Simple, policy friendly, ecological interaction models from uncertain data and expert opinion. (December 2015)

• Record Type: Journal Article
• Authors: Stafford, Richard; Williams, Rachel L.; Herbert, Roger J.H.
• Is Part Of: Ocean & coastal management. Volume 118:Part A(2016)
• Page Start: 88
• Page End: 96
• Publication Date: 2015-12
• Subjects: Rocky shore -- Predictive model -- Sparse data -- Marine management -- Fisheries -- Bayesian belief network -- Conservation; Marine resources -- Management -- Periodicals; Coastal zone management -- Periodicals; Coastal ecology -- Periodicals; Ressources marines -- Périodiques; Littoral -- Aménagement -- Périodiques; Écologie littorale -- Périodiques; Coastal ecology; Coastal zone management; Marine resources -- Management; Periodicals; Electronic journals; 551.46
• DOI: 10.1016/j.ocecoaman.2015.04.013
• Languages: English
• ISSNs: 0964-5691

Abstract

Abstract: In the marine environment, humans exploit natural ecosystems for food and economic benefit. Challenging policy goals have been set to protect resources, species, communities and habitats, yet ecologists often have sparse data on interactions occurring in the system to assess policy outcomes. This paper presents a technique, loosely based on Bayesian Belief Networks, to create simple models which 1) predict whether individual species within a community will decline or increase in population size, 2) encapsulate uncertainty in the predictions in an intuitive manner and 3) require limited knowledge of the ecosystem and functional parameters required to model it. We develop our model for a UK rocky shore community, to utilise existing knowledge of species interactions for model validation purposes. However, we also test the role of expert opinion, without full scientific knowledge of species interactions, by asking non-UK based marine scientists to derive parameters for the model (non-UK scientists are not familiar with the exact communities being described and will need to extrapolate from existing knowledge in a similar manner to model a poorly studied system). We find these differ little from the parameters derived by ourselves and make little difference to the final model predictions. We also test our model against simple experimental manipulations, and find that the most important changes in community structure as a result of manipulations correspond well to the model predictions with both our, and non-UK expert parameterisation. The simplicity of the model, nature of the outputs, and the user-friendly interface makes it potentially suitable for policy, conservation and management work on multispecies interactions in a wide range of marine ecosystems. Highlights: Marine policy can involve crude measures of population changes (e.g. decreasing). Data on species interactions can be poor and difficult to model. We create models to predict species decline using limited knowledge of interactions. We demonstrate expert groups can parameterise these models. Models correctly predict community changes, verified by experiments.