SEARCH

SEARCH BY CITATION

Keywords:

  • Bayesian belief networks;
  • Ambiguity;
  • Group model building;
  • Adaptive groundwater management;
  • Guadiana

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

In integrated groundwater management, different knowledge frames and uncertainties need to be communicated and handled explicitly. This is necessary in order to select efficient adaptive groundwater management strategies. In this connection, Bayesian belief networks allow for integration of knowledge, for engaging stakeholders and for dealing with multiple knowledge frame uncertainties. This is illustrated for the case of the Upper Guadiana Basin, Spain, where Bayesian belief networks with stakeholder involvement were used for dealing with the ambiguities related to sustainable groundwater exploitation. Integr Environ Assess Manag 2012; 8: 430–444. © SETAC


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

Uncertainties in groundwater management

Groundwater management should be based on scientific knowledge and sustainability principles (CCA 2009) and should follow the principles of environmental ethics (Delli Priscoli and Llamas 2001). Yet, groundwater management decisions are made in situations involving multiple criteria (Xu and Tung 2008) and often without explicit consideration of uncertainty in data and models. This makes the decision process less transparent and credible for stakeholders, because propagation of uncertainties has not been incorporated into solutions to problems. Furthermore, due to lack of time and a paucity of face-to-face meeting opportunities, water managers are often not able to consider the many uncertainties and how best to cope with them in detail.

The European Union (EU) NeWater research project defines adaptive water management as “a systematic process for improving management policies and practices by learning from the outcomes of implemented management strategies” (Pahl-Wostl et al. 2008). A key element in adaptive management is implementation of more robust and flexible practices to cope with the uncertainties and inevitable surprises. Furthermore, the focus is on social learning, which refers to the capacity of the stakeholders to deal, in a constructive and meaningful way, with conflicting interests and points of view, and to collectively manage groundwater resources in a sustainable way (Pahl-Wostl et al. 2008).

The NeWater synthesis product 2 “Uncertainty Guidelines” defines 3 types of uncertainties and proposes different strategies for dealing with those uncertainties (Brugnach et al. 2008, 2009; Table 1):

  • 1.
    Accept not knowing (unpredictability)
  • 2.
    Work on improving knowledge (incomplete knowledge)
  • 3.
    Learning to deal with differences (ambiguity)
Table 1. Definitions of 3 types of uncertainty and strategies for dealing with uncertainties (Brugnach et al. 2009)
Type of uncertaintyDefinition of uncertaintyStrategy for dealing with
UnpredictabilityRefers to the inherently unpredictable aspects of a system that are due to variability or complex system behavior. Expected not to change in the foreseeable future.Accept not knowing
Incomplete knowledgeSituations where our knowledge is incomplete due to lack of reliable information or data understanding or ignorance. Uncertainty can be reduced, given sufficient time and resources.Work on improving knowledge
Multiple knowledge framesRefers to different, sometimes conflicting, views about the system regarding boundaries, focus, or urgency of matters. Also termed “ambiguity.” Originating from differences in backgrounds, disciplines, value systems, and positions, among other factors.Learning to deal with differences

The presence of ambiguity can prevent a shared understanding of the problem among stakeholders due to differences in problem framing (Brugnach 2011) and thus causes wrong or unfortunate behavior, such as ignorance, institutional inertia (negligence), arrogance (or professional bias), and vested interests, which may result in conflicts and corrupt practices (Gray 2004; Llamas et al. 2007), that obstructs proper identification of the problem and articulation of its solutions (Brugnach 2011).

Therefore, integrated groundwater management requires tools that can support social learning and can distinguish between alternative management options. Tools that facilitate integration of technical, socioeconomic, and value-based knowledge could contribute to reducing or highlighting ambiguity.

Adaptive groundwater management

Adaptive water management can be viewed as an extension of the Integrated Water Resources Management (IWRM) concept (in advanced IWRM) (GWP 2004; Mysiak et al. 2009). The application of IWRM involves a 7-step cycle (Jønch-Clausen 2004; van der Keur et al. 2008): 1) establish status, 2) build commitment to reform, 3) analyze gaps, 4) prepare strategy and action plan, 5) build commitment to action, 6) implement frameworks, and 7) monitor and evaluate progress. After evaluation, adaptive water management would also include “adapt strategy and action plan to new emerged learning.” Nevertheless, the central contribution of adaptive management to IWRM is the explicit acknowledgement of uncertainty. The key to adaptive IWRM, therefore, is to establish “learning cycles” for each step of the IWRM cycle to deal with uncertainties and ambiguities that require a special learning effort.

Adaptive water management means a shift in focus and locus from traditional authority-driven management to stakeholder-driven decision making under uncertainty. Instead of trying to isolate risk factors based on objectivity, reductionism, and technical problem solving and viewing uncertainty as something abnormal, adaptive management attempts to renovate the whole system by also acknowledging its uncertainties. Hence, uncertainty awareness through learning about the system becomes essential (issues of climate, land use, groundwater and surface water quantity and quality, and socioeconomics). However, in situations in which uncertainties such as ambiguity are important, the ability to make collective decisions becomes challenging, and it becomes necessary to improve the collective decision-making capacity.

According to business leadership theory, a distinction can be made between technical problems and adaptive challenges. Heifetz et al. (2009) proposed that technical problems are problems that can be diagnosed and solved, generally within a short time frame, by applying established know-how and procedures. Technical problems are amenable to authoritative expertise and management of routine processes.” Contrary to technical problems, adaptive challenges require an understanding of the differences in views and values due to the gap between the values that people stand for (that constitute thriving) and the reality that they face (their current lack of capacity to realize those values in their environment)” (Heifetz et al. 2009). Dealing with adaptive challenges, therefore, is a social and human process in order to properly define and find solutions to problems that in the literature have been referred to as “wicked problems” (Rittel and Weber 1973; Allan 2008). Freshwater allocation problems and surface and groundwater connectivity problems are examples of wicked problems that should be addressed as adaptive challenges, due to the uncertainty and complexity of consequences of management actions (Allan 2008). For such challenges, group model building is important for coping with ambiguity and creating a shared view on the complex issues.

Bayesian belief networks as a tool for group model building

Group model building has been identified as an important tool to cope with ambiguity in collective decision-making processes. In the group model building field, many conceptual tools are available for identifying problems and solutions (Vennix 1996; Sendzimir et al. 2007), such as cognitive mapping (Axelrod 1976; Eden 1989), rich pictures (Checkland 2000), mind maps (Bryson et al. 2004), causal loop diagrams (Vennix 1996; Sterman 2000), role playing games, agent-based models (Janssen 2002; Barreteau 2003), and Bayesian belief networks (BBNs).

BBNs are graphical models of (causal) interactions among a set of variables, where the variables are represented as nodes of a graph and the interactions (direct dependences) are represented as edges (links) between the nodes. BBNs are probabilistic networks used to represent systems and analyze their behavior. BBNs can take into account simultaneously the social, physical, and economic variables of the problem under consideration. Nodes and links in a BBN are referred to as the graphical (qualitative) aspects of the BBN. The probabilistic (numerical) part consists of the conditional probability distributions (represented by probability tables) also referred to as quantitative aspects of the BBN. Working with probabilities per se accounts for the level of uncertainty of any particular result, implying that BBNs are particularly suited to complex problems when dealing with adaptive challenges in problem framing and identifying novel solutions. Below, we will briefly describe how BBNs have been used for IWRM.

In natural resource management, BBNs have been used to analyze effects from land-planning alternatives on wildlife, to predict and aid water-quality management (Borsuk 2004; Shihab 2005), to aid water-resource planning (Domínguez Padilla 2004; Bromley et al. 2005; Martín de Santa Olalla et al. 2005; Molina et al. 2009), or to optimize management of groundwater contamination (Henriksen et al. 2007a). Farmani et al. (2009) showed how to combine BBNs with an evolutionary multiobjective optimization (EMO) to help managers assess multiple evolutionary decision pathways under uncertainty. Matott et al. (2009) described BBNs as probabilistic graphical models that can combine prior distributions of uncertainty with general knowledge and site-specific data (including local knowledge) to yield an updated (posterior) set of distributions. In this process, BBNs simultaneously treat uncertain input and response data, model parameter distributions, and qualitative errors in model structure (Matott et al. 2009).

