Editorial
10 January 2023
Marie Anne Eurie Forio
, Elena Moltchanova
and
Peter L. M. Goethals
Editorial: Application of Bayesian modeling in environmental management
Editors
3
Impact
Loading...
Original Research
03 November 2022
Runwei Li
and
Jacqueline MacDonald Gibson
In the past two decades, global manufacturing of per- and polyfluoroalkyl substances (PFAS) has shifted from long-chain compounds to short-chain alternatives in response to evidence of the health hazards of long-chain formulations. However, accumulating data indicate that short-chain PFAS also pose health risks and are highly mobile and persistent in the environment. Because short-chain PFAS are relatively new chemicals, comprehensive knowledge needed to predict their environmental fate is lacking. This study evaluated the capacity of machine-learned Bayesian networks (BNs) to predict risks of exposure to short-chain PFAS in a Minnesota region affected by PFAS releases from the 3M Cottage Grove facility. Models were trained using long-term monitoring data provided by the Minnesota Department of Health (n = 12,406), which we coupled to a comprehensive dataset created by curating 88 other variables that describe potential PFAS sources, soil and hydrogeologic characteristics, and land use. Model performance was assessed using the area under the receiver-operating characteristic curve (AUC), a common measure of the accuracy of machine-learned classification algorithms. In addition, exposure risks were visualized spatially by coupling model predictions to a geographic information system. We found that machine-learned BN models had robust predictive performance, with AUCs above 0.96 in cross-validation. Significant risk factors identified by the BNs include distance to the 3M factory, distance to a former landfill, and areal extent of wetlands and developed land. We also found that risks of exposure to and the areal extent of perfluorosulfonic acids were greater than for perfluorocarboxylic acids with the same carbon number. The results suggest that machine-learned BNs could provide a promising screening tool for assessing short-chain PFAS exposure risks in groundwater.
Perspective
11 November 2022
Andrée De Cock
, 2 more and
Peter L. M. Goethals
2,437 views
5 citations
Original Research
13 October 2022
A systems approach to modelling phosphorus pollution risk in Scottish rivers using a spatial Bayesian Belief Network helps targeting effective mitigation measures
Miriam Glendell
, 5 more and
Malcolm Coull
Water quality remains a main reason for the failure of waterbodies to reach Good Ecological Status (GES) under the European Union Water Framework Directive (WFD), with phosphorus (P) pollution being a major cause of water quality failures. Reducing P pollution risk in agricultural catchments is challenging due to the complexity of biophysical drivers along the source-mobilisation-delivery-impact continuum. While there is a need for place-specific interventions, the evidence supporting the likely effectiveness of mitigation measures and their spatial targeting is uncertain. We developed a decision-support tool using a Bayesian Belief Network that facilitates system-level thinking about P pollution and brings together academic and stakeholder communities to co-construct a model appropriate to the region of interest. The expert-based causal model simulates the probability of soluble reactive phosphorus (SRP) concentration falling into the WFD high/good or moderate/poor status classifications along with the effectiveness of three mitigation measures including buffer strips, fertiliser input reduction and septic tank management. In addition, critical source areas of pollution are simulated on 100 × 100 m raster grids for seven catchments (12–134 km2) representative of the hydroclimatic and land use intensity gradients in Scotland. Sensitivity analysis revealed the importance of fertiliser inputs, soil Morgan P, eroded SRP delivery rate, presence/absence of artificial drainage and soil erosion for SRP losses from diffuse sources, while the presence/absence of septic tanks, farmyards and the design size of sewage treatment works were influential variables related to point sources. Model validation confirmed plausible model performance as a “fit for purpose” decision support tool. When compared to observed water quality data, the expert-based causal model simulated a plausible probability of GES, with some differences between study catchments. Reducing fertiliser inputs below optimal agronomic levels increased the probability of GES by 5%, while management of septic tanks increased the probability of GES by 8%. Conversely, implementation of riparian buffers did not have an observable effect on the probability of GES at the catchment outlet. The main benefit of the approach was the ability to integrate diverse, and often sparse, information; account for uncertainty and easily integrate new data and knowledge.
Original Research
03 October 2022
Steven F. Wilson
, 2 more and
Daniel A. Guertin
1,538 views
1 citations
Original Research
15 September 2022
Probabilistic risk assessment of pesticides under future agricultural and climate scenarios using a bayesian network
Sophie Mentzel
, 4 more and
S. Jannicke Moe
The use of Bayesian networks (BN) for environmental risk assessment has increased in recent years as they offer a more transparent way to characterize risk and evaluate uncertainty than the traditional risk assessment paradigms. In this study, a novel probabilistic approach applying a BN for risk calculation was further developed and explored by linking the calculation a risk quotient to alternative future scenarios. This extended version of the BN model uses predictions from a process-based pesticide exposure model (World Integrated System for Pesticide Exposure - WISPE) in the exposure characterization and toxicity test data in the effect characterization. The probability distributions for exposure and effect are combined into a risk characterization (i.e. the probability distribution of a risk quotient), a common measure of the exceedance of an environmentally safe exposure threshold. The BN model was used to account for variabilities of the predicted pesticide exposure in agricultural streams, and inter-species variability in sensitivity to the pesticide among freshwater species. In Northern Europe, future climate scenarios typically predict increased temperature and precipitation, which can be expected to cause an increase in weed infestations, plant disease and insect pests. Such climate-related changes in pest pressure in turn can give rise to altered agricultural practices, such as increased pesticide application rates, as an adaptation to climate change. The WISPE model was used to link a set of scenarios consisting of two climate models, three pesticide application scenarios and three periods (year ranges), for a case study in South-East Norway. The model was set up for the case study by specifying environmental factors such as soil properties and field slope together with chemical properties of pesticides to predict the pesticide exposure in streams adjacent to the agricultural fields. The model was parameterized and evaluated for five selected pesticides: the three herbicides clopyralid, fluroxypyr-meptyl, and 2-(4-chloro-2-methylphenoxy) acetic acid (MCPA), and the two fungicides prothiocanzole and trifloxystrobin. This approach enabled the calculation and visualization of probability distribution of the risk quotients for the future time horizons 2050 and 2085. The risk posed by the pesticides were in general low for this case study, with highest probability of the risk quotient exceeding 1 for the two herbicides fluroxypyr-meptyl and MCPA. The future climate projections used here resulted in only minor changes in predicted exposure concentrations and thereby future risk. However, a stronger increase in risk was predicted for the scenarios with increased pesticide application, which can represent an adaptation to a future climate with higher pest pressures. In the current study, the specific BN model predictions were constrained by an existing set of climate projections which represented only one IPCC scenario (A1B) and two climate models. Further advancement of the BN modelling demonstrated herein, including more recent climate scenarios and a larger set of climate models, is anticipated to result in more relevant risk characterization also for future climate conditions. This probabilistic approach will have the potential to aid targeted management of ecological risks in support of future research, industry and regulatory needs.