Bayesian Networks in Environmental Risk Assessment: A Review

Causal, predictive or observational? Different understandings of key event relationships for adverse outcome pathways and their implications on practice

The Adverse Outcome Pathways (AOPs) framework is pivotal in toxicology, but the, terminology describing Key Event Relationships (KERs) varies within AOP guidelines.This study examined the usage of causal, observational and predictive terms in AOP, documentation and their adaptation in AOP development. A literature search and text, analysis of key AOP guidance documents revealed nuanced usage of these terms, with KERs often described as both causal and predictive. The adaptation of, terminology varies across AOP development stages. Evaluation of KER causality often, relies targeted blocking experiments and weight-of-evidence assessments in the, putative and qualitative stages. Our findings highlight a potential mismatch between,terminology in guidelines and methodologies in practice, particularly in inferring,causality from predictive models. We argue for careful consideration of terms like, causal and essential to facilitate interdisciplinary communication. Furthermore, integrating known causality into quantitative AOP models remains a challenge.

Probabilistic ecological risk assessment for deep-sea mining: A Bayesian network for Chatham Rise, Pacific Ocean

Increasing interest in seabed resource use in the ocean is introducing new pressures on deep-sea environments, the ecological impacts of which need to be evaluated carefully. The complexity of these ecosystems and the lack of comprehensive data pose significant challenges to predicting potential impacts. In this study, we demonstrate the use of Bayesian networks (BNs) as a modeling framework to address these challenges and enhance the development of robust quantitative predictions concerning the effects of human activities on deep-seafloor ecosystems. The approach consists of iterative model building with experts, and quantitative probability estimates of the relative decrease in abundance of different functional groups of benthos following seabed mining. The model is then used to evaluate two alternative seabed mining scenarios to identify the major sources of uncertainty associated with the mining impacts. By establishing causal connections between the pressures associated with potential mining activities and various components of the benthic ecosystem, our model offers an improved comprehension of potential impacts on the seafloor environment. We illustrate this approach using the example of potential phosphorite nodule mining on the Chatham Rise, offshore Aotearoa/New Zealand, SW Pacific Ocean, and examine ways to incorporate knowledge from both empirical data and expert assessments into quantitative risk assessments. We further discuss how ecological risk assessments can be constructed to better inform decision-making, using metrics relevant to both ecology and policy. The findings from this study highlight the valuable insights that BNs can provide in evaluating the potential impacts of human activities. However, further research and data collection are crucial for refining and ground truthing these models and improving our understanding of the long-term consequences of deep-sea mining and other anthropogenic activities on marine ecosystems. By leveraging such tools, policymakers, researchers, and stakeholders can work together toward human activities in the deep sea that minimize ecological harm and ensure the conservation of these environments.

Development of a probabilistic risk model for pharmaceuticals in the environment under population and wastewater treatment scenarios

Preparing for future environmental pressures requires projections of how relevant risks will change over time. Current regulatory models of environmental risk assessment (ERA) of pollutants such as pharmaceuticals could be improved by considering the influence of global change factors (e.g., population growth) and by presenting uncertainty more transparently. In this article, we present the development of a prototype object-oriented Bayesian network (BN) for the prediction of environmental risk for six high-priority pharmaceuticals across 36 scenarios: current and three future population scenarios, combined with infrastructure scenarios, in three Norwegian counties. We compare the risk, characterized by probability distributions of risk quotients (RQs), across scenarios and pharmaceuticals. Our results suggest that RQs would be greatest in rural counties, due to the lower development of current wastewater treatment facilities, but that these areas consequently have the most potential for risk mitigation. This pattern intensifies under higher population growth scenarios. With this prototype, we developed a hierarchical probabilistic model and demonstrated its potential in forecasting the environmental risk of chemical stressors under plausible demographic and management scenarios, contributing to the further development of BNs for ERA. Integr Environ Assess Manag 2024;20:1715-1735. © 2024 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

Investigating PCB degradation by indigenous fungal strains isolated from the transformer oil-contaminated site: degradation kinetics, Bayesian network, artificial neural networks, QSAR with DFT, molecular docking, and molecular dynamics simulation

The widespread prevalence of polychlorinated biphenyls (PCBs) in the environment has raised major concerns due to the associated risks to human health, wildlife, and ecological systems. Here, we investigated the degradation kinetics, Bayesian network (BN), quantitative structure-activity relationship-density functional theory (QSAR-DFT), artificial neural network (ANN), molecular docking (MD), and molecular dynamics stimulation (MS) of PCB biodegradation, i.e., PCB-10, PCB-28, PCB-52, PCB-138, PCB-153, and PCB-180 in the soil system using fungi isolated from the transformer oil-contaminated sites. Results revealed that the efficacy of PCB biodegradation best fits the first-order kinetics (R2 ≥ 0.93). The consortium treatment (29.44-74.49%) exhibited more efficient degradation of PCBs than those of Aspergillus tamarii sp. MN69 (27.09-71.25%), Corynespora cassiicola sp. MN69 (23.76-57.37%), and Corynespora cassiicola sp. MN70 (23.09-54.98%). 3'-Methoxy-2, 4, 4'-trichloro-biphenyl as an intermediate derivative was detected in the fungal consortium treatment. The BN analysis predicted that the biodegradation efficiency of PCBs ranged from 11.6 to 72.9%. The ANN approach showed the importance of chemical descriptors in decreasing order, i.e., LUMO > MW > IP > polarity no. > no. of chlorine > Wiener index > Zagreb index > HOMU > Pogliani index > APE in PCB removal. Furthermore, the QSAR-DFT model between the chemical descriptors and rate constant (log K) exhibited a high fit and good robustness of R2 = 99.12% in predicting ability. The MD and MS analyses showed the lowest binding energy through normal mode analysis (NMA), implying stability in the interactions of the docked complexes. These findings provide crucial insights for devising strategies focused on natural attenuation, holding substantial potential for mitigating PCB contamination within the environment.

Examining site intervention efficacy and uncertainties with conceptual Bayesian networks: preventing offsite migration of DNAPL and contaminated groundwater

For contaminated sites, conceptual site models (CSMs) guide the assessment and management of risks, including remediation strategies. Recent research has expanded diagrammatic CSMs with structural causal modeling to develop what are nominally called conceptual Bayesian networks (CBNs) for environmental risk assessment. These CBNs may also be useful for problems of controlling and preventing offsite contaminant migration, especially for sites containing dense nonaqueous phase liquids (DNAPLs). In particular, the CBNs provide greater clarity on the causal relationships between source term, onsite and offsite migration, and remediation effectiveness characterization for contaminated DNAPL sites compared to traditional CSMs. These ideas are demonstrated by the inclusion of modifying variables, causal pathway analysis, and interventions in CBNs. Additionally, several new extensions of the CBN concept are explored including the representation of measurement variables as lines of evidence and alignment with conventional pictorial CSMs for groundwater modeling. Taken as a whole, the CBNs provide a powerful and adaptable knowledge representation tool for remediating subsurface systems contaminated by DNAPL.