The first experience of BBN applications with stakeholder involvement in European water problems was gained in connection with the MERIT project (EVK1-CT-2000-00085) funded by the European Commission (Henriksen 2004, 2007a; Bromley et al. 2005; Martín de Santa Olalla et al. 2005). Interviews with water managers described by Henriksen and Barlebo (2008) revealed water managers find that tools such as BBNs are required to evaluate the efficiency of resources used in environmental and groundwater management. The water managers requested an integrated assessment tool that can combine groundwater modeling, monitoring data, and planning restrictions, e.g., wetlands, habitats, and other administrative data for adaptation planning related to Copenhagen, Denmark, wellfields (Henriksen and Barlebo 2008). BBNs could help delineate the complexities and handle some of the uncertainties confronting water managers with respect to the value of clean groundwater. The Copenhagen groundwater managers expressed that they had always focused on adaptation planning, but that the integration (as requested by the Water Framework Directive and its daughter groundwater directive) of the different physical, ecological, economic, and social issues presented a new challenge (Henriksen and Barlebo 2008). Recently, under the Water Framework Directive requirements, BBNs have been tested as a tool for participation modeling (Carmona et al. 2009; Molina et al. 2009; Mysiak et al. 2009; Zorrilla 2009; Zorrilla et al. 2010).

Objectives of this article

The objectives of this article are: 1) to highlight the value of BBNs as a dialogue and communication tool for dealing with ambiguity in groundwater management and 2) to illustrate, using the Upper Guadiana Basin, Spain, as a case study, how BBNs have been a useful tool for group model building and for dealing with trade-offs in relation to a management plan for the basin.

The structure of the article is as follows: In the section DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID, we introduce the knowledge layers and sources of uncertainty in integrated groundwater management. In USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES, we explain the enhancement of BBN as a tool for dealing with adaptive challenges in groundwater management. In THE UPPER GUADIANA RIVER BASIN CASE, we describe the results of exploring ambiguity from the case study. Finally, in the DISCUSSION, we discuss the results from Guadiana and the experience gained from using BBNs for adaptive and integrated groundwater management.

DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

In this section, we briefly introduce a knowledge pyramid (Figure 1) that allow us to categorize different levels of knowledge (data, geological interpretation, models, scenarios, and so forth) and integrate technical, natural, and social system knowledge required for scenario development and decision making under uncertainty, as part of adaptive and integrated groundwater management. Our version of the knowledge pyramid (left side of Figure 1) has been modified (CCA 2009) by adding the level “scenario development,” which covers the horizontal integration and tradeoffs between socioeconomic (e.g., land use), technical (e.g., irrigation), and natural subsystems (e.g., climate change and groundwater). In addition, we have renamed the top level from “decision making” (CCA 2009) to “decision making under uncertainty” in order to stress basic characteristics of adaptive groundwater management. As illustrated by the right side of Figure 1, BBNs are used as a tool for group model building, engagement of stakeholders, integration, and for dealing with ambiguity (van der Keur et al. 2008, 2010; Brugnach et al. 2009; Mysiak et al. 2009).

thumbnail image

Figure 1. Groundwater management scientific knowledge pyramid (modified after CCA 2009) and illustration of use of Bayesian belief networks (BBNs) for scenario development and dealing with ambiguity (multiple knowledge frame uncertainties).

Download figure to PowerPoint

The “numerical groundwater models” layer provides simulation results that must be further considered as part of scenario development. Regardless of whether a traditional (technical problem solving) or an adaptive groundwater management approach (or protocol) is used, model accuracy is affected by different sources of uncertainty, such as lack of knowledge of the conceptual model, lack of knowledge of the continuity of aquifers, or the effectiveness of aquitards as flow barriers (CCA 2009). Another important source of uncertainty is gaps in the quantitative understanding of the hydrogeology. Lack of spatial and temporal hydraulic head, flow, and age data are other sources of uncertainty in calibration and validation of numerical groundwater models. As a result of these sources of uncertainty, groundwater modeling has to be viewed as an ongoing learning process (Henriksen et al. 2003; Beven 2008; CCA 2009), where performance criteria analysis and uncertainty analysis become key issues in the modeling process (Refsgaard et al. 2007). However, knowledge and uncertainty of groundwater models cannot be described as universal, but are always limited to the site-specific situation and its framed terms of reference, as well as to the reliability of the numerical modeling process.

The “scenario development” layer incorporates the broader perspectives of integrated groundwater management, e.g., the need to integrate different systems (climate, land use, and surface water and groundwater), different sectors (having different overall goals and paradigms), and different actors (having different evaluation criteria and mental models). The entire water system is subdivided into 3 subsystems: social, technical (infrastructure), and ecosystem.

When applied to intensive groundwater uses, which we are dealing with in the Upper Guadiana Basin (UGB), Spain, sustainability is a manifold concept” (Llamas et al. 2007). This may be viewed from different angles: hydrological, ecological, economic, social welfare, legal, institutional, political, and intragenerational solidarity.

Llamas et al. (2007) recognize the difficulty in providing a general guide to groundwater sustainability. This means that emphasis on one or the other of the above described point of view is likely to depend on economic, social, cultural and political constraints” (Llamas et al. 2007). Adaptive groundwater management is based on a high degree of stakeholder involvement, because sustainable management cannot be achieved solely by means of top-down “command and control” measures (driven by the locus of authorities), but also need to be nurtured by bottom-up participatory processes (driven by the locus of stakeholders). The above justifies the necessity of adding the layer “scenario development” in order to evaluate the social acceptability and economic appraisal of various water management options (or actions). In this layer, important uncertainties in groundwater management, especially ambiguity, will take on form.

USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

Many tools can be used to integrate knowledge from the different layers and to classify and explicitly assess various uncertainties, e.g., multicriteria analysis (Barton et al. 2005; Berbel et al. 2010). Yet, learning to deal with differences requires a different set of tools and guidelines, and BBNs constructed with active involvement of stakeholders is a group model building effort within this category (van der Keur et al. 2010). What is challenging as part of adaptive groundwater management is the ability to keep a balanced focus, on the one hand dealing with a complex technical task in an aggregated and/or probabilistic way and on the other hand supporting the social relations that are key to social learning.

The EU MERIT research project (Bromley 2005) resulted in a set of guidelines for construction of BBNs with stakeholder involvement for integrated resource management. Thus, the ability of BBNs to integrate knowledge and involve stakeholders following the MERIT guidelines (Bromley 2005) has been documented (Zorrilla et al. 2010).

The MERIT guidelines recommend a process with stakeholder involvement through 7 stages, with 2 interacting flowcharts for each stage, one representing the “network construction process” and the other the “stakeholder involvement process” (Bromley 2005; Henriksen et al. 2007a, 2007b). A fundamental aspect of dealing with adaptive challenges and ambiguity is the adoption of a participatory process that 1) ensures transparency of process and decision, 2) is open to all interests, 3) involves participants representing the broad range of likely interests, 4) ensures clarity of purpose, and 5) gives opportunity for dialogue and discussion. In particular, the first of the 7 MERIT steps—framing the issue, identification and selection of stakeholders, and identifying their views and concerns, roles, and responsibilities (Bromley 2005)—is important when using MERIT guidelines for adaptive and integrated groundwater management.

When applying BBNs and stakeholder involvement, the approach is as follows: First, the stakeholders discuss and identify which actions are available for dealing with a specific groundwater management adaptive challenge that requires the development and analysis of a new solution. Next, they identify a set of indicators suitable for evaluating the consequences for system state variables and planning goals. Between alternative actions and indicators, researchers and stakeholders now add the necessary intermediate variables to allow for analysis of probabilistic knock-on effect of actions on indicators and evaluate causal relationships between variables. Each variable, action, indicator, and intermediate variable requires a definition of states. Finally, data is collected from databases, models, expert opinions, and so forth in order to populate the conditional probability tables defined by the variables and links. Once validated, the BBNs can be used in the analysis of different management scenarios.

