Review
Reviewing Bayesian Networks potentials for climate change impacts assessment and management: A multi-risk perspective

https://doi.org/10.1016/j.jenvman.2017.07.044Get rights and content

Highlights

  • Bayesian Networks applications for environmental management are critically reviewed.

  • Potentials of Bayesian Networks for climate change multi-risk assessment are discussed.

  • Main limitations are highlighted and future developments suggested.

Abstract

The evaluation and management of climate change impacts on natural and human systems required the adoption of a multi-risk perspective in which the effect of multiple stressors, processes and interconnections are simultaneously modelled. Despite Bayesian Networks (BNs) are popular integrated modelling tools to deal with uncertain and complex domains, their application in the context of climate change still represent a limited explored field. The paper, drawing on the review of existing applications in the field of environmental management, discusses the potential and limitation of applying BNs to improve current climate change risk assessment procedures. Main potentials include the advantage to consider multiple stressors and endpoints in the same framework, their flexibility in dealing and communicate with the uncertainty of climate projections and the opportunity to perform scenario analysis. Some limitations (i.e. representation of temporal and spatial dynamics, quantitative validation), however, should be overcome to boost BNs use in climate change impacts assessment and management.

Introduction

Climate change risk assessment represents a challenging task for environmental management due to the inherent complexity of socio-ecological systems, the multiplicity of processes and the high degree of uncertainty, variability and randomness involved (Döll and Romero-Lankao, 2017, Gallina et al., 2016).

Multiple climatic and non-climatic stressors interact, inducing impacts which can be highly correlated (i.e. cumulative, synergic or antagonistic effect) or strongly dependent (i.e. cascading or triggering effects) one each other (Gill and Malamud, 2014, Kappes et al., 2012, Liu et al., 2014). Neglecting these interactions can lead to an underestimation of the overall risk and further to inefficient or controversial risk management strategies (i.e. maladaptation, unexpected environmental impacts, increase vulnerability or exposure toward other kind of stressors) (Liu et al., 2014). An effective risk management should be built on a good understanding of all relevant threats affecting the target of interest thus enabling decision makers and practitioner to develop efficient adaptation plans based on a robust prioritization of risk reduction measures (Komendantova et al., 2014).

However, due to the differing characteristics of hazards, few quantitative models that suite a fully multi-risk perspective exist. Most climate change studies still are predominantly mono-disciplinary, designed to consider impacts triggered by individual shocks (i.e. hazards) and analyzing mono-causal and mono-temporal cause-effect relationships (Gallina et al., 2016; Næss, 2010).

At the same time, considering uncertainty as a pervasive issue in climate change, it should be included as a key component of each risk assessment model.

Despite this, most of risk assessment models still rely on the traditional definition of risk considering the probability of an event and its negative consequences (UNISDR, 2009). Instead, deep uncertainty about future risk could be better addressed by risk scenarios describing the range of plausible future environmental and socio-economic conditions (Döll and Romero-Lankao, 2017, Stirling, 2010, Willows and Connell, 2003).

The aforementioned considerations clearly suggest that climate change risk assessment and management, required a shift from traditional risk assessment based on a single stressor approach, toward a more challenging multi risk and adaptive paradigm (Döll and Romero-Lankao, 2017, Landis et al., 2013a).

New approaches should be included in the current environmental risk assessment and management procedures considering the adoption of integrated models (i.e. Bayesian Networks (BNs), System Dynamics (SD), Agent-Based Models (ABMs), Artificial Neural Networks (ANNs) and Expert Systems) able to: incorporate multiple stressors and endpoints (i.e. social, economic and environmental objectives and priorities); ii) deal with uncertainty; iii) take into account the effect of policy and adaptation in changing final system states (Hamilton et al., 2015, Kelly et al., 2013).

Bayesian Networks (BNs) in the last decade have become a recognized tool to deal with environmental problems and decision making under uncertainty (Varis and Kuikka, 1997) and some authors (Catenacci and Giupponi, 2010, Döll and Romero-Lankao, 2017, Hart and Pollino, 2008, Landis et al., 2013a, Pollino and Hart, 2008), suggested their use for risk assessment.

Bayesian Networks (BNs), also known as Bayesian Belief Networks (BBNs) or Belief Networks, are probabilistic graphical models representing a set of random variables and their conditional interdependencies via a Directed Acyclic Graph (DAG) (Pearl, 1988) thus using probabilistic expressions to describe the relationships among system components (Borsuk et al., 2004).

