Using Bayesian decision networks to guide restoration of freshwater mussels in Illinois: a step-by-step guide to creating and using BDNs for ecological applications

Research output: Book/Report/Conference proceedingTechnical report

Abstract

We highlight the use of Bayesian Decision Networks to formalize the decision process and suggest a management strategy for restoring Ellipse and Spike to target areas. A BDN is particularly useful in complicated situations like this, because it allows for the combination of prior knowledge of mussel distributions and habitat relationships in Illinois with expected value of management outcomes. To build the Bayesian Decision Network, we used long term mussel presence data paired with a suite of environmental and biotic variables to elucidate important factors for each focal species and structure preliminary models (Chiavacci et al. 2018). We then built multiple versions for each species using three levels of information 1) data subset (target streams, non-target streams, or both; Figure 9), 2) expert opinion values (median, minimum, or maximum), and 3) precision of mussel data (long term presence, 2018 presence or 2018 density). All model versions were compared using sensitivity analyses to determine sources of potential model performance bias and decide whether a need for quantitative mussel density sampling in future model iterations was needed. All models built in this project were created using Netica by Norsys Software Corp., a program specifically designed to create Bayesian networks. Netica is available for download for 285 dollars for an individual application, or for 600 dollars for commercial applications (as of 2019).The 7 following walk-through serves as a step-by-step tutorial of how to build BDNs using Netica, while also detailing the methods and results of the Ellipse and Spike models created for this project. For each step, we outline a “General Description”, which is a broad description for managers to consider for their own projects, and “Applied Project Result” is a detailed explanation of the process completed for this project.We highlight the use of Bayesian Decision Networks to formalize the decision process and suggest a management strategy for restoring Ellipse and Spike to target areas. A BDN is particularly useful in complicated situations like this, because it allows for the combination of prior knowledge of mussel distributions and habitat relationships in Illinois with expected value of management outcomes. To build the Bayesian Decision Network, we used long term mussel presence data paired with a suite of environmental and biotic variables to elucidate important factors for each focal species and structure preliminary models (Chiavacci et al. 2018). We then built multiple versions for each species using three levels of information 1) data subset (target streams, non-target streams, or both; Figure 9), 2) expert opinion values (median, minimum, or maximum), and 3) precision of mussel data (long term presence, 2018 presence or 2018 density). All model versions were compared using sensitivity analyses to determine sources of potential model performance bias and decide whether a need for quantitative mussel density sampling in future model iterations was needed. All models built in this project were created using Netica by Norsys Software Corp., a program specifically designed to create Bayesian networks. Netica is available for download for 285 dollars for an individual application, or for 600 dollars for commercial applications (as of 2019).The 7 following walk-through serves as a step-by-step tutorial of how to build BDNs using Netica, while also detailing the methods and results of the Ellipse and Spike models created for this project. For each step, we outline a “General Description”, which is a broad description for managers to consider for their own projects, and “Applied Project Result” is a detailed explanation of the process completed for this project.
Original languageEnglish (US)
PublisherIllinois Natural History Survey
StatePublished - Dec 6 2019

Publication series

NameINHS Technical Report 2019 (29)
No.29

Keywords

  • INHS

Fingerprint

Dive into the research topics of 'Using Bayesian decision networks to guide restoration of freshwater mussels in Illinois: a step-by-step guide to creating and using BDNs for ecological applications'. Together they form a unique fingerprint.

Cite this