The MERIT cases showed (Bromley 2005) that construction of BBNs has a number of distinguishing features “that promote the adoption of dialogue-based methods” (workshops or focus or discussion groups). This adds an additional quality compared with more traditional information-gathering processes (one-to-one interviews, questionnaires, and so forth) such as the use of the BBN tool for documenting dialogue and participatory management processes, where the aim is the process rather than the development of a decision support tool per se (Bromley 2005; Henriksen et al. 2007a, 2007b). These distinguishing features are:

  • The ability to represent scientific and technical complexity in a meaningful way, including the translation of technical or expert language into easily understandable graphical models.

  • Inherent data uncertainty that needs to be characterized and understood represented by explicit variables, links, and/or in conditional probability tables, providing a quantitative participatory integrated assessment.

  • Providing transparency about data availability problems; data, both published and unpublished, can be merged by incorporating data, modeling results, and expert knowledge.

  • Balancing divergent expert and/or scientific opinions by allowing the different points of view to be explicitly incorporated in cases where ambiguity influences interpretation of causal relationships.

  • The iterative nature of BBN construction to combine knowledge and data into a single model is important, because it not only provides a graphical model as a product but also provides a narrative based on the network construction process and stakeholder engagement process.

  • The underlying divergent values, interests, and experience that may need to be reconciled and balanced, where the BBN tool can be powerful in highlighting the sometimes hidden values, interests, and experiences that pose as barriers to change.

  • The need to make tradeoffs, which calls for a tool that can be rapidly updated based on new evidence, and the creation of workshop settings where stakeholders can discuss and present uncertainties that are not yet fully understood.

BBNs as part of scenario development can be used for calculating joint probabilities for decision options and predicting outcomes of management policies. This is of benefit to most adaptive-management teams by promoting group model building and mutual understanding of the system being managed, and also by encouraging rigorous examination of alternative management policies.

However, when dealing with uncertainties, BBNs may also have some weaknesses, such as the fact that feedback loops (common in causal loop diagrams) cannot be handled. It is a requirement for a BBN to design it as a directed acyclic graph (Jensen 2001). The following case discusses the results of testing BBNs for adaptive management in the UGB and focuses on how BBNs facilitated dialogical learning about framing uncertainties.

THE UPPER GUADIANA RIVER BASIN CASE

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

The river basin

The transboundary Guadiana River Basin is shared by Spain and Portugal. It is one of the largest river basins on the Iberian Peninsula, with 55 220 km2 (83%) on Spanish territory. One of the main problems of the river basin is the overexploitation of la Mancha Occidental aquifer in the UGB through the large extractions of water from the groundwater aquifers for agricultural use (Figure 2). Since the 1970s, groundwater table drawdown due to the intensive abstraction of water for irrigation has caused severe negative impacts on wetlands, streams, and rivers, and has resulted in the lowering of groundwater levels by up to 50 m in some places. The drying-out of the upper river and associated wetlands, such as the Tablas de Daimiel National Park, has intensified the main conflicts in the area between farmers and conservationists, among central, regional, and local government water agencies, and between owners of small and large farms (Zorrilla et al. 2010; Brugnach et al. 2011).

thumbnail image

Figure 2. The Guadiana River Basin and Las Tablas de Daimiel wetland area (Henriksen et al. 2007c).

Download figure to PowerPoint

In order to recover these ecosystems, the Spanish government, at the proposal of the Ministry of the Environment, approved a Special Plan for the Upper Guadiana (Plan Especial del Alto Guadiana [PEAG], or simply the Special Plan) on January 11, 2008 (CHG 2008; Aldaya and Llamas 2008). The formal approval of this plan includes a budget of 5500 million Euros for the next 20 years.

Using BBNs to address ambiguity in UGB

Below, the main uncertainties identified and handled using BBNs in the UGB are explained. In the UGB, ambiguity exists with respect to what the water problem is and how it is defined by different actors (Brugnach et al. 2011). Before the BBN process, the problem was understood differently by environmentalists, farmers, policy makers, and the public. Environmentalists, for example, viewed the problem as one of overexploitation of water and focused on the natural areas and wetlands but not on the relationship established between the authorities and farmers. Meanwhile, policy makers focused on the results of bad administration and lack of law reinforcement. Farmers, on the other hand, considered themselves entitled to the use of groundwater and see the problem as one of lack of water, although there are also divergent views among farmers. Some farmers extract groundwater illegally, whereas others are against illegal exploitation. Some farmers recognize the severe situation for the wetlands and seek solutions in more rainfed crops and new irrigation technology, whereas other farmers (a majority) tend to follow a business-as-usual strategy. Furthermore, each of the stakeholder groups used different data to support their own viewpoint.

BBNs helped address ambiguity by 1) making explicit the different frames stakeholders held about the water problem, 2) facilitating dialogue among stakeholders, and 3) enabling the systematization of the UGB problem and an understanding of the interactions among the different elements of the system. Application of the BBN allowed participants to work on a shared view of the problem, where the different perspectives were considered and integrated into a single model. For example, hydrological models carried out by researchers with the cooperation of the Guadiana River Basin Authority were unknown to the stakeholders until the BBN process, and a result of the BBN process was increased understanding and acceptance of the scientific knowledge of the aquifer. Equally, the Water Authority and planning officers were exposed to the experiential, tacit knowledge of farmers as well as problems experienced by farmers on a daily basis, e.g., in relation to conflicting regulatory frameworks.

The integrative capacity of the BBN to include variables, such as how to achieve aquifer recovery and possible consequences of implementation of the Special Plan on regional farming income from reduced water abstractions also revealed some uncertainties concerning the Special Plan, and particularly its potential and limitations to recover both aquifer levels and degraded wetlands. The process of developing BBNs and the meetings that were held during the process provided a unique setting for discussion and debate that facilitated communication through the exchange of ideas and opinions. This supported the process of social learning among the different stakeholders in the UGB.

Some of the disagreements (ambiguities) apply to eliciting probabilities for the BBN. For example, assessing the probabilities of some parent variables exposed to basic disagreements between stakeholders:

  • Conditional probabilities of “water transfer” variable: some of the stakeholders expected a 60 hm3 outcome, whereas other participants expected 300 hm3, and yet others expected nothing at all.

  • Conditional probability of the “vineyard plan” variable: environmental stakeholders believed that the vineyard plan would increase the area occupied by illegal irrigated agriculture, whereas the regional agricultural government expected it to only influence existing legal irrigated farms. Consequently, the regional agricultural government wanted to attach a high probability to the “no new illegal irrigated farms” state of the variable “vineyard plan,” whereas ecologists wanted to attach a low probability value for this state.

BBN construction process: Emerging information and diminishing ambiguity

Through the participatory construction of the BBN, the graphical model was shown to the stakeholders several times. In each of the meetings, 1 to 2 h were devoted to studying, variable by variable, the structure of the BBN in order to come to an agreement about the relevance of each variable and how variables were linked. A total of 93% of the participants agreed that the participatory process and the engagement of stakeholders in the construction of BBNs had enabled the process to “incorporate stakeholder values, assumptions, and preferences into decision-making” (Zorrilla et al. 2010).

Six BBNs were built. The first 2 meetings took place concurrently on May 10, 2007, by organizing stakeholders into 2 groups, one attended by farmers and the regional government and the other by environmental nongovernmental organizations and the water authority. The aim of the meetings was to ensure that all elements and aspects of the system were included. This initial process focused on the identification of all variables and their relationships (the qualitative aspects of the BBN construction). Each group came forward with a proposal for variables and links. Based on an inspection of the initial 2 BBNs (BBN 1 and BBN 2), 5 main variables represented in both of the BBNs were identified by the researchers: regional crop distribution, farming productivity, groundwater abstractions, socio-economic welfare, and ecological integrity. A merged BBN was then developed by the researchers (BBN 3; May 31, 2007) from the 2 BBNs developed by the parallel stakeholder groups. The merged BBN (BBN 3) had 51 variables and was quite complex with many links, especially containing many links specific to the 5 main variables.

