Predation and spatial connectivity interact to shape ecosystem resilience to an ongoing regime shift

The study investigates how spatial connectivity and local environmental conditions interact to shape resilience to a spatially propagating ecosystem regime shift along the Swedish Baltic Sea coast. The ecosystem is undergoing a shift from dominance by predatory fish (European perch and northern pike) to dominance by three-spined stickleback, driven by predator-prey reversal dynamics and associated food-web cascades. The authors hypothesize that connectivity among predator spawning habitats increases resilience by supporting higher local predator densities via immigration and recolonization, while local drivers such as predation by grey seals and great cormorants, fishing pressure, and spawning-season temperature modulate this effect. They predict positive effects of connectivity (especially under low predation/fishing pressure) and temperature (accelerating predator larval growth and reducing vulnerability to stickleback), and link the stressor gradient to offshore stickleback populations and distance to the open sea.

The paper builds on theory that spatial connectivity can either facilitate the spread of alternative regimes or enhance resilience by supporting species that maintain the original state. Empirical studies on connectivity’s role in regime shifts are limited, though connectivity is widely emphasized in conservation. Prior work in the Baltic Sea indicates strong bimodality between predator- and stickleback-dominated states, predator-prey reversal mechanisms, and cascading effects to lower trophic levels. Movement ecology suggests pike and perch disperse over relatively short coastal distances, while stickleback populations are connected over large spatial scales with seasonal migrations from offshore. Previous studies also indicate that offshore processes (e.g., reduced predation on stickleback) and local factors (habitat degradation, fishing, seal and cormorant predation) have reduced resilience of predatory fish populations.

Data: The study compiles >7000 juvenile fish samples collected in shallow coastal areas along ~680 km of the Swedish Baltic Sea coast (55–60.5°N) during late July–September, 2001–2020, using standardized low-impact explosive sampling. Juveniles of perch (Perca fluviatilis), pike (Esox lucius), and stickleback (Gasterosteus aculeatus) were counted; corrections were applied for floating vs sunken fish and detonation strength to standardize to ~80 m² sampling area. Samples beyond 40 km from the open sea and above 60.5°N were excluded. Juvenile relative dominance mirrors adult dominance during spawning and is used as a proxy for local state. Stressor variables (incoming stickleback): Distance to open sea computed via cost-distance in GIS; wave exposure from the Simplified Wave Model (10×10 m); offshore mature stickleback biomass (≥5.5 cm) from Baltic International Acoustic Survey, distance-weighted to the nearest archipelago-open sea border with a Gaussian kernel emphasizing ≤150 km. Connectivity metrics: Predator spawning habitat delineated by depth ≤3 m and wave exposure thresholds (log10 ≤3.2 and ≤3.5), using high-resolution landmask and raster layers; patches <1 ha removed. Two connectivity measures within a 10 km movement radius (~95% of adult movement): (1) network-based sum of connected habitat areas weighted by distance-dependent link weights estimated from adult perch tagging data (Michaelis-Menten cumulative dispersal function); (2) raster-based distance-weighted sum of suitable habitat cells within 10 km (later used for main plots due to best cross-model performance). Connectivity represents potential dispersal among spawning habitats for adult predators. Top predator predation pressure: Annual grey seal counts (corrected for haul-out detectability and foraging range up to 60 km) and great cormorant colony data (years 2006 and 2012, interpolated across years; foraging range 20 km). Converted to fish extraction (kg km−2 yr−1) using literature-based consumption rates and combined into a 1×1 km annual raster. Fishing pressure: Recreational catches mapped using questionnaire-derived catch rates distributed by population density and travel distance (years 2006, 2010, 2013 averaged); commercial catches at ICES rectangle scale (1999–2015 averaged). Combined into a single raster of kg km−2 yr−1 at 1×1 km resolution (perch-focused, assumed broadly representative for pike spatially). Temperature: Copernicus Baltic Sea L4 daily SST (0.02° grid). Calculated degree-day sums above 10 °C from Jan 1 to start of juvenile sampling at the nearest grid cell. Statistical analyses: Baseline model of relative predator dominance (predator juveniles / [predator + stickleback juveniles]) using generalized linear mixed models (GLMM) with binomial error and spatio-temporal random effects (best structure: year-specific spatial fields) to test stressor variables (offshore stickleback biomass, distance to open sea, wave exposure) (N=3491). Extended models assessed resilience drivers: connectivity, predation pressure, fishing pressure, temperature, and interactions (connectivity×predation/fishing; temperature×distance to open sea), plus stressor variables, using GLMMs with random year effects (glmmTMB). Parallel negative binomial models for absolute predator (N=7415) and stickleback (N=7167) densities (quadratic variance). Model selection by AIC; supported effects defined as present in all candidate models with ΔAIC < 4. Model diagnostics via DHARMa; multicollinearity checks via mctest. The raster-based weighted habitat availability within 10 km with wave exposure cutoff 3.5 was the best-performing connectivity metric used for figures.