Strategies for integrating scientific evidence in water policy and law in the face of uncertainty

Understanding how human actions and environmental change affect water resources is crucial for addressing complex water management issues. The scientific tools that can produce the necessary information are ecological indicators, referring to measurable properties of the ecosystem state; environmental monitoring, the data collection process that is required to evaluate the progress towards reaching water management goals; mathematical models, linking human disturbances with the ecosystem state to predict environmental impacts; and scenarios, assisting in long-term management and policy implementation. Paradoxically, despite the rapid generation of data, evolving scientific understanding, and recent advancements in systems modeling, there is a striking imbalance between knowledge production and knowledge utilization in decision-making. In this paper, we examine the role and potential capacity of scientific tools in guiding governmental decision-making processes and identify the most critical disparities between water management, policy, law, and science. We demonstrate how the complex, uncertain, and gradually evolving nature of scientific knowledge might not always fit aptly to the legislative and policy processes and structures. We contend that the solution towards increased understanding of socio-ecological systems and reduced uncertainty lies in strengthening the connections between water management theory and practice, among the scientific tools themselves, among different stakeholders, and among the social, economic, and ecological facets of water quality management, law, and policy. We conclude by tying in three knowledge-exchange strategies, namely - adaptive management, Driver-Pressure-Status-Impact-Response (DPSIR) framework, and participatory modeling - that offer complementary perspectives to bridge the gap between science and policy.

Key challenges and developments in wildlife ecological risk assessment: Problem formulation

Problem formulation (PF) is a critical initial step in planning risk assessments for chemical exposures to wildlife, used either explicitly or implicitly in various jurisdictions to include registration of new pesticides, evaluation of new and existing chemicals released to the environment, and characterization of impact when chemical releases have occurred. Despite improvements in our understanding of the environment, ecology, and biological sciences, few risk assessments have used this information to enhance their value and predictive capabilities. In addition to advances in organism-level mechanisms and methods, there have been substantive developments that focus on population- and systems-level processes. Although most of the advances have been recognized as being state-of-the-science for two decades or more, there is scant evidence that they have been incorporated into wildlife risk assessment or risk assessment in general. In this article, we identify opportunities to consider elevating the relevance of wildlife risk assessments by focusing on elements of the PF stage of risk assessment, especially in the construction of conceptual models and selection of assessment endpoints that target population- and system-level endpoints. Doing so will remain consistent with four established steps of existing guidance: (1) establish clear protection goals early in the process; (2) consider how data collection using new methods will affect decisions, given all possibilities, and develop a decision plan a priori; (3) engage all relevant stakeholders in creating a robust, holistic conceptual model that incorporates plausible stressors that could affect the targets defined in the protection goals; and (4) embrace the need for iteration throughout the PF steps (recognizing that multiple passes may be required before agreeing on a feasible plan for the rest of the risk assessment). Integr Environ Assess Manag 2024;20:658-673. © 2022 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC). This article has been contributed to by U.S. Government employees and their work is in the public domain in the USA.

Elicitation of Rank Correlations with Probabilities of Concordance: Method and Application to Building Management

Constructing Bayesian networks (BN) for practical applications presents significant challenges, especially in domains with limited empirical data available. In such situations, field experts are often consulted to estimate the model's parameters, for instance, rank correlations in Gaussian copula-based Bayesian networks (GCBN). Because there is no consensus on a 'best' approach for eliciting these correlations, this paper proposes a framework that uses probabilities of concordance for assessing dependence, and the dependence calibration score to aggregate experts' judgments. To demonstrate the relevance of our approach, the latter is implemented to populate a GCBN intended to estimate the condition of air handling units' components-a key challenge in building asset management. While the elicitation of concordance probabilities was well received by the questionnaire respondents, the analysis of the results reveals notable disparities in the experts' ability to quantify uncertainty. Moreover, the application of the dependence calibration aggregation method was hindered by the absence of relevant seed variables, thus failing to evaluate the participants' field expertise. All in all, while the authors do not recommend to use the current model in practice, this study suggests that concordance probabilities should be further explored as an alternative approach for the elicitation of dependence.

Operationalizing a fisheries social-ecological system through a Bayesian belief network reveals hotspots for its adaptive capacity in the southern North sea

Fisheries social-ecological systems (SES) in the North Sea region confront multifaceted challenges stemming from environmental changes, offshore wind farm expansion, and marine protected area establishment. In this paper, we demonstrate the utility of a Bayesian Belief Network (BN) approach in comprehensively capturing and assessing the intricate spatial dynamics within the German plaice-related fisheries SES. The BN integrates ecological, economic, and socio-cultural factors to generate high-resolution maps of profitability and adaptive capacity potential (ACP) as prospective management targets. Our analysis of future scenarios, delineating changes in spatial constraints, economics, and socio-cultural aspects, identifies factors that will exert significant influence on this fisheries SES in the near future. These include the loss of fishing grounds due to the installation of offshore wind farms and marine protected areas, as well as reduced plaice landings due to climate change. The identified ACP hotspots hold the potential to guide the development of localized management strategies and sustainable planning efforts by highlighting the consequences of management decisions. Our findings emphasize the need to consider detailed spatial dynamics of fisheries SES within marine spatial planning (MSP) and illustrate how this information may assist decision-makers and practitioners in area prioritization. We, therefore, propose adopting the concept of fisheries SES within broader integrated management approaches to foster sustainable development of inherently dynamic SES in a rapidly evolving marine environment.

Incorporating climate projections in the environmental risk assessment of pesticides in aquatic ecosystems

Global climate change will significantly impact the biodiversity of freshwater ecosystems, both directly and indirectly via the exacerbation of impacts from other stressors. Pesticides form a prime example of chemical stressors that are expected to synergize with climate change. Aquatic exposures to pesticides might change in magnitude due to increased runoff from agricultural fields, and in composition, as application patterns will change due to changes in pest pressures and crop types. Any prospective chemical risk assessment that aims to capture the influence of climate change should properly and comprehensively account for the variabilities and uncertainties that are inherent to projections of future climate. This is only feasible if they probabilistically propagate extensive ensembles of climate model projections. However, current prospective risk assessments typically make use of process-based models of chemical fate that do not typically allow for such high-throughput applications. Here, we describe a Bayesian network model that does. It incorporates a two-step univariate regression model based on a 30-day antecedent precipitation index, circumventing the need for computationally laborious mechanistic models. We show its feasibility and application potential in a case study with two pesticides in a Norwegian stream: the fungicide trifloxystrobin and herbicide clopyralid. Our analysis showed that variations in pesticide application rates as well as precipitation intensity lead to variations in in-stream exposures. When relating to aquatic risks, the influence of these processes is reduced and distributions of risk are dominated by effect-related parameters. Predicted risks for clopyralid were negligible, but the probability of unacceptable future environmental risks due to exposure to trifloxystrobin (i.e., a risk quotient >1) was 8%-12%. This percentage further increased to 30%-35% when a more conservative precautionary factor of 100 instead of 30 was used. Integr Environ Assess Manag 2024;20:384-400. © 2023 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

Incorporation of climate change into a multiple stressor risk assessment for the Chinook salmon (Oncorhynchus tshawytscha) population in the Yakima River, Washington, USA