The researchers decided to further simplify and rearrange the merged network in order to present a pilot BBN that covered the main variables, honored the requirements for a directed acyclic graph, presented an easy overview for stakeholders to work within a group model-building setting, and which could be used for presenting the quantitative aspects of a BBN and showing how a BBN can be used for inference and analysis (e.g., by entering evidence). The researchers therefore added states and numbers to the probability tables to selected variables (BBN 4; Figure 3) in order to explain to stakeholders what was required as inputs to BBNs for populating probability tables with numbers. The pilot BBN (BBN 4; Figure 3) was finished on January 11, 2008, and contained 38 variables grouped into 3 main categories: actions, indicators, and intermediate variables.

thumbnail image

Figure 3. BBN 4, as presented to the stakeholders at a meeting in Madrid, Spain, on January 11, 2008 (Zorrilla 2009).

Download figure to PowerPoint

BBN 4 (Figure 3) was presented at a number of individual interview meetings with each stakeholder in January 2008. The feedback to qualitative aspects of the merged BBN was collected and suggestions for states and knowledge sources for populating the BBNs with numbers for the probability tables was identified. On the basis of these feedbacks and inputs, a major revision of the BBN was implemented (BBN 5; Figure 4).

thumbnail image

Figure 4. BBN 5, as presented to the stakeholders at a meeting in Madrid, Spain, on February 5, 2008 (Zorrilla 2009).

Download figure to PowerPoint

The main changes from BBN 4 to BBN 5 considered:

  • During the integration of both BBNs, different sources of ambiguity were identified. For example, there was no agreement about the determining factors in purchasing the rights to use water. Some participants argued that this right cannot depend solely on price but also on factors such as the age of the farmer, farmer's occupation (full-time or part-time farmer), type of crops cultivated, and the market price of each crop type. Accordingly, it was decided that the amount of rights purchased were to be introduced directly into BBN 5 by the user of the BBN, in order to study this factor as a “what if” question, where the consequences of purchasing different quantities of rights could be analyzed by the BBN and discussed among the stakeholders. Subsequently, the participants decided on a realistic amount of rights likely to be purchased. The variable representing the rights held were named “buying water rights” and “afforestation plan.”

  • Another ambiguity that emerged during the BBN construction process was the realization that farmer cultivation was not constrained by water availability (because farmers would abstract as much water as they needed) due to insufficient control by the water administration. Moreover, farmers often did not change crops from year to year depending on water availability, when cultivating horticultural woody crops or cereals (and vice versa), but rather due to changing market prices and costs related to each crop. When cultivating wood, grain, and vegetables, there are fewer changes from year to year, compared to crops such as cereals and horticultural products. This led to a rearrangement of variables and links influencing “crop distribution ha/crop.”

  • During the construction of BBN 5, it was decided to dispense with the variables related to catastrophic events for crops (Politica Agraria Común [PAC]), because it was felt by stakeholders that such events will continue to occur in the future, and also due to the fact that stakeholders are not evaluating policies to mitigate or to be included in the process. This was partly the reason behind the restructuring of the BBN regarding the variable “agricultural policy scenarios,” which was recategorized into an action variable from an intermediate variable that is impacted by the parent variables PAC, agrarian policies, and global economies.

  • Initially, the participants distinguished between “irrigation technology” and “irrigation knowledge” (BBN 3). After individual interview consultations, the 2 variables were merged into a single “irrigation modernization” variable (BBN 4), which was used as a parent variable for allowing the BBN to assess impacts on other variables, based on an assessment of the efficacy of modernization plans.

  • It was also decided to add new variables describing areas without irrigation rights in order to include the new potential illegal irrigations, due to the majority of stakeholders agreeing that such illegal irrigations could continually emerge in the basin. Through introducing this variable in the network, a previous ambiguity about the appearance of new illegal farming was explicitly recognized by all the stakeholders, including the River Basin Authority, which is the party responsible for closing illegal wells.

  • A difference from and an example of how ambiguities were dealt with, as well as an important knowledge contribution by the participants, was their recommendation to quantify the buying of water rights as related to the area (ha) of farmland and not as a volume (m3) of water. By shifting from an issue of water volumes to land-use relationships, it was easier to come to a shared understanding and to evaluate, construct, and integrate the quantitative aspects of the BBN, allowing the participants to better appreciate the viewpoint of the other side.

The updated BBN (BBN 5; Figure 4) was presented at a meeting with stakeholders on February 5, 2008. It was explained that the research team had used the information, opinions, and feedbacks from stakeholders as input and as a guide to merge and provide the updated network. Also, the criteria that guided the reconstruction of the updated network were explained.

Based on the outcome of the stakeholder meeting on February 5, 2008, the following changes were decided:

  • To include all uses of water. In addition to water use by agriculture, “urban water use” and “industrial water use” from the aquifers were included.

  • To include the variable “new water uses.” These uses represented examples such as new golf courses, the renewable energy industry, large tourist facilities, and similar developments. The majority of participants were worried about these new uses. There was a great uncertainty about the importance of these new uses, but it was possible to consider them, giving a similar probability to each alternative state.

  • To include the “productivity (Euro per hectare)” of crops as an explicit variable influencing “regional agricultural production.”

The final BBN was presented at the last meeting on April 29, 2008 (BBN 6; Figure 5). The final major changes were to include potential income from the Special Upper Guadiana Plan (based on acquisition of water rights and afforestation) as well as new dry-land farming that could develop as result of the Special Upper Guadiana Plan.

thumbnail image

Figure 5. BBN 6, as presented to the stakeholders at a meeting in Madrid, Spain, on April 29, 2008 (Zorrilla 2009).

Download figure to PowerPoint

Farmers directly influenced the design and framing of the issues represented by the final BBN, changing the acquisition of water rights into a main variable and helping to design the probability tables. Farmers helped to decide which variables to eliminate, such as extreme events.

Social learning in the Upper Guadiana Basin

One of the greatest achievements in using BBNs for UGB was to bring together conflicting stakeholder groups and give them an opportunity to express their different views in an atmosphere of reliability and trust. In fact, the level of distrust meant that prior to the BBN process, there was a preliminary step where stakeholders learned to work with each other, thus creating a safe environment where they could express their opinions. This process enabled the different parties to understand and accept more easily the position of others and to take steps toward finding solutions that were acceptable to everyone (Llamas et al. 2009). This was reflected in the BBNs that were developed (Figures 3 to 6).

thumbnail image

Figure 6. Constructed final BBN for the Upper Guadiana Basin, Spain (Zorilla 2009).

Download figure to PowerPoint

The main outcome in terms of social learning was increased appreciation of other positions, which crystallized into shared beliefs, primarily the inclusion of all water users in the group model building process. The technical aspects of the BBN also helped (i.e., systematic, beyond fragmentation, integration of perspectives from different actors, blending scientific and experiential knowledge) because of its integrative capacity, ultimately adding to ownership and trust.

Participatory BBNs used as part of the NeWater activities increased the farmers' knowledge of the negative effects of groundwater exploitation. Furthermore, knowledge transfer and the shaping of farmers' and the public's knowledge and awareness of key issues and ambiguities (e.g., multiple nonoverlapping knowledge frames). Several agricultural and water policy scenarios were simulated and evaluated with groundwater models and BBNs and were further analyzed using other tools, such as water footprint analysis.

The potential benefits and the success so far of this stakeholder involvement in the BBN construction process supported by the NeWater project has been recognized by the Guadiana River Basin Authority and the Ministry of Environment by inclusion of selected local NeWater team members in the participatory process for the elaboration of the PEAG by 2009 (Martinez-Santos 2007; Zorrilla 2009; Zorrilla et al. 2010).