BNs, originally emerged from research into artificial intelligence (Charniak, 1991, Heckerman et al., 1995, Jensen, 1996, Pearl, 2011), have been applied with a risk assessment perspective to many different environmental issues (e.g. Integrated Water Resource Management, ecology, maritime spatial planning, fishery, agronomy) (Castelletti and Soncini Sessa, 2007; Barton et al., 2008, Borsuk et al., 2004, Borsuk et al., 2003, Bromley et al., 2005, Farmani et al., 2009, Gudimov et al., 2012, Henriksen et al., 2007, Lecklin et al., 2011, Renken and Mumby, 2009, Stelzenmüller et al., 2010, Zorrilla et al., 2010; Little et al., 2004).

However, the application of BNs in the context of climate change still represent a limited explored field if compare with others where their use has exponential growth during recent years. Only very few studies which explicitly apply BNs to assess and manage climate change impacts on natural resources (i.e. water scarcity and deterioration, soil erosion, biodiversity loss, eutrophication, sea-level rise) can be found in literature (Catenacci and Giupponi, 2013, Dyer et al., 2011, Gutierrez et al., 2011, Kotta et al., 2010, Molina et al., 2013, Tighe et al., 2007).

The purpose of the present work is, therefore, to discuss the application of BNs to improve climate change risks assessment and management.

The paper wants to explore to what extent BNs can be useful to enhance environmental risk assessment, through the analysis of multi-stressors issues in several case studies mainly related to natural hazards, sustainable resources management and pollution prevention in view of climate change.

Finally, it describes the use of BNs for the environmental management of climate change impacts, providing some insights of their functionalities for risks prioritization, uncertainty communication and to support practitioners in the selection of optimal adaptation measures at the regional and local scale.

To do so, a systematic analysis of existing literature is presented in Section 1. In Section 2, main advantages and limitations of BNs are discussed according to each steps of the general framework of multi-risk assessment and management (Fig. 1) thus providing a sort of “road map” for the integration of BNs also in climate change risk assessment and management procedures.

Section snippets

Reviewing Bayesian Networks applications

According to the aim of the present work, a set of case studies dealing with BNs applications in the field of environmental modelling and management were selected and analyzed.

To facilitate a comparative analysis and discussion, case studies have been categorized according to specific features (i.e. model objectives, model structure, model parametrization, model evaluation, scenarios analysis, stakeholders involvement, representation of spatial and temporal dynamics) recognized as fundamental

Potentials and limits of BNs for climate change multi-risk assessment and management

The assessment and management of environmental and climate impacts from a multi-risk perspective can be seen as an extension of common environmental risk assessment and management frameworks in which particular effort have to be posed to the identification and quantification of stressors interactions (Dawson, 2015, Gallina et al., 2016). Artificial Intelligence based models (e.g. Artificial Neural Networks (ANN), Bayesian Networks (BN) and fuzzy modelling approaches), integrating different

Conclusions

The paper provides a review of BNs applications in the field of sustainable resources management and natural hazard with the aim to explore the potential use of BNs as modelling tools to improve current climate change risk assessment and management procedure. The results highlight that, despite BNs have been applied to a large variety of problems and contexts (i.e. water resource management, oil spills, transport of pollutants, hydrological hazards), their application in climate change studied

Acknowledgments

The research leading to these results has received funding from the Euro-Mediterranean Center on Climate Change (CMCC), Strategic Projects Funds (2016–2018).

References (89)

  • R. Giordano et al.

    Bayesian belief network to support conflict analysis for groundwater protection: the case of the Apulia region

    J. Environ. Manage.

    (2013)
  • P. Haapasaari et al.

    Formalizing expert knowledge to compare alternative management plans: sociological perspective to the future management of Baltic salmon stocks

    Mar. Policy

    (2010)
  • S.H. Hamilton et al.

    Integrated assessment and modelling: overview and synthesis of salient dimensions

    Environ. Model. Softw.

    (2015)
  • I. Helle et al.

    A probabilistic approach for a cost-benefit analysis of oil spill management under uncertainty: a Bayesian network model for the Gulf of Finland

    J. Environ. Manage.

    (2015)
  • H.J. Henriksen et al.

    Public participation modelling using Bayesian networks in management of groundwater contamination

    Environ. Model. Softw.

    (2007)
  • D. Inman et al.

    Perceived effectiveness of environmental decision support systems in participatory planning: evidence from small groups of end-users

    Environ. Model. Softw.

    (2011)
  • A.R. Keshtkar et al.

    Application of Bayesian networks for sustainability assessment in catchment modeling and management (Case study: the Hablehrood river catchment)

    Ecol. Modell.

