Estimating Thresholds for Good Status in Marine Ecosystem Management

Abstract

International and national marine legislations are currently in place to improve, maintain, and restore the condition of marine biodiversity using ecosystem-based management (e.g., Marine Strategy Framework Directive (MSFD), UK Marine Strategy (UKMS), Nature Restoration Regulation (NRR)). Thresholds for good status are used to distinguish between good and degraded ecosystem states and are required under several legislations to assess ecosystem condition (e.g., MSFD: Good Environmental Status; NRR: good condition). However, good-state thresholds are currently not available for many ecosystem components, preventing the full operationalisation of condition assessments and the implementation of ecosystem-based management. This PhD aimed to evaluate quantitative and qualitative methods for estimating thresholds for good status to evaluate the robustness and applicability of approaches, and to outline operational frameworks for decision-makers and managers which could be used to estimate thresholds within ecosystem assessments.
Two groups of quantitative approaches for setting thresholds, either using ecosystem state data across a gradient of increasing pressure (‘pressure-state’) or long-term monitoring from minimally impacted areas (‘reference conditions’) demonstrated key differences in statistical robustness within a simulation study. Methods using pressure-state data could not reliably estimate thresholds with moderate levels of noise and small sample sizes as this precluded the fitting of statistical relationships, whereas methods using reference condition data estimated thresholds for all datasets and responded predictably to varying data. As a result, reference condition methods, ‘Range of natural variation’ (RNV) and ‘Statistically detectable change’ (SDC), were considered more robust, transparent, and likely to estimate thresholds for noisy, small-sample sized monitoring data. Both methods were applied to empirical reference condition data from seagrass, plankton, tropical reef fish, and benthic invertebrate species groups and estimated thresholds for all species, whilst explicitly defining an associated probability of being in good status. Thresholds range from 0-1 as they are a relative proportion of a reference state mean (i.e., 1). RNV thresholds ranged from 0.40-0.85, and SDC thresholds from 0.65-0.94 (with 50% probability of being in good state). Thresholds were modelled against environmental gradients (e.g., depth, temperature) to demonstrate how general relationships can be used to predict thresholds for regions where minimally impacted reference data may not available (e.g., thresholds for benthic invertebrate biomass were predicted using depth and maximum average lifespan of the community). These methods were also applied to estimate upper thresholds, which can identify increases in species abundance outside of natural ranges that may be causing increased predation or competition on other species, resulting in an overall degraded state. Upper thresholds are a proportion of the reference state but range from 1 to infinity. RNV thresholds ranged from 1.19-2.98 and SDC thresholds from 1.08-1.83 (50% probability good state). The envelope of variation used to estimate thresholds was wider under RNV compared to SDC, which resulted in lower minimum and higher maximum thresholds. A qualitative, structured expert elicitation approach (IDEA protocol) was applied and resulted in thresholds with high levels of variation. Average biomass thresholds ranged from 0.38-0.61 and extent thresholds from 0.67-0.81. This variation was partially explained by two different expert interpretations of good status. This transparent, systematic expert judgment approach presents a more objective, replicable method to estimate thresholds than unstructured elicitation approaches, but differences in interpretations of key terms (i.e., what is good) require a level of consideration to ensure thresholds are ecologically meaningful and transparent. Biomass thresholds estimated using structured elicitation were on average lower than those estimated using quantitative methods, suggesting that there are differences in the way in which these two approaches define ‘good’ status, at least in their application here.
This thesis outlined differences in strengths, weaknesses, and applicability of all evaluated methods in certain contexts. From conceptual, statistical, and empirical evaluations, ‘range of natural variation’ was highlighted as the most transparent, robust, ecologically meaningful and operational approach for estimating thresholds for ecologically focused objectives of good condition. A framework was outlined on how this systems-, indicator-, and pressure-independent method could be applied in practice, to estimate thresholds with an associated probability of being in good status, and to up-scale these to relevant local or regional spatial assessment units. These results provide a pathway towards enabling more complete assessments of the condition of marine environments and the implementation of ecosystem-based management practices to improve, maintain, and restore marine biodiversity.

Date of Award	3 Sept 2026
Original language	English
Awarding Institution	Bangor University
Sponsors	Natural Environment Research Council (NERC) & NIRAS
Supervisor	Jan Geert Hiddink (Supervisor), Abigail McQuatters-Gollop (Supervisor), Tomas Chaigneau (Supervisor), Sebastian Valanko (Supervisor), Cristina Vina-Herbon (Supervisor), Sally Kazer (Supervisor) & Ian Gloyne-Phillips (Supervisor)