Conceptual model (a) and schematic illustration (b) of self-amplifying feedback loops during the regime shift observed in NW Sweden, from a stable vegetated eelgrass state with high light conditions, to the stable sediment state with low light conditions when perennial drift algae dominate. In (a), the green line shows the change in state variables of eelgrass (i.e., areal extent, biomass, shoot density) over time, and green arrows denote a positive feedback with a positive effect on eelgrass (i.e., sediment stabilization having a positive effect on light conditions). Small black arrows denote feedbacks with negative effects on eelgrass (sediment resuspension decreasing light conditions and drift algae shadowing and dislodging eelgrass). The black line shows the change in turbidity and algal cover over time. The red dotted line denotes the threshold in the eelgrass state variables where the feedbacks switch from self-generating positive to negative effect on light and eelgrass growth. The large black arrow indicates an external pulse disturbance (e.g., storm, filamentous algal bloom causing anoxia, dredging activities) decreasing the eelgrass state variables below the threshold and resulting in a regime shift from the eelgrass state to the sediment state with drift algae. In (b), the feedback loops are illustrated during the stable eelgrass state (i), the transient stage after the feedbacks have switch to self-generating negative effect on light and eelgrass growth (ii), and during the stable sediment state with drift algae (iii). Drifting algal mats moving at the bottom increase the resuspension of sediment through physical abrasion, and increased turbidity facilitates for drift algae to enter meadows by decreasing growth and density of eelgrass, resulting in a self-generating interaction between the two feedback mechanisms

Changes in eelgrass distribution 1981–2015. Map showing the distribution of eelgrass in the Marstrand area mapped in the early 1980s, 2000–2004 (Baden et al. 2003; Nyqvist et al. 2009), and 2015 (this study). Colored areas denote areas where eelgrass covered ≥ 5% of the bottom 1981–2015. Areas with eelgrass from later inventories are placed on top of older ones. All areas mapped in 2015 overlap with earlier inventories, and a large majority of the areas mapped in 2000–2004 overlap with areas from 1980s. The enlarge sections with sites 4–5 and 7–8 include areas with the largest losses of eelgrass since 2004