a problem for many years, different measures have already been developed by different organisations, varying in the basic approach to solving the aquifer management problem (López-Gunn, 2003; Martín de Santa Olalla Mañas et al., 1999; Peña-Haro et al., 2010, 2014). After all potential measures were identified, they were grouped into fundamental actions (FA).

FA1: Control/restriction of groundwater use a.

Reduction in irrigated agricultural area, change to dry farming

Although the direct ranking of a high number of complex alternatives with different outcomes would be a very difficult, unrealistic and probably meaningless task for the stakeholders, the MAVT will allow for the indirectly ranking of the alternatives for each stakeholder by obtaining their values and preferences in terms of objectives, and by assessing the performance of the alternatives on those objectives.

To assess how well each alternative meets the objectives, the impact assessment was forecasted for a certain time horizon. As the principal objectives are based on the guidelines of the WFD, the deadlines for its implementation were chosen. The date of evaluation was set in 2027, since this is one of the stages to be addressed. The base situation described the average situation between 2000 and 2008 to avoid possible anomalies of one specific year. Some of the proposed measures have already been implemented or are in the process of realisation.

To analyse the influence of climate change and other changes in the hydrological system, two scenarios with different assessment (consequence) matrices were introduced:

scenario 1: static scenario; and

Scenario 2 incorporated changes in urban water demand and climate. The possible impact of climate change on water resources (Chirivella-Osma, 2010) and the possible increase in urban water supply, handled as a restriction, were introduced into the calculation of the impact assessment. By comparing the results, the vulnerability of the acceptance of different measures and changes in preferences could be evaluated as well as the importance of the outcome by considering such variables.

For the assessment of the physical impact assessment, we used an existing groundwater model (Sahuquillo et al., 2008; Sanz et al., 2011), groundwater budgets from the annual statistics of the Jucar River Basin Agency (CHJ) and piezometric level observations (Lopez Sanz, 2010). Fitted regression equations (Sanz et al., 2009) and nitrate measurements (Moratalla et al., 2009) were used to estimate the influence of management options on nitrate concentration. For the assessment of possible climate change impacts on the water resources, we used estimations from Chirivella-Osma (2010) and Chirivella-Osma et al. (2015).

For the assessment of economic and socioeconomic impact, we used input–output matrices from the statistics institute of Castilla–La Mancha, and fitted regressions from previous hydroeconomic models of the groundwater system (Peña-Haro et al., 2010, 2014) to estimate the effects of different policies such as fertiliser taxation and water prices, and other relevant data from previous studies in the same case study (López Fuster, 1999; Martín de Santa Olalla Mañas, 1999)

The value function and the weights of the alternatives were elicited through interviews with representatives of every stakeholder group. The valuation of the attributes was realised for the range between the lowest and highest levels of the attribute for all alternatives in both scenarios. This range differed slightly between the two scenarios, and it would have required separate valuation interviews for every scenario. However, it was confusing for stakeholders to realise an interview twice with almost identical ranges of attributes. Because of this, a little error has been accepted, as it was not of high significance to the outcomes. Also, the use of qualitative scales of attribute levels reduced this problem.

Personal computer-aided interviews (Sect. 2.2) were conducted with representatives of the stakeholder groups. Either the interview was with the official spokesman of a stakeholder group or with at least two stakeholders per stakeholder group, to maintain the objective character. Within one stakeholder group, the values of the representatives were averaged. If a stakeholder was not available for a personal interview, it was conducted by electronic interview, although personal interviews were preferred to avoid misinterpretations or errors in the valuation. In addition to the elicitation of the value function and weights, some general questions about the aquifer management were asked. Through the interview questions, the interviewee's holistic preferences were evaluated. Afterwards, a workshop in Albacete (Spain) was held, where the results were presented and stakeholders were asked to evaluate their prior valuation of preferences.

