SEFS14: Early warning signals have limited applicability to empirical lake data

An invited talk given at the European Federation for Freshwater Science's biannual meeting.

Transcript

1. Duncan O’Brien duncan.obrien@bristol.ac.uk duncanobrien.bsky.social https://duncanobrien.github.io
2. Regime shifts duncanobrien.bsky.social “Sudden or abrupt shift in the state of the system resulting from the influence of an external control parameter/driver or by the system’s internal dynamics, where core ecosystem functions, structures and processes are fundamentally changed.” “Sudden or abrupt shift in the state of the system resulting from the influence of an external control parameter/driver or by the system’s internal dynamics, where core ecosystem functions, structures and processes are fundamentally changed.” “Sudden or abrupt shift in the state of the system resulting from the influence of an external control parameter/driver or by the system’s internal dynamics, where core ecosystem functions, structures and processes are fundamentally changed.” Stephen R. Carpenter Thomas Wernberg Low stressor High stressor Low stressor High stressor
3. Regime shifts mechanisms duncanobrien.bsky.social Ball-and-cup heuristic & critical transitions Wright 1931 (Genetics)
4. Regime shifts mechanisms duncanobrien.bsky.social Predictable – critical slowing down Wissel 1984 (Oecologia) Early warning signals - inconsistent Burthe et al. 2016 (J. Appl. Ecol) Gsell et al. 2014 (PNAS) Critical transition/bifurcation
5. Regime shifts are not always critical transitions Regime shifts duncanobrien.bsky.social “Sudden or abrupt shift in the state of the system resulting from the influence of an external control parameter/driver or by the system’s internal dynamics, where core ecosystem functions, structures and processes are fundamentally changed.” A regime shift is not necessarily predictable Dakos et al. 2015 (Proc. R. Soc. B) Hillebrand et al. 2020 (NEE) Davidson et al. 2023 (Nat. Commun)
6. Regime shifts mechanisms duncanobrien.bsky.social Understanding driver dynamics vital for predicting regime shifts Critical slowing down only present prior to bifurcations
7. Regime shifts classification duncanobrien.bsky.social 1. Time series shift? 2. Response to a driver 3. Multimodal distribution? Extension of: • Scheffer & Carpenter 2003 (TREE) • Anderson et al. 2009 (TREE)
8. Results classification duncanobrien.bsky.social • Nine term plankton monitoring programmes • Whole ecosystem coverage • Yearly phyto- and zooplankton time series • Identification of critical transitions • Testing of EWSs
9. Results classification duncanobrien.bsky.social
10. Results early warning signals duncanobrien.bsky.social • Indicator ability rarely better than ‘guessing’ • Combining information improves classification • Machine learning unexpectedly low probabilities
11. Conclusions duncanobrien.bsky.social • Identifying critical transitions is complicated • In complex systems, different transitions possible and critical transitions not ubiquitous • Understanding drivers vital • Early warning signals require ‘best-case’ data • Multivariate > univariate LOCAL >> GENERIC
12. 12 Funding Dr Chris Clements Dr Steve Thackeray Dr Gideon Gal Dr Partha Dutta Smita Deb Dr Ichiro Matsuzaki Data bodies Acknowledgments duncanobrien.bsky.social Acknowledgements duncan.obrien@bristol.ac.uk https://duncanobrien.github.io