One outcome of the 2022 Society of Environmental Toxicology and Chemistry Pellston Workshop on incorporating climate change predictions into ecological risk assessments was the key question of how to integrate ecological risk assessments that focus on contaminants with the environmental alterations from climate projections. This article summarizes the results of integrating selected direct and indirect effects of climate change into an existing Bayesian network previously used for ecological risk assessment. The existing Bayesian Network Relative Risk Model integrated the effects of two organophosphate pesticides (malathion and diazinon), water temperature, and dissolved oxygen levels on the Chinook salmon population in the Yakima River Basin (YRB), Washington, USA. The endpoint was defined as the entity, Yakima River metapopulation, and the attribute was defined as no decline to a subpopulation or the overall metapopulation. In this manner, we addressed the management objective of no net loss of Chinook salmon, an iconic and protected species. Climate change-induced changes in water quality parameters (temperature and dissolved oxygen levels) used models based on projected climatic conditions in the 2050s and 2080s by the use of a probabilistic model. Pesticide concentrations in the original model were modified assuming different scenarios of pest control strategies in the future, because climate change may alter pest numbers and species. Our results predict that future direct and indirect changes to the YRB will result in a greater probability that the salmon population will continue to fail to meet the management objective of no net loss. As indicated by the sensitivity analysis, the key driver in salmon population risk was found to be current and future changes in temperature and dissolved oxygen, with pesticide concentrations being not as important. Integr Environ Assess Manag 2024;20:419-432. © 2023 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

Integrating climate model projections into environmental risk assessment: A probabilistic modeling approach

The Society of Environmental Toxicology and Chemistry (SETAC) convened a Pellston workshop in 2022 to examine how information on climate change could be better incorporated into the ecological risk assessment (ERA) process for chemicals as well as other environmental stressors. A major impetus for this workshop is that climate change can affect components of ecological risks in multiple direct and indirect ways, including the use patterns and environmental exposure pathways of chemical stressors such as pesticides, the toxicity of chemicals in receiving environments, and the vulnerability of species of concern related to habitat quality and use. This article explores a modeling approach for integrating climate model projections into the assessment of near- and long-term ecological risks, developed in collaboration with climate scientists. State-of-the-art global climate modeling and downscaling techniques may enable climate projections at scales appropriate for the study area. It is, however, also important to realize the limitations of individual global climate models and make use of climate model ensembles represented by statistical properties. Here, we present a probabilistic modeling approach aiming to combine projected climatic variables as well as the associated uncertainties from climate model ensembles in conjunction with ERA pathways. We draw upon three examples of ERA that utilized Bayesian networks for this purpose and that also represent methodological advancements for better prediction of future risks to ecosystems. We envision that the modeling approach developed from this international collaboration will contribute to better assessment and management of risks from chemical stressors in a changing climate. Integr Environ Assess Manag 2024;20:367-383. © 2023 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

Using a Bayesian Network Model to Predict Risk of Pesticides on Aquatic Community Endpoints in a Rice Field-A Southern European Case Study

Bayesian network (BN) models are increasingly used as tools to support probabilistic environmental risk assessments (ERAs), because they can better account for uncertainty compared with the simpler approaches commonly used in traditional ERA. We used BNs as metamodels to link various sources of information in a probabilistic framework, to predict the risk of pesticides to aquatic communities under given scenarios. The research focused on rice fields surrounding the Albufera Natural Park (Valencia, Spain), and considered three selected pesticides: acetamiprid (an insecticide), 2-methyl-4-chlorophenoxyacetic acid (MCPA; a herbicide), and azoxystrobin (a fungicide). The developed BN linked the inputs and outputs of two pesticide models: a process-based exposure model (Rice Water Quality [RICEWQ]), and a probabilistic effects model (Predicts the Ecological Risk of Pesticides [PERPEST]) using case-based reasoning with data from microcosm and mesocosm experiments. The model characterized risk at three levels in a hierarchy: biological endpoints (e.g., molluscs, zooplankton, insects, etc.), endpoint groups (plants, invertebrates, vertebrates, and community processes), and community. The pesticide risk to a biological endpoint was characterized as the probability of an effect for a given pesticide concentration interval. The risk to an endpoint group was calculated as the joint probability of effect on any of the endpoints in the group. Likewise, community-level risk was calculated as the joint probability of any of the endpoint groups being affected. This approach enabled comparison of risk to endpoint groups across different pesticide types. For example, in a scenario for the year 2050, the predicted risk of the insecticide to the community (40% probability of effect) was dominated by the risk to invertebrates (36% risk). In contrast, herbicide-related risk to the community (63%) resulted from risk to both plants (35%) and invertebrates (38%); the latter might represent (in the present study) indirect effects of toxicity through the food chain. This novel approach combines the quantification of spatial variability of exposure with probabilistic risk prediction for different components of aquatic ecosystems. Environ Toxicol Chem 2024;43:182-196. © 2023 The Authors. Environmental Toxicology and Chemistry published by Wiley Periodicals LLC on behalf of SETAC.

Do we really remember the view? The cellar door schema and its contribution to memorable experiences: Recommendations for cellar door practices

The importance of enjoyable, memorable cellar door experiences is well-established in the literature. The winescape, which incorporates views, building design, and ambience is recognised as a central motivation for wine tourism and the most repeated content in word-of-mouth communication, a valuable marketing tool. Recent research has prioritised human interaction, which develops a connection as the most important component of the cellar door experience (d'Ament, Nayeem, & Saliba, 2022). The current study expands previous research methodologies, adopting memory work and cellar door surveys in a mixed methods approach to explore the cellar door schema and its influence on cellar door expectations, assessments, purchases, and future positive word-of-mouth communication. A constructivist grounded theory approach was adopted to analyse participant memories. A Bayesian network was produced from 136 cellar door surveys to determine the influence of cellar door schema on purchases and intention to engage in word-of-mouth communication. The results supported recent findings that the human element is the most remembered and valued; it fosters a connection, strengthens brand attachment and creates enduring customers. The winescape, while important for grounding the memory, is less prominent in recollections. Additionally, the results demonstrate the importance of word-of-mouth as a contributor to cellar door schemas. Recommendations are made for cellar door managers and staff who strive to create memorable cellar door experiences for their customers.

Providing an approach to analyze the risk of central oxygen tanks in hospitals during the COVID-19 pandemic

The central oxygen unit of hospitals is considered a high-risk unit, requiring high safety standards to maintain the integrity of the system during the COVID-19 pandemic. The linear reasoning assumption of conventional risk analysis methods cannot adequately describe these modern systems, which are characterized by tight connections and complex interactions between technical, human, and organizational aspects. Therefore, this study presents a new and comprehensive approach to oxygen tanks in hospitals during the COVID-19 pandemic. In this study, trapezoidal fuzzy numbers were used to calculate failure rates. After determining the probability of basic events (BEs), intermediate events (IE), and top event (TE) with fuzzy logic and transferring it into Bayesian Network (BN), deductive and inductive reasoning, and sensitivity analysis were performed using RoV in GeNIe software. The results of the case study showed that the IE of "Human Error" had the highest probability of fuzzy fault tree (FFT) and the probability of oxygen leakage was lower using FBN than FFT. According to the results, BE16 (failure to use standard and updated instructions) and BE12 (defects in the inspection and testing program of tank devices) had the highest posterior probability, while based on the FFT results, BE4 (defects in the external coating system of the tank) and, BE3 (Corrosive environment (acidity state)) had the least probability. According to the sensitivity analysis, basic events 10, 11, and 16 were the most important in the oxygen leakage event with a very small difference, which was almost in line with the results of posterior FBN (FBNPO). Updating the existing guidelines, fixing defects in the inspection of all types of tank gauges, and testing related equipment can greatly help the reliability of these tanks. Root cause analysis of these events provides opportunities for prevention and emergency response in critical situations, such as the COVID-19 pandemic.