Dealing with tradeoffs and uncertainties in the Upper Guadiana basin

The BBN was designed to cope with ambiguity related to the hydrological, social, and economic impacts of the Special Plan at the scale of the Mancha Occidental Aquifer. The impacts of 2 adaptation actions proposed in the Special Plan—purchase of irrigation rights and imposition of water volume restrictions—were analyzed in detail. The results showed that with the full implementation of the Special Plan, there was only a 40% chance of aquifer recovery before 2027 (the deadline established by the Water Framework Directive). Full implementation of the Special Plan would lead to a certain reduction of current agrarian economic production as well as a reduction in the number of farmers (Llamas et al. 2009; Zorrilla et al. 2010).

In Table 2, results from the Upper Guadiana case study have been compiled, drawing on the BBN construction process and stakeholder meetings. The table shows examples of groundwater management uncertainty in the UGB identified for natural, technical, and social subsystems. They are described and classified according to layers of the knowledge pyramid and the 3 types of uncertainties mentioned (Table 1).

Table 2. Examples of uncertainties identified in each of the 3 knowledge relationships in the Upper Guadiana case
Uncertainty classification/pyramid levelNatural systemTechnical systemSocial systemSelected strategyMade decision under uncertainty
Scenario developmentAnthropic “impact” on aquifer behaviorAquifer response to PEAG implementationGroundwater sustainabilityReduce uncertaintyTo apply new tools in order to improve the system problem
Numerical groundwater modelsLack of knowledge regarding deepest aquifer (lower layer)Lack of knowledge regarding groundwater abstractionsStakeholders are concerned about quantity more than qualityAccept not knowingTo accept a range of illegal wells (not officially defined)
Hydrogeological regimeLack of knowledge regarding surface–groundwater interactions in ecosystems (mainly wetlands)Affection caused by artificial recharge or water transfers from other basinsProtection of drinking water from areas with nitratesAccept not knowingTo integrate groundwater models into the BBN approach
Geological frameworkLack of knowledge regarding main aquifer boundaries connectionLack of information from test drilling database availableGroundwater users associations without connection among themLearn to deal with uncertaintyBBN was used to help and promote trust and dialogue learning from water managers and farmers
DatabaseLack of water quality information (wells)No public access to pumping informationGroundwater users associations have their own databaseLearn to deal with differences and learn to deal with uncertaintyGuadiana Water Authority promotes dialogue and social participation processes in order to achieve PEAG goals
Ability to integrateStakeholders accepted on the ability of BBN to integrate different types of knowledge 
Ability to engage stakeholdersStakeholders were involved in the whole process of BBN construction, they understood the way BBNs functioned, and they thought the process was useful for them and for their organizations 
Ability to analyze uncertaintiesThe unknown area of irrigated farms was identified as an important uncertainty, and the “purchase of irrigation rights” was identified as one of the most important actions, in which success in its application is one of the biggest uncertainties in the near future 

“Reduce uncertainty” (Table 2) seems to be the best option when uncertainties could significantly affect the performance of the policy measures considered as part of scenario development. The decision made under uncertainty is to apply new tools in order to improve the common understanding of the problem. “Reducing uncertainty” implies a reduction of “strategic and institutional uncertainty” (Koppenjan and Klijn 2004; Bredenhoff-Bijlsma 2010). The construction of BBNs with stakeholder engagement is a way to allow learning and better understanding of the problem as well as identification of measures and solutions that can solve the problem.

In the Guadiana Basin, groundwater modeling was useful in order to better describe the water balance, the groundwater flow system, and impacts from irrigation on groundwater level and wetlands. Groundwater models were used to help managers and water users better understand the hydro-ecological sustainability of the aquifer–wetland system. Rather than postponing needed actions until better data becomes available regarding illegal irrigation amounts, it was decided to “accept not knowing” on these data gaps as the strategy for coping with such uncertainties. Groundwater models only cover a limited part of the problem domain, and other issues such as socioeconomics, land use, groundwater relationships, institutions, values, and beliefs were considered equally or more important. BBNs could be used for data integration available model results, and expert knowledge in order to deal with the urgent problems that required actions to be taken, whereby different policy responses and how these affect the aquifer, wetlands, and farmers' economy and social parameters could be evaluated by BBNs, along with the engagement of stakeholders.

For geological interpretations and data uncertainty, the selected strategy is to “learn to deal with uncertainty.” In this case, a policy that satisfies all involved parties is the goal, despite persistent uncertainty. Tools such as problem analysis and qualitative model building are instrumental for this purpose. By initiating a new process—in this case, the BBN construction process with stakeholder engagement—a way is found to cope with uncertainty by focusing on social learning (e.g., by stimulating different ideas and keeping an open process), by communicating different understandings, and by finding new ways forward. One finding from the Guadiana case shows that the “language of volumes” (previous attempts in enforcing restrictions based on the amount of abstracted water) was evaluated as a failed strategy by stakeholders. Instead, the stakeholders suggested a way forward with a focus on land use (irrigated versus rainfed agricultures). Stakeholders believed this to be a more realistic approach for adaptive management of natural resources (and land use impacts on groundwater quantity and quality) in the Guadiana case, where data uncertainty was significant.

Ambiguity is rooted in future consequences regarding the execution of the Special Plan, and originates from the farmers' implementation of proposed actions, from implementation of actions by the water authority, and especially from the central government's willingness to provide the funds needed for the execution of the Special Plan. At the time of writing, the funds from the central government have not yet been received, partly because of the national economic crisis. Therefore, the Special Plan is implemented in only a limited version. This issue was highlighted as one of the most significant uncertainties during the BBN participatory process.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

Decision support and/or policy advantages of group model building with explicit focus on ambiguity

Dealing with ambiguity requires learning how to deal with differences. Ambiguity in natural resource management has increased during the past decades. Even though new approaches of dealing with uncertainties have involved multiple actors, this has not reduced the scope of uncertainties but rather created ambiguity because it has revealed different viewpoints on the problems. Thus, ambiguity occurs when different actors, such as policy makers, environmental groups, and local groups, express multiple, valid, and sometimes conflicting views on a problem.

The UGB in the Spanish Central Plateau illustrates how ambiguity has hampered sustainable groundwater management in the basin (Brugnach et al. 2011). The different and conflicting views on water availability and sustainable use of water have not been solved, because policy makers, farmers, environmentalists, and the public all viewed the problems differently and to some extent from equally valid angles given their reference to specific or sector-oriented knowledge. Nevertheless, during the participatory process, each stakeholder was able to explain its point of view in a constructive way and to better appreciate the viewpoints of others (although they still do not share them); agreement was even reached on some common solutions.

Adaptive groundwater management processes include some pitfalls not to be overlooked (Beven 2008; Kallis et al. 2009). Even though adaptive groundwater management is an attractive concept, it requires adaptive leadership and institutional frameworks supporting this proactive management approach. Solving distribution or water allocation problems often at the core of contemporary water conflicts (Visscher 2008; Kallis et al. 2009) may not be possible if political and institutional support is lacking. Furthermore, courageous leadership is needed in order make it happen and to keep focus on a desire to learn and change (Allan 2008). On the other hand, bottom-up participatory processes and integrated assessments are needed, especially when dealing with groundwater management and agriculture, where divergent views and understandings are well known (Molden 2009).

BBN modeling strategies specific to the environmental management problem

Within recent decades, when implementing water control measures in the UGB, the authorities adopted a rational problem-solving approach. They did not consider the issues as an adaptive challenge and instead imposed technical solutions to the problems based on scientific evidence and monitoring of data. For example, in the 1980s, the River Basin Authority tried to decrease water withdrawals by banning farmers from drilling new wells. This measure failed because farmers did not understand the thinking behind such a measure, among other reasons, and because they had no other easy farming alternative without irrigation (as the water level fell, they needed to dig new and deeper wells). As a result, there were thousands of new illegal wells, and water withdrawals did not decrease.

