Robust Decision Analysis under Severe Uncertainty and Ambiguous Tradeoffs: An Invasive Species Case Study.

Bayesian decision analysis is a useful method for risk management decisions, but is limited in its ability to consider severe uncertainty in knowledge, and value ambiguity in management objectives. We study the use of robust Bayesian decision analysis to handle problems where one or both of these issues arise. The robust Bayesian approach models severe uncertainty through bounds on probability distributions, and value ambiguity through bounds on utility functions. To incorporate data, standard Bayesian updating is applied on the entire set of distributions. To elicit our expert's utility representing the value of different management objectives, we use a modified version of the swing weighting procedure that can cope with severe value ambiguity. We demonstrate these methods on an environmental management problem to eradicate an alien invasive marmorkrebs recently discovered in Sweden, which needed a rapid response despite substantial knowledge gaps if the species was still present (i.e., severe uncertainty) and the need for difficult tradeoffs and competing interests (i.e., value ambiguity). We identify that the decision alternatives to drain the system and remove individuals in combination with dredging and sieving with or without a degradable biocide, or increasing pH, are consistently bad under the entire range of probability and utility bounds. This case study shows how robust Bayesian decision analysis provides a transparent methodology for integrating information in risk management problems where little data are available and/or where the tradeoffs are ambiguous.

"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).

Assessment of uncertainty in chemical models by Bayesian probabilities: Why, when, how?

A prediction of a chemical property or activity is subject to uncertainty. Which type of uncertainties to consider, whether to account for them in a differentiated manner and with which methods, depends on the practical context. In chemical modelling, general guidance of the assessment of uncertainty is hindered by the high variety in underlying modelling algorithms, high-dimensionality problems, the acknowledgement of both qualitative and quantitative dimensions of uncertainty, and the fact that statistics offers alternative principles for uncertainty quantification. Here, a view of the assessment of uncertainty in predictions is presented with the aim to overcome these issues. The assessment sets out to quantify uncertainty representing error in predictions and is based on probability modelling of errors where uncertainty is measured by Bayesian probabilities. Even though well motivated, the choice to use Bayesian probabilities is a challenge to statistics and chemical modelling. Fully Bayesian modelling, Bayesian meta-modelling and bootstrapping are discussed as possible approaches. Deciding how to assess uncertainty is an active choice, and should not be constrained by traditions or lack of validated and reliable ways of doing it.

Assessing predictive uncertainty in comparative toxicity potentials of triazoles.

Comparative toxicity potentials (CTPs) quantify the potential ecotoxicological impacts of chemicals per unit of emission. They are the product of a substance's environmental fate, exposure, and hazardous concentration. When empirical data are lacking, substance properties can be predicted. The goal of the present study was to assess the influence of predictive uncertainty in substance property predictions on the CTPs of triazoles. Physicochemical and toxic properties were predicted with quantitative structure-activity relationships (QSARs), and uncertainty in the predictions was quantified with use of the data underlying the QSARs. Degradation half-lives were based on a probability distribution representing experimental half-lives of triazoles. Uncertainty related to the species' sample size that was present in the prediction of the hazardous aquatic concentration was also included. All parameter uncertainties were treated as probability distributions, and propagated by Monte Carlo simulations. The 90% confidence interval of the CTPs typically spanned nearly 4 orders of magnitude. The CTP uncertainty was mainly determined by uncertainty in soil sorption and soil degradation rates, together with the small number of species sampled. In contrast, uncertainty in species-specific toxicity predictions contributed relatively little. The findings imply that the reliability of CTP predictions for the chemicals studied can be improved particularly by including experimental data for soil sorption and soil degradation, and by developing toxicity QSARs for more species.

Experimental assessment of the environmental fate and effects of triazoles and benzotriazole.

The environmental fate and effects of triazoles and benzotriazoles are of concern within the context of chemical regulation. As part of an intelligent testing strategy, experimental tests were performed on endpoints that are relevant for risk assessment. The experimental tests included the assessment of ecotoxicity to an alga, a daphnid and zebrafish embryos, and the assessment of ready biodegradability. Triazole and benzotriazole compounds were selected for testing, based on existing toxicity data for vertebrate and invertebrate species, as well as on the principal component analysis of molecular descriptors aimed at selecting the minimum number of test compounds in order to maximise the chemical domain spanned for both compound classes. The experimental results show that variation in the toxicities of triazoles and benzotriazole across species was relatively minor; in general, the largest factor was approximately 20. The study conducted indicated that triazoles are not readily biodegradable.

A Risk Assessment Perspective of Current Practice in Characterizing Uncertainties in QSAR Regression Predictions.

The European REACH legislation accepts the use of non-testing methods, such as QSARs, to inform chemical risk assessment. In this paper, we aim to initiate a discussion on the characterization of predictive uncertainty from QSAR regressions. For the purpose of decision making, we discuss applications from the perspective of applying QSARs to support probabilistic risk assessment. Predictive uncertainty is characterized by a wide variety of methods, ranging from pure expert judgement based on variability in experimental data, through data-driven statistical inference, to the use of probabilistic QSAR models. Model uncertainty is dealt with by assessing confidence in predictions and by building consensus models. The characterization of predictive uncertainty would benefit from a probabilistic formulation of QSAR models (e.g. generalized linear models, conditional density estimators or Bayesian models). This would allow predictive uncertainty to be quantified as probability distributions, such as Bayesian predictive posteriors, and likelihood-based methods to address model uncertainty. QSAR regression models with point estimates as output may be turned into a probabilistic framework without any loss of validity from a chemical point of view. A QSAR model for use in probabilistic risk assessment needs to be validated for its ability to make reliable predictions and to quantify associated uncertainty.