Bayesian Network Modelling provides Spatial and Temporal Understanding of Ecosystem Dynamics within Shallow Shelf Seas

Neda Trifonova* (Corresponding Author), Beth Scott, Michela De Dominicis, James J. Waggitt, Judith Wolf

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

6 Citations (Scopus)
8 Downloads (Pure)


Understanding ecosystem dynamics within shallow shelf seas is of great importance to support marine spatial management of natural populations and activities such as fishing and offshore renewable energy production to combat climate change. Given the possibility of future changes, a baseline is needed to predict ecosystems responses to such changes. This study uses Bayesian techniques to find the data-driven estimates of interactions among a set of physical and biological variables and a human pressure within the last 30 years in a well-studied shallow sea (North Sea, UK) with four contrasting regions and their associated ecosystems. A hidden variable is incorporated to model functional ecosystem change, where the underlying interactions dramatically change, following natural or anthropogenic disturbance. Data-driven estimates of interactions were identified, highlighting physical (e.g. bottom temperature, potential energy anomaly) and biological variables (e.g. sandeel larvae, net primary production) to be strong indicators of ecosystem change. There was consistency in the physical and biological variables, identified as good indicators in three of the regions, however the shallower region (with depths < 50 m, that is targeted for static offshore wind developments) was the most dissimilar. The use of contrasting regions provided useful insights on responses linked to ecosystem disturbances and identified the top predators as better indicators for each region, with the harbour porpoise being a particularly valuable indicator of ecosystem change across most regions. Another important finding was the dramatic changes in the strength of many interactions over time. This suggests that physical and biological indicators should only be used with additional temporal information, as changes in strength led to the identification of two potentially significant periods of ecosystem change (after 2005 and after 2010), linked to physical pressures (e.g. cold-water anomalies, seen in bottom temperatures; salinity changes, seen in the potential energy anomaly) and primary production changes. The hidden variable also modelled a change in the early 2000s for all the regions and identified maximum chlorophyll-a and sea surface temperature as some of the better indicators of these ecosystem changes.
Original languageEnglish
Article number107997
Number of pages16
JournalEcological Indicators
Early online date24 Jul 2021
Publication statusPublished - Oct 2021

Bibliographical note

This work was supported by the Supergen Offshore Renewable Energy (ORE) Hub, funded by the Engineering and Physical Sciences Research Council (EPSRC EP/S000747/1) and the NERC/DEFRA funded Marine Ecosystems Research Programme (MERP: NE/L003201/1). The authors would also like to thank the following people for providing data to this study: Debbie Russel, Signe Sveegaard, Mirko Hauswirth, Ruben Fijn, Chelsea Bradbury, Mark Lewis, Steve Geelhoed, Nicolas Vanermen, Oliver Boisseau, Dave Wall, Mark Jessopp, Jared Wilson, Alex Banks, Graham Pierce, Sally Hamilton, Jan Haelters, Suzanne Henderson, Peter Evans, Anita Gilles, Eric Stienen, Paul Thompson, Nicola Hodgins and Andrea Salkeld. For detailed information on their organizations and contacts, please refer to the SI. The authors would like to thank Ella-Sophia Benninghaus (University of Aberdeen) for providing the images in Fig. 6.


  • climate change
  • Hidden variable
  • functional ecosystem change
  • top predator dynamics
  • Fisheries effects


Dive into the research topics of 'Bayesian Network Modelling provides Spatial and Temporal Understanding of Ecosystem Dynamics within Shallow Shelf Seas'. Together they form a unique fingerprint.

Cite this