Based on the total value of the alternatives for the stakeholder groups, we proceeded to rank the different alternatives between 1 (best) and 36 (worst). The different alternatives were ranked quite similarly (see Fig. 5) across the stakeholder groups, although some clear discrepancies can also be detected. According to the results of the MAVT method, stakeholders tend to prefer a mix of different fundamental actions for problem resolution. The potential of combined measures might be preferred for economic reasons, and might cause smaller impacts distributed in various sectors instead of one rigorous measure. For an alternative to be considered good, ecological, economic and also social interests have to be considered. Of course, among the different stakeholder groups the focus differs, but the best-ranked alternatives are still the same. Figure 6 shows the evaluation of the alternatives and the range of variation between stakeholders. The best overall alternatives were chosen in two steps. Firstly, all dominated alternatives were eliminated. Secondly, the sum of the stakeholder rankings was calculated for every alternative. The lower the sum of ranking, the better the overall evaluation.

By analysing the composition of the total preference value, it could be seen that economic, ecological and social interests had to be fulfilled by an alternative to achieve a good ranking.

The representatives of all stakeholder groups agreed that a restriction on water access is necessary to obtain sustainable aquifer management. The impediment of a further increase in the water demand by legal instruments was fundamental for a high ranking of an alternative. It can be reasoned that any other measures will be declined if there is no restriction on water access. Such measures are the reduction in the irrigated agricultural area, the reduction in the water allotment in drought periods, restrictions for high water needing crops, the improvement in extraction controls and the improvement in the irrigation efficiency. Alternatives 1–9 do not include such measures to limit the water access, and received consequently a low ranking. Also, the compliance of the ecological objectives relied mainly on this restriction.

Groundwater substitution by surface water was evaluated as an appropriate measure, although the preferred implementation factor differed among the stakeholders. Especially the high costs and the long realisation time of a full implementation impacted negatively on the results of environmental organisations and rural development. To agricultural representatives, it represented a necessary measure to guarantee water supply and economic activity.

Regarding the economic instruments, fertiliser taxation is considered the most adequate solution, especially because of its positive influence on the nitrate concentration in groundwater. However, the influence of water taxation on the results is quite low, and is considered less adequate than other measures.

In general, it can be observed that the main question considered a good alternative is not only “what is the best measure”, but also “what is the implementation stage of a measure” and “how is the combination with other measures”.

A conflict potential might arise when there is a large variation in the ranking of certain alternatives across all the stakeholder groups. Analysing the conflict potential helps to focus on the proper measures for reaching a consensus solution. The conflict potential expressed by the mean deviation of the evaluation of alternatives by the stakeholders is shown in Fig. 7.

Analysing the fundamental actions, conflict potential could be found in the full implementation of the planned measures. High implementation stages of measures are just accepted by all stakeholder groups if they are combined with other fundamental measures. For instance, the option of full implementation of groundwater substitution is assigned a low ranking by ecologists unless the option also includes water use restrictions to prevent an overall increase in irrigation. On the other hand, agricultural representatives would accept a water use restriction provided that groundwater substitution measures are also implemented to avoid economic drawbacks.

The acceptance of economic measures varies significantly and has high conflict potential. According to the results, water taxes are not decisive for a good alternative ranking. Fertiliser taxes contain high conflict potential, but they are provided in the best-ranked alternatives, due to their regulation capacity of nitrate concentration in groundwater. However, in reality, the acceptance of taxes, especially by agricultural representatives (obtained by direct interviews), might be low, because of the economic losses as a consequence.

Uncertainties in the future by dynamic variables (scenario 2) are another source of conflicts. This might be because of the uncertainty of future development. Static external variables (scenario 1) provide more precise results, but they might not represent the future reality.

The scenario comparison served to show up conflict points coming up because of uncertainties in variables in the future. In scenario 2, due to the inclusion of a greater urban water demand and climate change, the achievement of the objective of sustainable aquifer management was more difficult than in scenario 1. Figure 8 shows changes in ranking for the three best evaluated scenarios 1. The main questions were whether there were significant changes in the preferences of outcomes and how the alternatives change between the two scenarios.

Looking at the best evaluated alternatives in both scenarios, it could be seen that the most preferred alternatives are not entirely the same, but that the focal points of well-evaluated alternatives were rather similar. The following similarities could be detected.

Alternatives with the focus on just one measure are evaluated worse than those with several different measures of the fundamental actions.