Baseline stressor effects: Predator dominance increased with distance from the open sea and decreased with offshore stickleback biomass and with wave exposure, linking the spatially propagating regime shift to offshore stickleback abundance and coastal openness. Connectivity and predation interaction: Connectivity increased predatory fish densities and the probability of predator dominance at low to medium top-predator (seal and cormorant) predation pressure. Under high predation pressure, the positive effect of connectivity disappeared for dominance and turned weakly negative for predator densities, indicating that connectivity benefits depend on the availability of adult predators to redistribute. Fishing: Fishing pressure negatively affected the probability of predator dominance but showed no clear effect on absolute predator or stickleback densities, and no supported interaction with connectivity. Temperature: Higher spawning-season degree-days (>10 °C) increased predator dominance and juvenile predator densities. For predator densities, warmer temperatures had a stronger positive effect closer to the open sea, consistent with earlier and larger stickleback arrivals there. Temperature effects on stickleback densities were inconsistent and tended to be opposite to predator responses. Relative importance (variance explained by fixed effects): - Predator dominance: R²baseline=0.34 (stressor-only), R²full=0.38 (including connectivity and local drivers). - Stickleback densities: R²baseline=0.36, R²full=0.38. - Predator densities: R²baseline=0.23, R²full=0.38. A model with only connectivity and local environmental variables (excluding incoming stickleback) explained R²≈0.36, indicating local coastal dynamics are key for absolute predator densities. For predator densities, removing predation from the full model reduced R² to 0.25 (largest drop), removing connectivity to 0.34, and removing temperature or fishing to 0.37. Overall, stressor-related drivers (offshore stickleback, distance to open sea, wave exposure) primarily explained variation in relative dominance and stickleback densities, while local coastal drivers (predation, connectivity, temperature) were more important for absolute predator densities.

Findings empirically support that resilience to a spatially propagating regime shift is jointly shaped by spatial connectivity and local environmental conditions. Connectivity promotes predator dominance and higher predator densities by facilitating redistribution and recolonization among spawning habitats, but this benefit depends on low to moderate top-predator predation; when seal and cormorant pressure is high, few adult predators remain to redistribute, potentially leading to a dilution or smearing effect and reduced local resilience. Top predators tended to shift communities toward stickleback dominance, consistent with seals and cormorants preferentially consuming larger perch and pike. Warmer spawning seasons likely enhance predator larval growth and shorten vulnerability to stickleback predation, particularly near the open sea where stickleback arrive earlier and in greater numbers; warming may thus reduce the risk of a shift locally, though broader offshore effects on stickleback remain uncertain. The regime shift dynamics appear driven by both increased stickleback influx (stressor) and loss of resilience of predatory fish populations, with evidence of flickering during transitions but no stable long-term reversals detected. Results highlight the nuanced role of connectivity—beneficial under conducive local conditions but limited or even counterproductive under high predation—and underscore the importance of integrating spatial connectivity with local management of top predators and environmental context to bolster resilience.

The study demonstrates that spatial connectivity and local environmental drivers interact to shape resilience to an ongoing regime shift from predatory fish to stickleback along the Baltic Sea coast. Connectivity can increase resilience by supporting higher predator dominance and densities, but predominantly where top-predator predation pressure is low. Warmer spawning-season temperatures further enhance resilience by accelerating predator larval development and potentially increasing consumption rates. Management should therefore couple efforts to limit the stickleback influx (e.g., managing offshore predator communities and access to spawning areas) with actions that strengthen the resilience of predator-dominated states: managing seal and cormorant impacts, prioritizing and maintaining areas of high connectivity and dispersal corridors, ensuring access to warm spring habitats, and reducing local fishing pressure. Future research should refine connectivity metrics (identify barriers, assess directional movement, quantify juvenile dispersal), incorporate eutrophication and other habitat quality factors, and leverage rigorous, spatially explicit monitoring to disentangle resistance versus recovery processes and to evaluate management interventions experimentally.

Key limitations include: coarse connectivity metrics that do not capture fine-scale habitat quality (vegetation, prey) or specific dispersal pathways (e.g., narrow passages); reliance on juvenile data (lack of concurrent adult predator abundance data) and inability to distinguish resistance from recovery in the resilience signal; correlative analyses with potential confounding among spatial variables (e.g., wave exposure correlates with habitat suitability) and spatial autocorrelation; coarse fishing pressure estimates (averaged across years, perch-focused, assumed representative for pike) and limited cormorant data (two years, interpolated); possible underestimation of spawning habitat where freshwater spawning is important; unavailability of reliable eutrophication data at the required spatial scale; and general uncertainties regarding movement directionality and the equivalence of adult-based dispersal estimates for juvenile stages.