Influence of climate change and pesticide use practices on the ecological risks of pesticides in a protected Mediterranean wetland: A Bayesian network approach

Pollution by agricultural pesticides is one of the most important pressures affecting Mediterranean coastal wetlands. Pesticide risks are expected to be influenced by climate change, which will result in an increase of temperatures and a decrease in annual precipitation. On the other hand, pesticide dosages are expected to change given the increase in pest resistance and the implementation of environmental policies like the European ´Farm-to-Fork` strategy, which aims for a 50 % reduction in pesticide usage by 2030. The influence of climate change and pesticide use practices on the ecological risks of pesticides needs to be evaluated making use of realistic environmental scenarios. This study investigates how different climate change and pesticide use practices affect the ecological risks of pesticides in the Albufera Natural Park (Valencia, Spain), a protected Mediterranean coastal wetland. We performed a probabilistic risk assessment for nine pesticides applied in rice production using three climatic scenarios (for the years 2008, 2050 and 2100), three pesticide dosage regimes (the recommended dose, and 50 % increase and 50 % decrease), and their combinations. The scenarios were used to simulate pesticide exposure concentrations in the water column of the rice paddies using the RICEWQ model. Pesticide effects were characterized using acute and chronic Species Sensitivity Distributions built with toxicity data for aquatic organisms. Risk quotients were calculated as probability distributions making use of Bayesian networks. Our results show that future climate projections will influence exposure concentrations for some of the studied pesticides, yielding higher dissipation and lower exposure in scenarios dominated by an increase of temperatures, and higher exposure peaks in scenarios where heavy precipitation events occur right after pesticide application. Our case study shows that pesticides such as azoxystrobin, difenoconazole and MCPA are posing unacceptable ecological risks for aquatic organisms, and that the implementation of the ´Farm-to-Fork` strategy is crucial to reduce them.

The Everglades vulnerability analysis: Linking ecological models to support ecosystem restoration

Understanding of the Everglades’ ecological vulnerabilities and restoration needs has advanced over the past decade but has not been applied in an integrated manner. To address this need, we developed the Everglades Vulnerability Analysis (EVA), a decision support tool that uses modular Bayesian networks to predict the ecological outcomes of a subset of the ecosystem’s health indicators. This tool takes advantage of the extensive modeling work already done in the Everglades and synthesizes information across indicators of ecosystem health to forecast long-term, landscape-scale changes. In addition, the tool can predict indicator vulnerability through comparison to user-defined ideal system states that can vary in the level of certainty of outcomes. An integrated understanding of the Everglades system is essential for evaluation of trade-offs at local, regional, and system-wide scales. Through EVA, Everglades restoration decision makers can provide effective guidance during restoration planning and implementation processes to mitigate unintended consequences that could result in further damage to the Everglades system.

Accuracies of various types of spinal robot in robot-assisted pedicle screw insertion: a Bayesian network meta-analysis

BACKGROUND

With the popularization of robot-assisted spinal surgeries, it is still uncertain whether robots with different designs could lead to different results in the accuracy of pedicle screw placement. This study aimed to compare the pedicle screw inserting accuracies among the spinal surgeries assisted by various types of robot and estimate the rank probability of each robot-assisted operative technique involved.

METHODS

The electronic literature database of PubMed, Web of Science, EMBASE, CNKI, WANFANG and the Cochrane Library was searched in November 2021. The primary outcome was the Gertzbein-Robbins classification of pedicle screws inserted with various operative techniques. After the data extraction and direct meta-analysis process, a network model was established in the Bayesian framework and further analyses were carried out.

RESULTS

Among all the 15 eligible RCTs, 4 types of robot device, namely Orthbot, Renaissance, SpineAssist and TiRobot, were included in this study. In the network meta-analysis, the Orthbot group (RR 0.27, 95% CI 0.13-0.58), the Renaissance group (RR 0.33, 95% CI 0.14-0.86), the SpineAssist group (RR 0.14, 95% CI 0.06-0.34) and the conventional surgery group (RR 0.21, 95% CI 0.13-0.31) were inferior to the TiRobot group in the proportion of grade A pedicle screws. Moreover, the results of rank probabilities revealed that in terms of accuracy, the highest-ranked robot was TiRobot, followed by Renaissance and Orthbot.

CONCLUSIONS

In general, current RCT evidence indicates that TiRobot has an advantage in the accuracy of the pedicle screw placement, while there is no significant difference among the Orthbot-assisted technique, the Renaissance-assisted technique, the conventional freehand technique, and the SpineAssist-assisted technique in accuracy.

'Small Data' for big insights in ecology

Big Data science has significantly furthered our understanding of complex systems by harnessing large volumes of data, generated at high velocity and in great variety. However, there is a risk that Big Data collection is prioritised to the detriment of 'Small Data' (data with few observations). This poses a particular risk to ecology where Small Data abounds. Machine learning experts are increasingly looking to Small Data to drive the next generation of innovation, leading to development in methods for Small Data such as transfer learning, knowledge graphs, and synthetic data. Meanwhile, meta-analysis and causal reasoning approaches are evolving to provide new insights from Small Data. These advances should add value to high-quality Small Data catalysing future insights for ecology.

Intelligent lung cancer MRI prediction analysis based on cluster prominence and posterior probabilities utilizing intelligent Bayesian methods on extracted gray-level co-occurrence (GLCM) features

Lung cancer is the second foremost cause of cancer due to which millions of deaths occur worldwide. Developing automated tools is still a challenging task to improve the prediction. This study is specifically conducted for detailed posterior probabilities analysis to unfold the network associations among the gray-level co-occurrence matrix (GLCM) features. We then ranked the features based on t-test. The Cluster Prominence is selected as target node. The association and arc analysis were determined based on mutual information. The occurrence and reliability of selected cluster states were computed. The Cluster Prominence at state ≤330.85 yielded ROC index of 100%, relative Gini index of 99.98%, and relative Gini index of 100%. The proposed method further unfolds the dynamics and to detailed analysis of computed features based on GLCM features for better understanding of the hidden dynamics for proper diagnosis and prognosis of lung cancer.

