1. Background. Cumulative human pressures can induce non-linear discontinuous dynamics in ecosystems, known as regime shifts. Regime shifts typically imply hysteresis, a lacking or delayed system response when pressures are reverted, which can frustrate restoration efforts.

2. The goal of the study. Here we investigate whether the northern Adriatic Sea fish and macroinvertebrate community, depicted by commercial fishery landings, has undergone irreversible regime shifts over the last 40 years.

3. What was done. We use a stochastic cusp model to show that, under the interactive effect of fishing and warming, the community reorganized through discontinuous changes.

4. What was found. Part of the community has now reached a new stable state, implying that a recovery towards previous baselines might be impossible. Interestingly, total landings remained constant across decades, masking the low resilience of the community.

5. The relevance. Our study reveals the importance of carefully assessing regime shifts and resilience in marine ecosystems and advocates for their inclusion into management.

The database includes landings data and only seafood products caught by the Chioggia’s fishing fleet, operating in the Northern Adriatic Sea. Landing categories represent single or multiple species/taxa and were recorded monthly (Barausse et al., 2011; Mazzoldi et al., 2014). Some categories were removed since they include freshwater or lagoon species, species coming mainly from local extensive aquaculture and/or species whose fishery is regulated (e.g., eel, tunas), and therefore variation in landing is not expected to represent variation in abundance (for a discussion on this matter regarding the Clodia database, see (Barausse et al., 2011; Mazzoldi et al., 2014)). To analyze community changes over time, monthly data were summed to annual totals. Species which were occasionally recorded, with more than 15 zero annual landings, were removed. In total 37 categories, including 92 species, were used to analyze the dynamics of the Northern Adriatic Sea community. Since the time series are fishery-dependent data and span many years in which the fishing capacity differed, landings were divided by fishing capacity (the gross tonnage of the fleet, a proxy for fishing pressure) to make them better indicators of relative biomass. Fishing capacity data were built using two datasets. The first one, including ISTAT (Italian National Institute of Statistics) and unpublished data, spanned from 1951 to 1993 and was reported in Gross Register Tonnage (GRT) (Barausse et al., 2011). The second one, obtained from the EU Fleet Register (https://ec.europe.eu//fisheries/cfp fishing_rules/fishing_fleet_en), contained data since 1991 reported in Gross Tonnage (GT). Since these two units of measurement cannot be directly converted one to another, the data provided in GRT were converted into GT using a conversion factor of 1.076 calculated based on the overlapping values of the two time series. Yearly landings in tonnes were thus divided by yearly Gross Tonnage. This choice reduced the period spanned by the dataset to 1951-2018.