    (2013)
  • N. Komendantova et al.

    Multi-hazard and multi-risk decision-support tools as a part of participatory risk governance: feedback from civil protection stakeholders

    Int. J. Disaster Risk Reduct.

    (2014)
  • N.N. Kourgialas et al.

    Statistical analysis and ANN modeling for predicting hydrological extremes under climate change scenarios: the example of a small Mediterranean agro-watershed

    J. Environ. Manage.

    (2015)
  • D. Landuyt et al.

    An ecosystem service approach to support integrated pond management: a case study using Bayesian belief networks–Highlighting opportunities and risks

    J. Environ. Manage.

    (2014)
  • T. Lecklin et al.

    A Bayesian network for analyzing biological acute and long-term impacts of an oil spill in the Gulf of Finland

    Mar. Pollut. Bull.

    (2011)
  • L.R. Little et al.

    Information flow among fishing vessels modelled using a Bayesian network

    Environ. Model. Softw.

    (2004)
  • J.-L. Molina et al.

    Dynamic Bayesian networks as a decision support tool for assessing climate change impacts on highly stressed groundwater systems

    J. Hydrol.

    (2013)
  • T.A. Morton et al.

    The future thatmay (or may not) come: how framing changes responses to uncertainty in climate change communications

    Glob. Environ. Change.

    (2011)
  • T.D. Phan et al.

    Applications of Bayesian belief networks in water resource management: a systematic review

    Environ. Model. Softw.

    (2016)
  • C.A. Pollino et al.

    Parameterisation and evaluation of a Bayesian network for use in an ecological risk assessment

    Environ. Model. Softw.

    (2007)
  • C.A. Pshenichny et al.

    The event bush as a semantic-based numerical approach to natural hazard assessment (exemplified by volcanology)

    Comput. Geosci.

    (2009)
  • J. Qiu et al.

    Modeling method of cascading crisis events based on merging Bayesian Network

    Decis. Support Syst.

    (2014)
  • H. Renken et al.

    Modelling the dynamics of coral reef macroalgae using a Bayesian belief network approach

    Ecol. Modell.

    (2009)
  • P.L. Spence et al.

    Effects of nitrogen inputs on freshwater wetland ecosystem services–A Bayesian network analysis

    J. Environ. Manage.

    (2013)
  • V. Stelzenmüller et al.

    Assessment of a Bayesian Belief Network–GIS framework as a practical tool to support marine planning

    Mar. Pollut. Bull.

    (2010)
  • L. Uusitalo

    Advantages and challenges of Bayesian networks in environmental modelling

    Ecol. Modell.

    (2007)
  • L. Uusitalo et al.

    Exploring methods for predicting multiple pressures on ecosystem recovery: a case study on marine eutrophication and fisheries

    Cont. Shelf Res.

    (2016)
  • S. Balbi et al.

    A spatial Bayesian network model to assess the benefits of early warning for urban flood risk to people

    Nat. Hazards Earth Syst. Sci. Discuss.

    (2015)
  • M.E. Borsuk et al.

    Integrated approach to total maximum daily load development for Neuse River Estuary using Bayesian probability network model (Neu-BERN)

    J. Water Resour. Plan. Manag.

    (2003)
  • M. Burgman

    Risks and Decisions for Conservation and Environmental Management

    (2005)
  • J. Cain

    Planning Improvements in Natural Resource Management. Guidelines for Using Bayesian Networks to Support the Planning and Management of Development Programmes in the Water Sector and beyond

    (2001)
  • J. Cain

    Planning improvements in natural resources management

    Cent. Ecol. Hydrol. Wallingford, U. K.

    (2001)
  • M. Catenacci et al.

    Potentials and Limits of Bayesian Networks to Deal with Uncertainty in the Assessment of Climate Change Adaptation Policies

    (2010)
  • T.U. Chan et al.

    Bayesian network models for environmental flow decision making in the Daly River, Northern Territory, Australia

    River Res. Appl.

    (2012)
  • E. Charniak

    Bayesian networks without tears

    AI Mag.

    (1991)
  • R.J. Dawson

    Handling interdependencies in climate change risk assessment

    Climate

    (2015)
  • F.J.M. de Santa Olalla et al.

    Integrated water resources management of the hydrogeological unit “Eastern Mancha” using Bayesian belief networks

    Agric. Water Manag.

    (2005)
  • C.K. den Heijer et al.

    Impact assessment of extreme storm events using a Bayesian network

    Coast. Eng. Proc.

    (2012)
  • Cited by (80)

    View all citing articles on Scopus
    View full text