Critical gaps in nanoplastics research and their connection to risk assessment

Reports of plastics, at higher levels than previously thought, in the water that we drink and the air that we breathe, are generating considerable interest and concern. Plastics have been recorded in almost every environment in the world with estimates on the order of trillions of microplastic pieces. Yet, this may very well be an underestimate of plastic pollution as a whole. Once microplastics (<5 mm) break down in the environment, they nominally enter the nanoscale (<1,000 nm), where they cannot be seen by the naked eye or even with the use of a typical laboratory microscope. Thus far, research has focused on plastics in the macro- (>25 mm) and micro-size ranges, which are easier to detect and identify, leaving large knowledge gaps in our understanding of nanoplastic debris. Our ability to ask and answer questions relating to the transport, fate, and potential toxicity of these particles is disadvantaged by the detection and identification limits of current technology. Furthermore, laboratory exposures have been substantially constrained to the study of commercially available nanoplastics; i.e., polystyrene spheres, which do not adequately reflect the composition of environmental plastic debris. While a great deal of plastic-focused research has been published in recent years, the pattern of the work does not answer a number of key factors vital to calculating risk that takes into account the smallest plastic particles; namely, sources, fate and transport, exposure measures, toxicity and effects. These data are critical to inform regulatory decision making and to implement adaptive management strategies that mitigate risk to human health and the environment. This paper reviews the current state-of-the-science on nanoplastic research, highlighting areas where data are needed to establish robust risk assessments that take into account plastics pollution. Where nanoplastic-specific data are not available, suggested substitutions are indicated.

A Risk Treatment Strategy Model for Oil Pipeline Accidents Based on a Bayesian Decision Network Model

Risk treatment is an effective way to reduce the risk of oil pipeline accidents. Many risk analysis and treatment strategies and models have been established based on the event tree method, bow-tie method, Bayesian network method, and other methods. Considering the characteristics of the current models, a risk treatment strategy model for oil pipeline accidents based on Bayesian decision network (BDNs) is proposed in this paper. First, the quantitative analysis method used in the Event-Evolution-Bayesian model (EEB model) is used for risk analysis. Second, the consequence weights and initial event likelihoods are added to the risk analysis model, and the integrated risk is obtained. Third, the risk treatment strategy model is established to achieve acceptable risk with optimal resources. The risk treatment options are added to the Bayesian network (BN) risk analysis model as the decision nodes and utility nodes. In this approach, the BN risk analysis model can be transformed into a risk treatment model based on BDNs. Compared to other models, this model can not only identify the risk factors comprehensively and illustrate the incident evolution process clearly, but also can support diverse risk treatment strategies for specific cases, such as to reduce the integrated risk to meet acceptable criterion or to balance the benefit and cost of an initiative. Furthermore, the risk treatment strategy can be updated as the risk context changes.

Bayesian dynamic profiling and optimization of important ranked energy from gray level co-occurrence (GLCM) features for empirical analysis of brain MRI

Accurate classification of brain tumor subtypes is important for prognosis and treatment. Researchers are developing tools based on static and dynamic feature extraction and applying machine learning and deep learning. However, static feature requires further analysis to compute the relevance, strength, and types of association. Recently Bayesian inference approach gains attraction for deeper analysis of static (hand-crafted) features to unfold hidden dynamics and relationships among features. We computed the gray level co-occurrence (GLCM) features from brain tumor meningioma and pituitary MRIs and then ranked based on entropy methods. The highly ranked Energy feature was chosen as our target variable for further empirical analysis of dynamic profiling and optimization to unfold the nonlinear intrinsic dynamics of GLCM features extracted from brain MRIs. The proposed method further unfolds the dynamics and to detailed analysis of computed features based on GLCM features for better understanding of the hidden dynamics for proper diagnosis and prognosis of tumor types leading to brain stroke.

Action selection in risk assessment with fuzzy Fine-Kinney-based AHP-TOPSIS approach: a case study in gas plant

In this study, the hazards occurring in a medium-sized gas filling facility were defined, and the risk scores of these hazards were determined by the expert team according to the Fine-Kinney risk analysis method. However, since the same risk significance score is obtained in different combinations of scale values in the classical Fine-Kinney risk analysis method and the characteristics/constraints of the company applied in the risk analysis are not taken into account, the hazards were evaluated using fuzzy Fine-Kinney risk analysis, and the most critical hazards were determined. Action plans are defined for critical hazards determined as a result of fuzzy Fine-Kinney risk analysis. Among the actions that require company resources, the action selection was performed with the TOPSIS method, taking into account their relationship with the hazards by integrating the weights, which was calculated with the AHP method, of affected groups. The effect of operating constraints is included in the last step of the study to calculate the final weights. Calculating the results by including the risk-affected groups and company constraints and ranking the actions reveals that the study is an original, objective, and applicable study.

A Bayesian Dynamic Inference Approach Based on Extracted Gray Level Co-Occurrence (GLCM) Features for the Dynamical Analysis of Congestive Heart Failure

The adoptability of the heart to external and internal stimuli is reflected by heart rate variability (HRV). Reduced HRV can be a predictor of post-infarction mortality. In this study, we propose an automated system to predict and diagnose congestive heart failure using short-term heart rate variability analysis. Based on the nonlinear, nonstationary, and highly complex dynamics of congestive heart failure, we extracted multimodal features to capture the temporal, spectral, and complex dynamics. Recently, the Bayesian inference approach has been recognized as an attractive option for the deeper analysis of static features, in order to perform a comprehensive analysis of extracted nodes (features). We computed the gray level co-occurrence (GLCM) features from congestive heart failure signals and then ranked them based on ROC methods. This study focused on utilizing the dissimilarity feature, which is ranked as highly important, as a target node for the empirical analysis of dynamic profiling and optimization, in order to explain the nonlinear dynamics of GLCM features extracted from heart failure signals, and distinguishing CHF from NSR. We applied Bayesian inference and Pearson’s correlation (PC). The association, in terms of node force and mapping, was computed. The higher-ranking target node was used to compute the posterior probability, total effect, arc contribution, network profile, and compression. The highest value of ROC was obtained for dissimilarity, at 0.3589. Based on the information-gain algorithm, the highest strength of the relationship was obtained between nodes “dissimilarity” and “cluster performance” (1.0146), relative to mutual information (81.33%). Moreover, the highest relative binary significance was yielded for dissimilarity for 1/3rd (80.19%), 2/3rd (74.95%) and 3/3rd (100%). The results revealed that the proposed methodology can provide further in-depth insights for the early diagnosis and prognosis of congestive heart failure.

Bayesian Network Applications for Sustainable Holistic Water Resources Management: Modeling Opportunities for South Africa

Anthropogenic transformation of land globally is threatening water resources in terms of quality and availability. Managing water resources to ensure sustainable utilization is important for a semiarid country such as South Africa. Bayesian networks (BNs) are probabilistic graphical models that have been applied globally to a range of water resources management studies; however, there has been very limited application of BNs to similar studies in South Africa. This article explores the benefits and challenges of BN application in the context of water resources management, specifically in relation to South Africa. A brief overview describes BNs, followed by details of some of the possible opportunities for BNs to benefit water resources management. These include the ability to use quantitative and qualitative information, data, and expert knowledge. BN models can be integrated into geographic information systems and predict impact of ecosystem services and sustainability indicators. With additional data and information, BNs can be updated, allowing for integration into an adaptive management process. Challenges in the application of BNs include oversimplification of complex systems, constraints of BNs with categorical nodes for continuous variables, unclear use of expert knowledge, and treatment of uncertainty. BNs have tremendous potential to guide decision making by providing a holistic approach to water resources management.

