ARPEGES: A Bayesian Belief Network to Assess the Risk of Pesticide Contamination for the River Network of France. (3rd November 2020)

• Record Type: Journal Article
• Authors: Piffady, Jeremy; Carluer, Nadia; Gouy, Veronique; le Henaff, Guy; Tormos, Thierry; Bougon, Nolwenn; Adoir, Emilie; Mellac, Katell
• Is Part Of: Integrated environmental assessment and management. Volume 17:Number 1(2021)
• Page Start: 188
• Page End: 201
• Publication Date: 2020-11-03
• Subjects: Expert Bayesian belief network -- Contamination potential -- Pesticide environmental behavior -- Flow pathways -- Confidence index; Environmental management -- Periodicals; Pollution -- Periodicals; Environmental toxicology -- Periodicals; Environmental risk assessment -- Periodicals; Environmental impact analysis -- Periodicals; 628
• DOI: 10.1002/ieam.4343
• Languages: English
• ISSNs: 1551-3777

Abstract

ABSTRACT: Pesticides are priority concerns in aquatic risk assessment due to their widespread use, ongoing development of new molecules, and potential effects from short‐ and long‐term exposures to aquatic life. Water quality assessments are also challenged by contrasting pesticide behaviors (e.g., mobility, half‐life time, solubility) in different environmental contexts. Furthermore, monitoring networks are not well adapted to the pesticide media transfer dynamics and therefore fail at providing a reliable assessment of pesticides. We present here a Bayesian belief network that was developed in a cooperative process between researchers specializing in Bayesian modeling, soil sciences, agronomy, and diffuse pollutants to provide a tool for stakeholders to assess surface water contamination by pesticides. It integrates knowledge on dominant transfer pathways according to basin physical context and climate for different pesticides properties, such as half‐life duration and affinity to organic C, to develop an assessment of risks of contamination for every watershed in France. The resulting model, ARPEGES (Analyse de Risque PEsticide pour la Gestion des Eaux de Surface; trans. Risk analysis of contamination by pesticides for surface water management), was developed in R. A user‐friendly R interface was built to enable stakeholders to not only obtain ARPEGES' results, but also freely use it to test management scenarios. Though it is applicable to any chemical, its results are illustrated for S‐Metolachlor, a pesticide that was widely used on cereals crops worldwide. In addition to providing contamination potential, ARPEGES also provides a way to diagnose its main explaining factors, enabling stakeholders to focus efforts in the most potentially affected basins, but also on the most probable cause of contamination. In this context, the Bayesian belief network allowed us to use information at different scales (i.e., regional contexts for climate, pedology at the basin scale, pesticide use at the municipality scale) to provide an expert assessment of the processes driving pesticide contamination of streams and the associated uncertainties. Integr Environ Assess Manag 2021;17:188–201. © 2020 SETAC KEY POINTS: We present a Bayesian belief network, developed in a cooperative process between researchers specializing in Bayesian modelling, soil sciences, agronomy, and diffuse pollutants, to provide a tool for stakeholders to assess surface water contamination by pesticides for every watershed in France. It integrates knowledge on dominant transfer pathways according to basin physical context and climate for different pesticides properties, such as half‐life duration and affinity to organic C, to develop an assessment of potential of contamination at the watershed scale. By considering the catchment vulnerability to transfers, ARPEGES is the first process‐based model for pesticides, applicable at a national scale and filling the gap between numerical plot scale water transfers models and large scale indicators needed by stakeholders for management sakes. Taking advantage of the Bayesian principles, ARPEGES also provides an explicit measurement of the confidence associated to its predictions.