The sensitivity analysis was applied to test the robustness of the results towards uncertainties in the inputs. The results of the MAVT depend on two principle input factors:

the attribute levels of the alternatives (consequence matrix); and

The uncertainty in the valuation of the attributes is given by different value functions and weights assigned by different representatives within one stakeholder group, which results in different preference values and alternative rankings. A similar valuation within a stakeholder group creates homogeneity and uncertainty becomes low. The sensitivity analysis for uncertainty in valuations was done for stakeholder groups with more than one representative. Uncertainties within the stakeholder groups were significant for environmental organisations (Fig. 9); meanwhile, municipalities (Fig. 10) and rural development had quite similar evaluations. This was interesting, as one would expect a lower discrepancy between ecologists than between other stakeholder groups.

Finding the optimal solution with the multi-attribute value theory in water resource problems seems to be difficult, requiring that complex structures be reduced to one value, expressing the acceptance or negotiation of an alternative. More complex methods like for instance ELECTRE (San Cristóbal, 2012) might be more appropriate for finding the best solution, but in the application of these there is a risk of non-transparency and a lack of understanding about the method among the participating stakeholders.

Nevertheless, the applied decision analysis framework based on the MAVT is a useful method for finding possible conflict points between the multiple stakeholders, helping also to identify possible consensus solutions. Furthermore, it is possible to define basic criteria for alternative planning, to guarantee a high acceptance of measures and to avoid future conflicts.

The acceptance of the method is quite high (see Fig. 11) because of its simplicity (see also Marttunen et al., 2013). The involvement of stakeholders at the beginning of the planning process, especially in the setting of the objectives, is considered important for obtaining high acceptance. It is important to weigh the necessary complexity of the model with the comprehensibility. Stakeholders have to be chosen carefully in terms of their knowledge of the issue, and they should have a good overview of the problem. If not, the valuations will be made without basing them on facts. If stakeholders understand the method, the acceptance of the results will be higher, and their contribution to conflict resolution too, since the results become more acceptable.

The holistically assessed preferences given by some stakeholders are mostly coherent with the results of the MAVT method. In comparison to a holistic ranking of options, the MAVT method has the advantage of creating a more detailed evaluation framework, which enables a more informative analysis to be undertaken. This includes a more detailed analysis of conflict potential and the ability to undertake uncertainty and sensitivity analysis.

The approach has been applied to the analysis of sustainable management of the MO aquifer, allowing one to elicit stakeholder groups' values and to evaluate groundwater management options. Stakeholders clearly preferred combined measures for economic reasons, and cause smaller impacts, because impacts are distributed in various sectors. For an alternative to be considered good in MO, ecologic, economic and also social interests have to be considered. Logically, among the different stakeholder groups, the focus differs, but the best-ranked alternatives are still the same due to their basic structure. The representatives of all stakeholder groups agree that a restriction on water access is necessary to obtain sustainable aquifer management. Other measures just have efficiency if there is no additional increase in the water use. Also, the compliance of the ecological objectives relies mainly on this restriction. Groundwater substitution by surface water is evaluated as an appropriate measure, although the preferred implementation factor can differ among the stakeholders. Especially the high costs and the long realisation time of a full implementation affect negatively the results of environmental organisations and rural development. To agricultural representatives, groundwater substitution represents a necessary measure to guarantee water supply and economic activity. According to economic instruments, fertiliser taxation is considered the most adequate solution to achieve a good ecological status, especially because of its ability to limit the concentration of nitrate in the groundwater. The influence of water taxation on the results however is quite low and is considered less adequate than other measures.

Although the number of alternatives seemed to be unmanageable for the stakeholders at the beginning, the MAVT approach helped them to get more sensitised to the complexity of groundwater management in this aquifer and to elicit stakeholders' preferences and potential conflict points. In summary, the applied MAVT method is a useful support tool for planning processes; not for finding the best solution, but for avoiding future conflicts and finding potential consensus solutions by a detailed analysis of the measure rank based on the stakeholder preferences and values. Also, it serves to sensitise stakeholders to competing interests in environmental problems.