A critical review of effect modeling for ecological risk assessment of plant protection products

A wide diversity of plant protection products (PPP) is used for crop protection leading to the contamination of soil, water, and air, which can have ecotoxicological impacts on living organisms. It is inconceivable to study the effects of each compound on each species from each compartment, experimental studies being time consuming and cost prohibitive, and animal testing having to be avoided. Therefore, numerous models are developed to assess PPP ecotoxicological effects. Our objective was to provide an overview of the modeling approaches enabling the assessment of PPP effects (including biopesticides) on the biota. Six categories of models were inventoried: (Q)SAR, DR and TKTD, population, multi-species, landscape, and mixture models. They were developed for various species (terrestrial and aquatic vertebrates and invertebrates, primary producers, micro-organisms) belonging to diverse environmental compartments, to address different goals (e.g., species sensitivity or PPP bioaccumulation assessment, ecosystem services protection). Among them, mechanistic models are increasingly recognized by EFSA for PPP regulatory risk assessment but, to date, remain not considered in notified guidance documents. The strengths and limits of the reviewed models are discussed together with improvement avenues (multigenerational effects, multiple biotic and abiotic stressors). This review also underlines a lack of model testing by means of field data and of sensitivity and uncertainty analyses. Accurate and robust modeling of PPP effects and other stressors on living organisms, from their application in the field to their functional consequences on the ecosystems at different scales of time and space, would help going toward a more sustainable management of the environment. Graphical Abstract Combination of the keyword lists composing the first bibliographic query. Columns were joined together with the logical operator AND. All keyword lists are available in Supplementary Information at https://doi.org/10.5281/zenodo.5775038 (Larras et al. 2021).

Bayesian Networks for Preprocessing Water Management Data

Environmental data often present inconveniences that make modeling tasks difficult. During the phase of data collection, two problems were found: (i) a block of five months of data was unavailable, and (ii) no information was collected from the coastal area, which made flood-risk estimation difficult. Thus, our aim is to explore and provide possible solutions to both issues. To avoid removing a variable (or those missing months), the proposed solution is a BN-based regression model using fixed probabilistic graphical structures to impute the missing variable as accurately as possible. For the second problem, the lack of information, an unsupervised classification method based on BN was developed to predict flood risk in the coastal area. Results showed that the proposed regression solution could predict the behavior of the continuous missing variable, avoiding the initial drawback of rejecting it. Moreover, the unsupervised classifier could classify all observations into a set of groups according to upstream river behavior and rainfall information, and return the probability of belonging to each group, providing appropriate predictions about the risk of flood in the coastal area.

ECORISK2050: An Innovative Training Network for predicting the effects of global change on the emission, fate, effects, and risks of chemicals in aquatic ecosystems

By 2050, the global population is predicted to reach nine billion, with almost three quarters living in cities. The road to 2050 will be marked by changes in land use, climate, and the management of water and food across the world. These global changes (GCs) will likely affect the emissions, transport, and fate of chemicals, and thus the exposure of the natural environment to chemicals. ECORISK2050 is a Marie Skłodowska-Curie Innovative Training Network that brings together an interdisciplinary consortium of academic, industry and governmental partners to deliver a new generation of scientists, with the skills required to study and manage the effects of GCs on chemical risks to the aquatic environment. The research and training goals are to: (1) assess how inputs and behaviour of chemicals from agriculture and urban environments are affected by different environmental conditions, and how different GC scenarios will drive changes in chemical risks to human and ecosystem health; (2) identify short-to-medium term adaptation and mitigation strategies, to abate unacceptable increases to risks, and (3) develop tools for use by industry and policymakers for the assessment and management of the impacts of GC-related drivers on chemical risks. This project will deliver the next generation of scientists, consultants, and industry and governmental decision-makers who have the knowledge and skillsets required to address the changing pressures associated with chemicals emitted by agricultural and urban activities, on aquatic systems on the path to 2050 and beyond.

A Bayesian Modelling Framework for Integration of Ecosystem Services into Freshwater Resources Management

Models of ecological response to multiple stressors and of the consequences for ecosystem services (ES) delivery are scarce. This paper describes a methodology for constructing a BBN combining catchment and water quality model output, data, and expert knowledge that can support the integration of ES into water resources management. It proposes "small group" workshop methods for elucidating expert knowledge and analyses the areas of agreement and disagreement between experts. The model was developed for four selected ES and for assessing the consequences of management options relating to no-change, riparian management, and decreasing or increasing livestock numbers. Compared with no-change, riparian management and a decrease in livestock numbers improved the ES investigated to varying degrees. Sensitivity analysis of the expert information in the BBN showed the greatest disagreements between experts were mainly for low probability situations and thus had little impact on the results. Conversely, in our applications, the best agreement between experts tended to occur for the higher probability, more likely, situations. This has implications for the practical use of this type of model to support catchment management decisions. The complexity of the relationship between management measures, the water quality and ecological responses and resulting changes in ES must not be a barrier to making decisions in the present time. The interactions of multiple stressors further complicate the situation. However, management decisions typically relate to the overall character of solutions and not their detailed design, which can follow once the nature of the solution has been chosen, for example livestock management or riparian measures or both.

Lessons from COVID-19 for managing transboundary climate risks and building resilience

COVID-19 has revealed how challenging it is to manage global, systemic and compounding crises. Like COVID-19, climate change impacts, and maladaptive responses to them, have potential to disrupt societies at multiple scales via networks of trade, finance, mobility and communication, and to impact hardest on the most vulnerable. However, these complex systems can also facilitate resilience if managed effectively. This review aims to distil lessons related to the transboundary management of systemic risks from the COVID-19 experience, to inform climate change policy and resilience building. Evidence from diverse fields is synthesised to illustrate the nature of systemic risks and our evolving understanding of resilience. We describe research methods that aim to capture systemic complexity to inform better management practices and increase resilience to crises. Finally, we recommend specific, practical actions for improving transboundary climate risk management and resilience building. These include mapping the direct, cross-border and cross-sectoral impacts of potential climate extremes, adopting adaptive risk management strategies that embrace heterogenous decision-making and uncertainty, and taking a broader approach to resilience which elevates human wellbeing, including societal and ecological resilience.

Causal Bayesian networks in assessments of wildfire risks: Opportunities for ecological risk assessment and management

Wildfire risks and losses have increased over the last 100 years, associated with population expansion, land use and management practices, and global climate change. While there have been extensive efforts at modeling the probability and severity of wildfires, there have been fewer efforts to examine causal linkages from wildfires to impacts on ecological receptors and critical habitats. Bayesian networks are probabilistic tools for graphing and evaluating causal knowledge and uncertainties in complex systems that have seen only limited application to the quantitative assessment of ecological risks and impacts of wildfires. Here, we explore opportunities for using Bayesian networks for assessing wildfire impacts to ecological systems through levels of causal representation and scenario examination. Ultimately, Bayesian networks may facilitate understanding the factors contributing to ecological impacts, and the prediction and assessment of wildfire risks to ecosystems. Integr Environ Assess Manag 2021;17:1168-1178. Published 2021. This article is a U.S. Government work and is in the public domain in the USA.

