EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Expert elicitation for parameterisation of a Bayesian Network model designed to simulate Faecal Indicator Organism (FIO) losses from septic tank systems in rural catchments

Chisha Mzyece1, Miriam Glendell2, Richard S Quilliam1, Ian Jones1, Eulyn Pagaling2, Lisa Avery2, and David M Oliver1
Chisha Mzyece et al.
  • 1University of Stirling, Faculty of Natural Sciences, Biological and Environmental Sciences, Stirling, FK9 4LA, United Kingdom (
  • 2James Hutton Institute, Environmental and Biochemical Sciences Group, Aberdeen, AB15 8QH Scotland, United Kingdom (

Bayesian Networks (BNs) are a modelling approach increasingly used in landscape management, e.g., to predict microbial water pollution risk and inform ecological risk assessment. BNs are widely acknowledged for their ability to integrate multiple data types in their structure, including expert knowledge derived through structured elicitation approaches and are therefore, advantageous when empirical evidence or large-scale datasets are scarce. Expert elicitation is a useful technique for quantifying and characterising expert knowledge regarding an uncertain quantity in situations where empirical data are missing, or additional information is required to augment available data. In this study, an expert elicitation approach utilising the Sheffield Elicitation Framework (SHELF) was employed to obtain expert judgements of an uncertain quantity included in a BN model designed to quantify faecal indicator organism (FIO) losses from septic tank systems by modifying an existing phosphorus risk BN model. The aim of the study was to quantify expert judgements on the proportions of FIOs likely to be delivered to a surface watercourse from septic tank systems based on soil hydrological properties, septic tank distance to watercourse and slope. The specific objectives were to:

  • Solicit expert feedback on the structure of the BN conceptual model developed to identify key factors influencing FIO pollution from septic tank systems;
  • Use the SHELF elicitation protocol to obtain individual expert judgements on FIO delivery coefficients in form of percentiles for a series of soil type, slope and distance to watercourse scenarios;
  • Fit probability density curves to individual expert judgements and derive consensus from across the range of expert judgements using facilitated group discussion.

The structure of the BN model including identification and justification of model variables, approaches to expert elicitation and consensus expert judgements are presented. The study demonstrates effective use of expert opinion in BN model parameterisation and BN FIO modelling to inform on options for addressing microbial pollution originating from septic tank systems in the Tarland catchment in North Eastern Scotland.

How to cite: Mzyece, C., Glendell, M., Quilliam, R. S., Jones, I., Pagaling, E., Avery, L., and Oliver, D. M.: Expert elicitation for parameterisation of a Bayesian Network model designed to simulate Faecal Indicator Organism (FIO) losses from septic tank systems in rural catchments, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-7414,, 2023.

Comments on the supplementary material

to access the discussion