Bayesian Networks and participatory modelling in water resource management
Introduction
The last decade has witnessed a growing interest in the application of graphical models, such as Bayesian Networks (Bns), in environmental and natural resource modelling. This new trend is closely linked to the recognition that participation and uncertainty have a key role in integrated natural resource management and that there is a need for tools and methodologies that make it easier to handle them.
In the environmental field, many authors have illustrated the use of Bns starting with Varis (1995), who first generalized and rearranged the mathematical framework provided by Pearl (1988) to fit the features of environmental resource systems. Applications range from ecological issues, such as fisheries and related problems (Kuikka et al., 1999, Borsuk et al., 2002, Little et al., 2004), to the assessment of the effects of climate changes on crop production (Gu et al., 1996). In the water resource context they have been used by Batchelor and Cain (1999) in irrigated and rainfed farming system modelling, by Varis and Kuikka (1997) to investigate the effect of climate change on surface waters and by Borsuk et al., 2001, Borsuk et al., 2004 in studying the eutrophication of river estuaries. The works of Baran and Jantunen (2004) and Bromley et al. (2005) focus particularly on Bns as tools to support and improve participation.
The majority of these works use Bns to model the whole system being studied and thus do not discuss their integration with other types of models. In this paper the pros and cons of adopting Bns for modelling either part of or the whole system are analyzed, by framing the modelling activity within the context of a participatory and integrated planning procedure, and exploring the integration of Bns with other types of models. We first consider the problem of decision making in the field of water management in a general perspective, regardless of the modelling approach adopted. Then we will go through the process of model construction and compare Bns with other types of models, indicating the distinct characteristics of the Bn modelling approach.
In this paper only two uses of Bns are considered:
- (1)
for modelling, when they are used to describe the system being studied;
- (2)
for aiding decision making, when they include decision and utility nodes, and are employed as a decision support system (DSS).
A third use does exist, however: Bns can be used as a visualization tool to summarize simply the outcomes of more complex models. This tool may be part of a more complex DSS, or the DSS itself, and in this case it can be traced back to the second use. However, in both cases the Bn will not be perceived by Decision Makers and stakeholders as a model of the real system, since this role is conceptually and psychologically played by the underlying more complex model. Since the paper aims at analysing the role of Bns in participatory modelling within a decision making process, it is apparent why it will not focus on this third use. The paper is organized as follows: the structure and alternative ways of using Bayesian Networks are described in the first section. We evaluate their usefulness by looking at the problem of decision making in water resource planning in a general perspective; therefore, in the second section we introduce a Participatory and Integrated Planning (PIP) procedure, highlighting the key role of participatory modelling. The third section is entirely devoted to this latter topic, and we go through the construction of a water system model, showing where participation comes in, how the model is an aggregation of models which describes the subsystems that constitute the water system, and how the type of each one of these models can be selected from four different types, one of which is the Bn. At that point we will have all the necessary ingredients to focus on the role of Bn in water management in the last section.
Section snippets
Bayesian Networks
Bayesian Networks (also known as Belief Networks or Bayesian Belief Networks) are a powerful modelling technique that replicates the essential features of plausible reasoning (reasoning in conditions of uncertainty) in a consistent, efficient and mathematically sound way (Charniak, 1991). They were first developed by the artificial intelligence and machine learning community (Pearl, 1986, Pearl, 1988 and Jensen, 1996) and successfully applied in the fields of medical diagnosis (Andreassen
Decision making in water management
Participation and integration are increasingly accepted by international organizations and national authorities as central principles for decision making in the environmental field. For instance, they were included in the Water Framework Directive (WFD, Directive/2000/60/EC), which defines the rules for planning and managing water resources in the EU. The promotion and application of these principles in practical decision making is effectively enhanced by adopting a procedural approach setting
Participatory modelling
Participatory modelling requires not only that the stakeholders be aware of a model's underlying assumptions, limitations and intended uses, but also that they go through the same thinking process as the modeller. They should be exposed to the same information and issues that are encountered during the model identification process. Only in this way they can share a quantitative understanding of the system and be assured that what is being modelled actually reflects their understanding of the
The Bns' role in water management
The previous sections focused on the key role of participatory modelling in water resource decision making and introduced four types of models, from which the modeller must select one to model each of the components of the water system being considered. Defining a few precise selection criteria is far from easy (see Jakeman et al., in press).
The social acceptance of a model, i.e. whether it is widely and regularly applied in the same or similar contexts, might undoubtedly make it more appealing
Conclusions
In the last decade Bns have captured the interest of the environmental modelling community thanks to their friendly semantics and the graphical support they provide, which is useful for interaction with stakeholders. The wide availability of ready-to-use software that allows Bn models to be designed and implemented on a PC has further contributed to their spread. In this paper, we have illustrated the pros and cons of their use in water resource planning and management, by contextualizing their
Acknowledgement
The present work was carried out within the Project COFIN 2004 Sistemi di supporto alle decisioni per la pianificazione e gestione di serbatoi e laghi regolati [Contract 2004132971_004] partly funded by the Italian Ministry of Education, University and Research.
References (38)
- et al.
Medical expert systems based on causal probabilistic networks
International Journal of Biomedical Computing
(1991) - et al.
Application of belief networks to water management studies
Agricultural Water Management
(1999) - et al.
A Bayesian hierarchical model to predict benthic oxygen demand from organic matter loading in estuaries and coastal zones
Ecological Modelling
(2001) - et al.
A Bayesian network of eutrophication models for synthesis, prediction, and uncertainty analysis
Ecological Modelling
(2004) - et al.
The use of Hugin to develop Bayesian Network as an aid to integrated water resource planning
Environmental Modelling & Software
(2005) - et al.
Coupling real time control and socio-economic issues in participatory river basin planning
Environmental Modelling & Software
(2007) - et al.
A procedural approach to strengthening integration and participation in water resource planning
Environmental Modelling & Software
(2006) Hexagons for systems thinking
European Journal of Operational Research
(1992)- et al.
Inference in belief networks: a procedural guide
International Journal of Approximate Reasoning
(1996) - et al.
Information flow among fishing vessels modelled using a Bayesian network
Environmental Modelling & Software
(2004)
Fusion, propagation, and structuring in Belief Networks
Artificial Intelligence
Fusion and propagation with multiple observations in belief networks
Artificial Intelligence
Stakeholder consultation for bayesian decision support systems in environmental management
A Bayesian belief network for modelling brown trout (Salmo trutta) populations in Switzerland
Comprehensive testing and application of the pip procedure: the Verbano project case study
Topics on System Analysis and Integrated Water Resource Management
Parsimonious modelling of water and suspended-sediment flux from nested-catchments affected by selective tropical forestry
Changes and Disturbance in Tropical Rainforest in South–East Asia
Bayesian networks without tears
AI Magazine
Directive 2000/60/EC of the European Parliament and of the Council establishing a framework for Community action in the field of water policy
Official Journal
A belief network-based system for predicting future crop production
AI Applications
Cited by (250)
Predicting the impacts of anthropogenic drivers on management scenarios using Bayesian belief network in the Zeribar freshwater wetland, Iran
2024, Journal for Nature ConservationHow Bayesian networks are applied in the subfields of climate change: Hotspots and evolution trends
2024, Environmental Modelling and SoftwareAssessing compound pluvial-fluvial flooding: Research status and ways forward
2023, Water SecurityDesigning an expert-led Bayesian network to understand interactions between policy instruments for adoption of eco-friendly farming practices
2023, Environmental Science and Policy