Constructing a BBN that included social acceptability and economic appraisal of how alternative measures impact selected indicators resulted in improved integration of physical, ecological, economic, and social factors, as well as a better understanding of other stakeholders' views on defining the problem, and how different solutions could be effective for different indicators in the basin. Furthermore, causal relationships could be explored together with stakeholders in this as part of the BBN group model-building process in order to address knowledge sources, knowledge gaps, and uncertainties between variables (e.g., from the actions in the Special Plan [PEAG] to the agreed indicators: years until aquifer recovery, number of agricultural jobs, and regional agricultural production). After a decade, a shared vision is beginning to emerge in the basin, one that could lead to social learning and collective decisions about a more sustainable groundwater abstraction regime in the region.

The probabilistic nature of BBNs can help capture ambiguity. By introducing different alternative options into one system (that is, the BBN) and eliciting real numbers in probability tables for each relationship, the importance of each alternative option is explicitly expressed. If the BBN is constructed participatorily, the participants can perceive what the others are actually thinking when they propose numbers for the probability tables. By eliciting probability numbers, people are explicitly expressing and learning the importance of available alternative options for the individual stakeholders.

New modeling techniques

The UGB testing of BBN modeling illustrated that BBNs are effective for dealing with ambiguity where dialogues and communication are necessary processes (Henriksen et al. 2007a, 2007b, 2007c; Henriksen and Barlebo 2008; Molina et al. 2009; Martinez-Santos et al. 2010). BBNs can integrate different types of knowledge and shape strategies and actions and provide valuable feedback about delicate variables, links, and conditional probability tables, where different views and perceptions have to be considered (Vennix 1996).

Stakeholder perceptions from the UGB documented that a BBN is an excellent tool for incorporating stakeholder values, assumptions, and preferences (>90% agreed). BBNs provided useful information, improved system understanding, and enhanced understanding of concerns of other stakeholders (>80%). Furthermore, BBNs increased data transparency and assured credibility of output (>70%), due to the graphical interface and the ability to structure the participation process and to encourage dialogue (Zorrilla et al. 2010).

Therefore, the novelty of BBNs is not as a technical tool (because numerous advanced tools exist for covering the physical–technical system) but rather as a tool for dealing with the adaptive challenges (Heifetz et al. 2009) and/or wicked problems (Allan 2008) in groundwater management. Here, the quality of BBNs is the graphical model with its strength in terms of communicating in a meaningful way, in qualitative as well as quantitative terms, as part of group model building (Andersen et al. 2007; Lynam et al. 2007). The tool also has some weaknesses; it cannot incorporate feedback loops, which are quite frequent in causal loop diagrams and system thinking approaches using more conceptual style approaches. However, the ability to incorporate stakeholder values, assumptions, and preferences seems, thanks to the probability functioning of BBNs (Charniak 1991), to make it more useful than advanced technical modeling of feedbacks when dealing with ambiguity.

Specific modeling issues

In the UGB, a large group of stakeholders participated in BBN group model building. Conflicts among the involved groups were overcome thanks to increased trust achieved through common dialogue and communication provided by the BBN construction. In such a difficult conflict with no easy win–win solutions, adaptation requires political will, flexible planning, and inclusive decision processes and tools. Past focus on reducing groundwater abstraction by improving irrigation efficiency had not brought many results. Adjusting actions without focusing on re-opening the problem and broadening the focus to land use and crops that are less water consumptive had proven inadequate. With the new participatory process enabled by BBN construction, the process and requirements for quantifying conditional probabilities created a way of thinking where stakeholders started to understand how their well water level related not only to precipitation but also to abstractions (due to the revisiting of the groundwater flow model which had been unknown to some of the stakeholders). They began to realize how water abstractions influence the regional groundwater table, and how groundwater levels have a direct impact on groundwater discharges to the wetlands in the basin. In the same way, ecologists became more conscious about the problems of farmers and the difficulties they face if they stop irrigating their crops.

The Special Plan (PEAG) did not address the ambiguity, and consequently the adaptation strategies did not target the real institutional barriers or fully explore the innovative possibilities in the basin. And, major uncertainties and sources of ambiguity still remained: will the money ever arrive from the EU for the purchase of irrigation rights? How can the problem with illegal irrigations be diminished (nearly half of the abstractions are illegal; how can illegal water abstractors be incorporated in the group model building process)? BBNs recognized ambiguity and lack of knowledge by allowing the assessment of probabilities in a graphical model. Even for situations that are complex and where events are unpredictable, BBNs and the law of probability helped envision and quantify possible outcomes and futures (Naim 2008). In this way, BBNs are powerful and practical rational approaches for decision making under uncertainty. They allowed us to combine different sources of knowledge and to deal with the ambiguity reflected in multiple frames and viewpoints of stakeholders.

We exemplify this by: 1) commencing construction of the BBN, using water volume as the key issue. In the first meeting, participants came to the conclusion that it was more useful to use land units in most of the variables and only at the end transform them to water abstractions, and 2) ecologists thought that irrigation technology should be accompanied by irrigation knowledge, but it was not important for farmers who thought they were correctly able to manage new technologies, driven by their own interest.

Elicitation and communication strategies

Views vary considerably, as demonstrated by the UGB case, due to conflicting ideas on water, food, and ecosystems in relation to groundwater quantity issues. A major reason for this is the divergent understandings of basic premises (Molden 2009). The Guadiana case shows that one basic premise involves how much groundwater is actually abstracted and used by agriculture for irrigation. The BBNs enabled uncertainties to be explicitly handled and communicated. For example:

  • It was stated that the “different initial (current) irrigation area” variable could have different states, from the smallest assessment of 130 000 ha, to the largest of 260 000 ha.

  • The variable “water transfer” had 3 different states, and stakeholders had to agree with the probability of each state (no water transfer, the biggest possible water transfer, and an intermediate water transfer).

  • The variable “climate change” had 25% of probabilities for each of the 4 scenarios of the Intergovernmental Panel on Climate Change. In addition, for the BBN, consequences of climate change impacts on crop yields were constructed, and irrigation necessities were quantified.

The participants then established that they did not have the same idea of irrigated area size, and that the differences of perceptions were very high. By introducing probabilities of each possible water transfer, they explicitly developed a shared understanding about the nature of difficulties implied by the construction of the water transfer. Finally, by quantifying climate change, they were able to explicitly evaluate possible consequences of climate change, which are usually considered extremely significant, even though it may not end up with a single number. Another source of divergent views relates to differences regarding the value of water used for agriculture and food production versus groundwater used for conserving ecosystems and for supplying drinking water. This dilemma is a key issue in the Guadiana River Basin.

An alternative approach for elicitation and communication about uncertainties could be to use the uncertainty matrix (Refsgaard et al. 2007; van der Keur et al. 2008) for characterizing data, model, and scenario uncertainty. This uncertainty matrix, with a few modifications, could include the layers of the uncertainty pyramid and the uncertainty classification table (Figure 1 and Table 1) as an alternative approach for dealing with uncertainties to the BBN approach. However, we believe that for elicitation and communication purposes about ambiguities and multiple knowledge frame uncertainty when dealing with adaptive challenges, a BBN is a much stronger tool compared to the uncertainty matrix for proper elicitation of ambiguity with stakeholder engagement. For more technical problems, an uncertainty matrix and numerical groundwater models would probably be better than a BBN.

CONCLUSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

The case of the UGB in Spain demonstrated how groundwater management and dealing with various uncertainties can be facilitated by group model building processes with construction of BBNs with stakeholder engagement. This enables integration of sector and domain knowledge, participatory learning processes, and uncertainty assessments. In the UGB case, divergent views and understanding prevailed due to the conflicting choices between water use for agricultural production, for ecosystems, and for drinking water supply. Dealing with this adaptive challenge for contemporary groundwater management in the basin was important in order to assure joint problem understanding and learning processes for exploring and balancing tradeoffs (related to possible actions in the Special Plan for the basin [PEAG]) and to provide increasingly sustainable exploitation of groundwater resources in the near future. BBNs constructed with stakeholder involvement have proven to be an efficient tool for incorporating stakeholder beliefs, values, and perceptions into the decision-making process, which is essential for dealing with divergent views and ambiguity. For analyzing impacts and uncertainties related to adaption strategies, the BBNs are therefore evaluated as an excellent tool for group model building, focused dialogue, bringing to light the uncertainties, and for communicating major sources of uncertainty.