A roadmap towards quantitative cumulative impact assessments: Every step of the way

Currently most Cumulative Impacts Assessments (CIAs) are risk-based approaches that assess the potential impact of human activities and their pressures on the ecosystem thereby compromising the achievement of policy objectives. While some of these CIAs apply actual data (usually spatial distributions) they often have to rely on categorical scores based on expert judgement if they actually assess impact which is often expressed as a relative measure that is difficult to interpret in absolute terms. Here we present a first step-wise approach to conduct a fully quantitative CIA based on the selection and subsequent application of the best information available. This approach systematically disentangles risk into its exposure and effect components that can be quantified using known ecological information, e.g. spatial distribution of pressures or species, pressure-state relationships and population dynamics models with appropriate parametrisation, resulting in well-defined assessment endpoints that are meaningful and can be easily communicated to the recipients of advice. This approach requires that underlying assumptions and methodological considerations are made explicit and translated into a measure of confidence. This transparency helps to identify the possible data-handling or methodological decisions and shows the resulting improvement through its confidence assessment of the applied information and hence the resulting accuracy of the CIA. To illustrate this approach, we applied it in a North Sea CIA focussing on two sectors, i.e. fisheries and offshore windfarms, and how they impact the ecosystem and its components, i.e. seabirds, seabed habitats and marine mammals through various pressures. The results provide a "proof of concept" for this generic approach as well as rigorous definitions of several of the concepts often used as part of risk-based approaches, e.g. exposure, sensitivity, vulnerability, and how these can be estimated using actual data. As such this widens the scope for increasingly more quantitative CIAs using the best information available.

Causal Approach to Determining the Environmental Risks of Seabed Mining

Mineral deposits containing commercially exploitable metals are of interest for seabed mineral extraction in both the deep sea and shallow sea areas. However, the development of seafloor mining is underpinned by high uncertainties on the implementation of the activities and their consequences for the environment. To avoid unbridled expansion of maritime activities, the environmental risks of new types of activities should be carefully evaluated prior to permitting them, yet observational data on the impacts is mostly missing. Here, we examine the environmental risks of seabed mining using a causal, probabilistic network approach. Drawing on a series of expert interviews, we outline the cause-effect pathways related to seabed mining activities to inform quantitative risk assessments. The approach consists of (1) iterative model building with experts to identify the causal connections between seabed mining activities and the affected ecosystem components and (2) quantitative probabilistic modeling. We demonstrate the approach in the Baltic Sea, where seabed mining been has tested and the ecosystem is well studied. The model is used to provide estimates of mortality of benthic fauna under alternative mining scenarios, offering a quantitative means to highlight the uncertainties around the impacts of mining. We further outline requirements for operationalizing quantitative risk assessments in data-poor cases, highlighting the importance of a predictive approach to risk identification. The model can be used to support permitting processes by providing a more comprehensive description of the potential environmental impacts of seabed resource use, allowing iterative updating of the model as new information becomes available.

An application of Bayesian Belief Networks to assess management scenarios for aquaculture in a complex tropical lake system in Indonesia

A Bayesian Belief Network, validated using past observational data, is applied to conceptualize the ecological response of Lake Maninjau, a tropical lake ecosystem in Indonesia, to tilapia cage farms operating on the lake and to quantify its impacts to assist decision making. The model captures ecosystem services trade-offs between cage farming and native fish loss. It is used to appraise options for lake management related to the minimization of the impacts of the cage farms. The constructed model overcomes difficulties with limited data availability to illustrate the complex physical and biogeochemical interactions contributing to triggering mass fish kills due to upwelling and the loss in the production of native fish related to the operation of cage farming. The model highlights existing information gaps in the research related to the management of the farms in the study area, which is applicable to other tropical lakes in general. Model results suggest that internal phosphorous loading (IPL) should be recognized as one of the primary targets of the deep eutrophic tropical lake restoration efforts. Theoretical and practical contributions of the model and model expansions are discussed. Short- and longer-term actions to contribute to a more sustainable management are recommended and include epilimnion aeration and sediment capping.

A Bayesian network to simulate macroinvertebrate responses to multiple stressors in lowland streams

Aquatic ecosystems are affected by multiple environmental stressors across spatial and temporal scales. Yet the nature of stressor interactions and stressor-response relationships is still poorly understood. This hampers the selection of appropriate restoration measures. Hence, there is a need to understand how ecosystems respond to multiple stressors and to unravel the combined effects of the individual stressors on the ecological status of waterbodies. Models may be used to relate responses of ecosystems to environmental changes as well as to restoration measures and thus provide valuable tools for water management. Therefore, we aimed to develop and test a Bayesian Network (BN) for simulating the responses of stream macroinvertebrates to multiple stressors. Although the predictive performance may be further improved, the developed model was shown to be suitable for scenario analyses. For the selected lowland streams, an increase in macroinvertebrate-based ecological quality (EQR) was predicted for scenarios where the streams were relieved from single and multiple stressors. Especially a combination of measures increasing flow velocity and enhancing the cover of coarse particulate organic matter showed a significant increase in EQR compared to current conditions. The use of BNs was shown to be a promising avenue for scenario analyses in stream restoration management. BNs have the capacity for clear visual communication of model dependencies and the uncertainty associated with input data and results and allow the combination of multiple types of knowledge about stressor-effect relations. Still, to make predictions more robust, a deeper understanding of stressor interactions is required to parametrize model relations. Also, sufficient training data should be available for the water type of interest. Yet, the application of BNs may now already help to unravel the contribution of individual stressors to the combined effect on the ecological quality of water bodies, which in turn may aid the selection of appropriate restoration measures that lead to the desired improvements in macroinvertebrate-based ecological quality.

Parameterization Framework and Quantification Approach for Integrated Risk and Resilience Assessments

A growing challenge for risk, vulnerability, and resilience assessment is the ability to understand, characterize, and model the complexities of our joint socioecological systems, often delineated with differing natural (e.g., watershed) and imposed (e.g., political) boundaries at the landscape scale. To effectively manage such systems in the increasingly dynamic, adaptive context of environmental change, we need to understand not just food web interactions of contaminants or the flooding impacts of sea level rise and storm surges, but rather the interplay between social and ecological components within the inherent and induced feedforward and feedback system mechanisms. Risk assessment, in its traditional implementation, is a simplification of a complex problem to understand the basic cause-and-effect relationships within a system. This approach allows risk assessors to distill a complex issue into a manageable model that quantifies, or semiquantifies, the effects of an adverse stressor. Alternatively, an integrated risk and resilience assessment moves toward a solution-based assessment with the incorporation of adaptive management practices as 1 of 4 parts of system resilience (i.e., prepare, absorb, recover, and adapt), and directly considers the complexities of the systems being modeled. We present the Multilevel Risk and Resilience Assessment Parameterization framework for the systematic parameterization and deconstruction of management objectives and goals into assessment metrics and quantifiable risk measurement metrics and complementary resilience measurement metrics. As a proof-of-concept, the presented framework is paired with the Bayesian Network-Relative Risk Model for a human-focused subset of a larger risk and resilience assessment of climate change impacts within the Charleston Harbor Watershed of South Carolina. This new parameterization framework goes beyond traditional simplification and embraces the complexity of the system as a whole, which is necessary for a more representative analysis of an open, dynamic complex system. Integr Environ Assess Manag 2021;17:131-146. © 2020 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

"This Is What We Don't Know": Treating Epistemic Uncertainty in Bayesian Networks for Risk Assessment

Failing to communicate current knowledge limitations, that is, epistemic uncertainty, in environmental risk assessment (ERA) may have severe consequences for decision making. Bayesian networks (BNs) have gained popularity in ERA, primarily because they can combine variables from different models and integrate data and expert judgment. This paper highlights potential gaps in the treatment of uncertainty when using BNs for ERA and proposes a consistent framework (and a set of methods) for treating epistemic uncertainty to help close these gaps. The proposed framework describes the treatment of epistemic uncertainty about the model structure, parameters, expert judgment, data, management scenarios, and the assessment's output. We identify issues related to the differentiation between aleatory and epistemic uncertainty and the importance of communicating both uncertainties associated with the assessment predictions (direct uncertainty) and the strength of knowledge supporting the assessment (indirect uncertainty). Probabilities, intervals, or scenarios are expressions of direct epistemic uncertainty. The type of BN determines the treatment of parameter uncertainty: epistemic, aleatory, or predictive. Epistemic BNs are useful for probabilistic reasoning about states of the world in light of evidence. Aleatory BNs are the most relevant for ERA, but they are not sufficient to treat epistemic uncertainty alone because they do not explicitly express parameter uncertainty. For uncertainty analysis, we recommend embedding an aleatory BN into a model for parameter uncertainty. Bayesian networks do not contain information about uncertainty in the model structure, which requires several models. Statistical models (e.g., hierarchical modeling outside the BNs) are required to consider uncertainties and variability associated with data. We highlight the importance of being open about things one does not know and carefully choosing a method to precisely communicate both direct and indirect uncertainty in ERA. Integr Environ Assess Manag 2021;17:221-232. © 2020 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

Increased Use of Bayesian Network Models Has Improved Environmental Risk Assessments

Environmental and ecological risk assessments are defined as the process for evaluating the likelihood that the environment may be impacted as a result of exposure to stressors. Although this definition implies the calculation of probabilities, risk assessments traditionally rely on nonprobabilistic methods such as calculation of a risk quotient. Bayesian network (BN) models are a tool for probabilistic and causal modeling, increasingly used in many fields of environmental science. Bayesian networks are defined as directed acyclic graphs where the causal relationships and the associated uncertainty are quantified in conditional probability tables. Bayesian networks inherently incorporate uncertainty and can integrate a variety of information types, including expert elicitation. During the last 2 decades, there has been a steady increase in reports on BN applications in environmental risk assessment and management. At recent annual meetings of the Society of Environmental Toxicology and Chemistry (SETAC) North America and SETAC Europe, a number of applications of BN models were presented along with new theoretical developments. Likewise, recent meetings of the European Geosciences Union (EGU) have dedicated sessions to Bayesian modeling in relation to water quality. This special series contains 10 articles based on presentations in these sessions, reflecting a range of BN applications to systems, ranging from cells and populations to watersheds and national scale. The articles report on recent progress in many topics, including climate and management scenarios, ecological and socioeconomic endpoints, machine learning, diagnostic inference, and model evaluation. They demonstrate that BNs can be adapted to established conceptual frameworks used to support environmental risk assessments, such as adverse outcome pathways and the relative risk model. The contributions from EGU demonstrate recent advancements in areas such as spatial (geographic information system [GIS]-based) and temporal (dynamic) BN modeling. In conclusion, this special series supports the prediction that increased use of Bayesian network models will improve environmental risk assessments. Integr Environ Assess Manag 2021;17:53-61. © 2020 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

Quantification of an Adverse Outcome Pathway Network by Bayesian Regression and Bayesian Network Modeling

The adverse outcome pathway (AOP) framework has gained international recognition as a systematic approach linking mechanistic processes to toxicity endpoints. Nevertheless, successful implementation into risk assessments is still limited by the lack of quantitative AOP models (qAOPs) and assessment of uncertainties. The few published qAOP models so far are typically based on data-demanding systems biology models. Here, we propose a less data-demanding approach for quantification of AOPs and AOP networks, based on regression modeling and Bayesian networks (BNs). We demonstrate this approach with the proposed AOP #245, "Uncoupling of photophosphorylation leading to reduced ATP production associated growth inhibition," using a small experimental data set from exposure of Lemna minor to the pesticide 3,5-dichlorophenol. The AOP-BN reflects the network structure of AOP #245 containing 2 molecular initiating events (MIEs), 3 key events (KEs), and 1 adverse outcome (AO). First, for each dose-response and response-response (KE) relationship, we quantify the causal relationship by Bayesian regression modeling. The regression models correspond to dose-response functions commonly applied in ecotoxicology. Secondly, we apply the fitted regression models with associated uncertainty to simulate 10 000 response values along the predictor gradient. Thirdly, we use the simulated values to parameterize the conditional probability tables of the BN model. The quantified AOP-BN model can be run in several directions: 1) prognostic inference, run forward from the stressor node to predict the AO level; 2) diagnostic inference, run backward from the AO node; and 3) omnidirectionally, run from the intermediate MIEs and/or KEs. Internal validation shows that the AOP-BN can obtain a high accuracy rate, when run is from intermediate nodes and when a low resolution is acceptable for the AO. Although the performance of this AOP-BN is limited by the small data set, our study demonstrates a proof-of-concept: the combined use of Bayesian regression modeling and Bayesian network modeling for quantifying AOPs. Integr Environ Assess Manag 2021;17:147-164. © 2020 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).

The Origin, Development, Application, Lessons Learned, and Future Regarding the Bayesian Network Relative Risk Model for Ecological Risk Assessment

In 2012, a regional risk assessment was published that applied Bayesian networks (BN) to the structure of the relative risk model. The original structure of the relative risk model (RRM) was published in the late 1990s and developed during the next decade. The RRM coupled with a Monte Carlo analysis was applied to calculating risk to a number of sites and a variety of questions. The sites included watersheds, terrestrial systems, and marine environments and included stressors such as nonindigenous species, effluents, pesticides, nutrients, and management options. However, it became apparent that there were limits to the original approach. In 2009, the relative risk model was transitioned into the structure of a BN. Bayesian networks had several clear advantages. First, BNs innately incorporated categories and, as in the case of the relative risk model, ranks to describe systems. Second, interactions between multiple stressors can be combined using several pathways and the conditional probability tables (CPT) to calculate outcomes. Entropy analysis was the method used to document model sensitivity. As with the RRM, the method has now been applied to a wide series of sites and questions, from forestry management, to invasive species, to disease, the interaction of ecological and human health endpoints, the flows of large rivers, and now the efficacy and risks of synthetic biology. The application of both methods have pointed to the incompleteness of the fields of environmental chemistry, toxicology, and risk assessment. The low frequency of exposure-response experiments and proper analysis have limited the available outputs for building appropriate CPTs. Interactions between multiple chemicals, landscape characteristics, population dynamics and community structure have been poorly characterized even for critical environments. A better strategy might have been to first look at the requirements of modern risk assessment approaches and then set research priorities. Integr Environ Assess Manag 2021;17:79-94. © 2020 SETAC.