EDITOR'S NOTE

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

This paper represents 1 of 7 review and case study articles generated as a result of a workshop entitled “Scenario and decision analysis in environmental management using Bayesian Belief Networks” (1-2 October 2009, Oslo, Norway) hosted by the Norwegian Institute for Nature Research (NINA) and the Strategic Institute Project “Nature 2020+” and funded by the Research Council of Norway. The main aim of the workshop was to compare Bayesian network applications to different environmental and resource management problems from around the world, identifying common modeling strategies and questions for further research.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES

This work was partly carried out within the project “New approaches for adaptive water management under uncertainty (NeWater)” funded by EC Contract 511179 (GOCE): Integrated Project in Priority 6.3 Global Change and Ecosystems in the 6th EU framework programme.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. DECISION MAKING UNDER UNCERTAINTY: THE KNOWLEDGE PYRAMID
  5. USE OF BAYESIAN BELIEF NETWORKS FOR DEALING WITH ADAPTIVE MANAGEMENT CHALLENGES
  6. THE UPPER GUADIANA RIVER BASIN CASE
  7. DISCUSSION
  8. CONCLUSION
  9. EDITOR'S NOTE
  10. Acknowledgements
  11. REFERENCES
  • Aldaya MM, Llamas MR. 2008. Water footprint analysis for the Guadiana river basin. No. 3. Papeles de Agua Virtual. Madrid (ES): Fundación Marcelino Botín.
  • Allan C. 2008. Can adaptive management help us? In: Pahl-Wostl C, Kabat P, Möltgen J, editors. Adaptive and integrated water management. Coping with complexity and uncertainty. New York (NY): Springer. p 6173.
  • Andersen DF, Vennix JAM, Richardson GP, Rouwette EAJA. 2007. Group model building: problem structuring, policy simulation and decision support. J Operational Res Soc 58: 691694.
  • Axelrod R. 1976. The structure of decision: The cognitive maps of political elite. Princeton (NJ): Princeton Univ.
  • Barreteau O. 2003. The joint use of role-playing games and models regarding negotiation processes: characterization of associations. J Artif Soc Social Simul 6(2): 3 [cited 2011 May 10]. Available from: http://asss.soc.surrey.ac.uk/6/2/3.html
  • Barton D, Saloranta T, Bakkent TH, Solheim AL, Moe J, Selvik JR, Vagstad N. 2005. Using Bayesian network models to incorporate uncertainty in the economic analysis of pollution abatement under the Water Framework Directive. Water Supply 5: 95104.
  • Berbel J, Martin-Ortega J, Mesa P. 2010. A cost-effectiveness analysis of water-saving measures for the water framework directive: the case of the Guadalquivir River Basin in Southern Spain. Water Resour Manag 25: 623640.
  • Beven K. 2008. Environmental modeling: An uncertain future? An introduction to techniques for uncertainty estimation in environmental prediction. London (UK): Taylor & Francis. 351 p.
  • Borsuk ME. 2004. Predictive assessment of fish health and fish kills in the Neuse River Estuary using elicited expert judgment. Hum Ecol Risk Assess 10: 415434.
  • Bredenhoff-Bijlsma R. 2010. Policy development under uncertainty. A framework inspired by cases of water management [PhD thesis]. Gildeprint, Enschede (NL). University of Twente.
  • Bromley J. 2005. Guidelines for the use of Bayesian networks as a participatory tool for Water Resource Management Based on results of the MERIT project (www.merit-eu.net). Wallingford (UK): Centre for Ecology and Hydrology. 117 p.
  • Bromley J, Jackson NA, Clymer OJ, Giacomello AM, Jensen FV. 2005. The use of Hugin to develop Bayesian networks as an aid to integrated water resource planning. Environ Model Software 20: 231242.
  • Brugnach M, Dewulf A, Pahl-Wostl C, Taillieu T. 2008. Towards a relational concept of uncertainty: about knowing too little, knowing too differently and accepting not to know. Ecol Soc 13: 30 [cited 2011 May 10]. Available from: http://www.ecologyandsociety.org/vol13/iss2/art30
  • Brugnach M, van der Keur P, Henriksen HJ, Mysiak J. 2009. Uncertainty in adaptive management. Concepts and guidelines. NeWater synthesis product 2. [cited 2011 May 10]. Available from: www.newater.info
  • Brugnach M, Dewulf C, Henriksen HJ, van der Keur P. 2011. More is not always better: Coping with ambiguity in natural resources management. J Environ Manag 92: 7884.
  • Bryson JM, Ackermann F, Eden C, Finn CB. 2004. Visible thinking: Unlocking causal mapping for practical business results. Chichester (UK): John Wiley & Sons.
  • Carmona G, Varela-Ortega C, Bromley J. 2009. Stakeholder involvement in water management using object-oriented Bayesian networks and economic models in Spain. Contributed Paper prepared for presentation at the International Association of Agricultural Economists Conference, Beijing, China; 2009 Aug 16-22.
  • [CCA] The Council of Canadian Academies. 2009. The sustainable management of groundwater in Canada. The expert panel on groundwater. Ottawa (CA): The Council of Canadian Academies. June 2009.
  • Charniak E. 1991. Bayesian networks without tears. AI Magazine 12: 5063.
  • Checkland P. 2000. The emergent properties of SSM in use: a symposium by reflective practitioners. Systemic Practice and Action Research 13: 779823.
  • [CHG] Guadiana River Basin Authority. 2008. “Confederación Hidrográfica del Guadiana” Guadiana River Basin Authority, Spanish Ministry of Environment [cited 2011 May 10]. Available from: http://www.chguadiana.es/
  • Delli Priscoli J, Llamas MR. 2001. International perspective in ethical dilemmas in the water industry. In: Navigating Rough Waters Davis CK, McGinn RE, editors. Denver (CO): American Water Works Association. p 4164.
  • Domínguez Padilla A. 2004. Gestión integrada de los recursos hídricos de la unidad hidrogeológica Mancha Oriental mediante la utilización de redes bayesianas [PhD thesis]. Universidad de Castilla-La Mancha. Departamento de Producción Vegetal y Tecnología Agraria. E.T.S.I. Agrónomos de Albacete. 252 p.
  • Eden C. 1989. Using cognitive mapping for strategic options development and analysis. In: Rosenhead J, editor. Rational analysis for a problematic world: Problem structuring methods for complexity, uncertainty and conflict. Chichester (UK): John Wiley & Sons. p 21–42.
  • Farmani R, Henriksen HJ, Savic D. 2009. An evolutionary Bayesian belief network methodology for optimum management of groundwater contamination. Environ Model Software 24: 303310.
  • Gray B. 2004. Strong opposition: frame-based resistance to collaboration. J Community Appl Social Psychol 14: 166176.
  • [GWP] Global Water Partnership. 2004. Integrated Water Resource Management, TAC Background Paper 4. Stockholm (SE): Global Water Partnership.
  • Heifetz R, Grashow A, Linsky M. 2009. The practice of adaptive leadership. Tools and tactics for changing your organization and the world. Boston (MA): Harvard Business School.
  • Henriksen HJ. 2004. Test of Bayesian belief network and stakeholder involvement. Groundwater management and protection at Havelse well field in Northern Zealand. EVK1- 2000-0008 5 – MERIT. Danish case study [cited 2011 May 10]. Available from: www.geus.dk/program-areas/water/denmark/rapporter/merit_aug_04-uk.htm
  • Henriksen HJ, Troldborg L, Nyegaard P, Sonnenborg TO, Refsgaard JC, Madsen B. 2003. Methodology for construction, calibration and validation of a national hydrological model for Denmark. J Hydrol 280: 5271.
  • Henriksen HJ, Barlebo H. 2008. Reflections on the use of Bayesian belief networks for adaptive management. J Environ Manag 88: 10251036.
  • Henriksen HJ, Rasmussen P, Brandt G, von Bülow D, Jensen FV. 2007a. Public participation modelling using Bayesian belief networks in management of groundwater contamination. Environ Model Software 22: 11011113.
  • Henriksen HJ, Rasmussen P, Brandt G, von Bulow D, Jensen FV. 2007b. Chapter 3 - Bayesian networks as a participatory modelling tool for groundwater protection. In: Castelletti A, Soncini-Sessa , editors. Topics on System Analysis and Integrated Water Resources Management. Elsevier. p 4972.
  • Henriksen HJ, Rasmussen P, Bromley J, de la Hera Portillo A, Llamas MR. 2007c. Bayesian belief networks as a tool for participatory integrated assessment and adaptive groundwater management: the Upper Guadiana Basin, Spain. Geological Survey of Denmark and Greenland Bulletin 13: 6972.
  • Janssen MA. 2002. Complexity and ecosystem management: the theory and practice of multi-agent systems. Cheltenham (UK): Edward Elgar Publishing.
  • Jensen FV. 2001. Bayesian networks and decision graphs. Statistics for engineering and information science. New York (NY): Springer.
  • Jønch-Clausen T. 2004. Integrated Water Resources Management (IWRM) and Water Efficiency Plans by 2005. Why, What and How? TEC Background Paper10. Stockholm (SE): Global Water Partnership.
  • Kallis G, Kiparsky M, Norgaard R. 2009. Collaborative governance and adaptive management: Lessons from California's CALFED Water Program. Environ Sci Policy 12: 632643.
  • Koppenjan J, Klijn EH. 2004. Managing uncertainties in networks. A network approach to problem solving and decison making. London (UK): Routledge.
  • Llamas MR, Martínez-Santos P, de la Hera A. 2007. The manifold dimensions of groundwater sustainability: An overview. In: Ragone S, Plumbey T, Hernández-Mora N, de la Hera A, McKay G, Bergkamp G, editors. The global importance of groundwater in the 21st Century: Proceedings of the international symposium on groundwater sustainability 2006 Jan 24-27; Alicante, Spain. Westerville (OH): National Ground Water Association. p 105116.
  • Llamas MR, Varela C, de la Hera A, Aldaya MM, Villarroya F, Martínez-Santos P, Blanco-Gutierrez IMM, Carmona-García G, Esteve-Bengoechea P, De Stefano L., et al. 2009. The Guadiana Basin. In: Mysiak J, Henriksen HJ, Sullivan C, Bromley J, Pahl-Wostl C, editors. The adaptive water resource management handbook. London (UK): Earthscan. p 104115.
  • Lynam T, de Jong W, Sheil D, Kusumanto T, Evans K. 2007. A review of tools incorportating community knowledge, preferences, and values into decision making in natural resources management. Ecol Soc 12: 5 [cited 2011 May 10]. Available from: http://www.ecologyandsociety.org/vol112/iss1/art5/
  • Martín de Santa Olalla F, Dominguez A, Artigao A, Fabeiro C, Ortega JF. 2005. Integrated water resources management of the hydrogeological unit “Eastern Mancha” using Bayesian Belief Networks. Agric Water Manag 77: 2136.
  • Martinez-Santos P. 2007. Hacia la gestión adaptable del acuífero de la Mancha Occidental. [PhD thesis]. Universidad Complutense de Madrid. 375 p.
  • Martínez-Santos P, Henriksen HJ, Zorrilla P, Mártinez-Alfaro PE. 2010. Comparative reflections on the use of modelling tools in conflictive water management settings: The Mancha Occidental aquifer, Spain. Environ Model Software 25: 14391449.
  • Matott LS, Babendreier JE, Purucker ST. 2009. Evaluating uncertainty in integrated environmental models: A review of concepts and tools. Water Resour Res 45: W06421. DOI: 10.1029/2008WR007301
  • Molden D. 2009. Water for food. Water for life. A comprehensive assessment of water management in agriculture. London (UK): Earthscan.
  • Molina JL, Arostegui JLLG, Benavente J, Varela C, de la Hera A, Geta JAL. 2009. Aquifers overexploitation in SE Spain: A proposal for the integrated analysis of water management. Water Resour Manag 23: 27372760.
  • Mysiak J, Pahl-Wostl C, Sullivan C, Bromley J, Henriksen HJ. 2009. The adaptive water resource management handbook. London (UK): Earthscan.
  • Naim P. 2008. In: Pourret O, Naim P, Marcot B, Bayesian networks. A practical guide to applications. Hoboken (NJ): John Wiley & Sons. p 377–384.
  • Pahl-Wostl C, Kabat P, Möltgen J. 2008. Adaptive and integrated water management. New York (NY): Springer.
  • Refsgaard JC, van der Sluijs , Højberg AL, Vanrolleghem PA. 2007. Uncertainty in the environmental modelling process: A framework and guidance. Environ Model Software 22: 15431556.
  • Rittel HWJ, Weber MM. 1973. Dilemmas in a general theory of planning. Policy Sci 4: 155169.
  • Sendzimir J, Magnuszewski P, Balogh P, Vari A. 2007. Anticipatory modelling of biocomplexity in the Tisza river: First steps to establish a participatory adaptive framework. Environ Model Software 25: 113135.
  • Shihab K. 2005. Modeling groundwater quality with Bayesian techniques. Intelligent Systems Design and Applications 2005 ISDA '05 Proceedings of the 5th International Conference on Intelligent Systems Design and Applications; 2005 Sep 8-10. Washington (DC): IEEE Computer Society. p 73–78.
  • Sterman J. 2000. Business dynamics: System thinking and modelling for a complex world. New York (NY): Irwin/McGraw-Hill.
  • van der Keur P, Henriksen HJ, Refsgaard JC, Brugnach M, Pahl-Wostl C, Dewulf A, Buiteveld H. 2008. Identification of major sources of uncertainty in current IWRM practice. Illustrated for the Rhine case. Water Resour Manag 22: 16771708.
  • van der Keur P, Brugnach M, DeWulf A, Refsgaard JC, Zorilla P, Poolman M, Henriksen HJ, Warmink JJ, Isendahl N, Raadgever GT., et al. 2010. Identifying uncertainty guidance for supporting policy making in water management illustrated for Upper Guadiana and Rhine basins. Water Resour Manag 22: 39013938.
  • Vennix JAM. 1996. Group model building: Facilitating team learning using system dynamics. East Sussex (UK): John Wiley & Sons.
  • Visscher HW. 2008. Internal: feedback and team performance. An empirical (survey) study into the relationship of internal (co-worker) feedback with team performance, with conflict and trust as mediating factors, and a comparison between the feedback environment for internal- and external feedback [Master thesis]. Heerlen (NL): School of Management, Open Universiteit Nederland.
  • Xu YP, Tung YK. 2008. Decision making in water managment under uncertainty. Water Resour Manag 22: 535550.
  • Zorrilla P. 2009. Análisis de la gestión del agua en el acuífero de la Mancha Occidental: construcción de una red bayesiana mediante procesos de participación pública [PhD thesis]. Madrid (ES): Faculty of Geological Sciences. Universidad Autónoma de Madrid.
  • Zorrilla P, Carmona G, de la Hera A, Varela-Ortega C, Martinez-Santos P, Bromley J, Henriksen HJ. 2010. Evaluation of Bayesian networks as a tool for participatory water resources management: Application to the Upper Guadiana Basin in Spain. Ecol Soc 15: 12 [cited 2011 May 10]. Available from: http://www.ecologyandsociety.org/vol15/iss3/art12/