Climate-driven invasion and incipient warnings of kelp ecosystem collapse

Global climate change is driving widespread species redistributions and nonlinear ecosystem changes with potential for catastrophic collapses. Traditional early-warning approaches emphasize detecting critical slowing down in high-frequency, long-term time series, but such data are rare for most ecosystems already affected by rapid warming. Early warnings for climate-driven invasions may instead be detected via spatial pattern formation as warm-environment species extend poleward. In southeastern Australia, strengthening of the East Australian Current over ~60 years has increased poleward penetration of warm water and quadrupled regional ocean warming rates relative to the global average. Within this context, the long-spined sea urchin Centrostephanus rodgersii has extended its range ~700 km to Tasmania, where it can overgraze kelp forests and create persistent urchin barrens. Early warning signs—small, locally grazed patches termed ‘incipient barrens’—were first observed in the early 2000s. Because overgrazing exhibits tipping points and hysteresis (higher threshold for collapse than for recovery), detecting such spatial early-warning signals could allow proactive adaptation before widespread kelp collapse.

The study builds on theory of catastrophic regime shifts and early-warning signals, including critical slowing down and self-organized spatial patchiness in ecosystems. Prior work documents climate-driven tropicalization of temperate systems, poleward species range shifts in eastern Tasmania, and the ecological impacts of C. rodgersii overgrazing resulting in alternative stable states (kelp forests vs. urchin barrens) with hysteresis. Research has shown the role of oceanographic change (strengthened East Australian Current), predator depletion (overfishing of spiny lobsters) reducing kelp forest resilience, and interactions among herbivores and competitors (native urchins, abalone). These studies motivate focusing on spatial early-warning indicators (incipient barrens) and the use of space-for-time substitutions to anticipate collapse along invasion fronts.

A repeated broad-scale survey replicated methods from 2001/02 in 2016/17 across 13 eastern Tasmanian coastal reef sites (~20 km apart; spanning ~400 km). At each site, 3 subsites (~2 km apart) were surveyed, each with 4 dive transects (~200 m apart), totaling 156 diver transects per period. Transects were set perpendicular to shore from ~4 m to 18 m depth (or up to 100 m length), averaging ~50 m. Divers surveyed contiguous 5 m by 1 m quadrats (5 m²), recording depth, abundances of C. rodgersii, Heliocidaris erythrogramma, Jasus edwardsii (spiny lobster), and Haliotis rubra (abalone). Substratum cover was estimated to nearest 5% and categorized (flat rock, large boulders, small boulders, cobble, pebble, gravel, sand). Urchin barrens cover was estimated similarly and distinguished from non-macroalgal areas due to other causes. To extend spatial and depth coverage, towed-underwater-video transects were conducted in each period: two perpendicular (4–40 m depth) and two parallel (~1 km at ~15 m depth) per subsite, exceeding 80 km of reef per period. Video was time-stamped with GPS and depth; camera towed ~1–2 m above seafloor with ~20–30 m towline. Video was post-processed into contiguous ~10 m intervals and classified into four barrens categories: continuous (>85% barrens), and patchy incipient barrens (small <20%, medium 20–40%, large >40–85%). Overall planar barrens cover per transect was derived by weighting each category by the mid-point of its cover range (continuous 0.925; small 0.10; medium 0.30; large 0.625) and summing. Additionally, from 2008–2011, divers conducted 76 georeferenced timed swims (5–15 m depth) at six sites to estimate sizes of 1,297 individual barrens patches using a calibrated 1×1 m quadrat. Data handling included converting counts to densities after excluding sand areas from quadrats or video intervals and matching transects by depth and length between periods. Statistical analyses in R included mixed-effects ANOVA (nested 2-factor) testing Time (fixed: 2001/02 vs 2016/17), Site (random; 13 sites), Time×Site, and Subsite nested within Time×Site for dive and video metrics; 3-way fixed-effects ANOVA assessed effects of depth strata (2 m bins from 4–18 m), substratum type (>50% dominant cover), and Time on C. rodgersii density and barrens cover (Sites 1–9). Multiple linear regression (n=3200 quadrats) evaluated contributions of Time, latitude, substratum, depth, and co-occurring invertebrates (lobsters, native urchins, abalone) to C. rodgersii density and barrens; variable importance used LMG (relaimpo). Box-Cox transformations (MASS) were applied as appropriate. Highly correlated predictors (|r|≥0.70) were reduced (e.g., only “large boulder” retained among substrata).

• Across all 13 sites (to 18 m depth), C. rodgersii density increased from 0.10 ± 0.05 SE to 0.18 ± 0.06 SE individuals m⁻² from 2001/02 to 2016/17 (x1.8; ~3.8% per annum). Spatial unevenness at the ~20 km site scale persisted through time; increases varied significantly at the ~2 km subsite scale.
• In eastern Tasmania (Sites 1–9), urchin density rose from 0.15 ± 0.06 SE to 0.26 ± 0.07 SE individuals m⁻² (x1.7). Over the same coast section, barrens cover increased x3.9 from 2.3% ± 1.6% to 9.0% ± 3.9% (compounding ~9.5% per annum), with significant subsite-scale variability.
• The kelp–urchin relationship steepened through time: kelp cover vs. urchin density trendlines shifted from y = −19.82x + 100.31 (R² = 0.56; n = 1600 quadrats) in 2001/02 to y = −33.58x + 99.85 (R² = 0.65; n = 1600) in 2016/17, indicating disproportionate kelp loss relative to urchin increases. Tipping points: overgrazing at ~2.2 urchins m⁻²; kelp recovery at ~0.36 urchins m⁻².
• Depth and habitat effects: greater increases in urchins and barrens on deeper reefs, especially on large boulder substrata (>8 m depth) compared with small boulders or flat rock.
• Biotic associations: C. rodgersii positively associated with native urchin H. erythrogramma; both urchin species negatively associated with predatory spiny lobsters J. edwardsii. Abalone H. rubra negatively associated with urchin barrens.
• Towed-video (4–40 m) showed significant increases from 2001/02 to 2016/17 in small, medium, and large incipient barrens, and in continuous barrens across eastern Tasmania. Small incipient barrens rose most overall but varied by site depending on progression to larger patch sizes. Planar barrens cover increased from 3.4% to 15.2% overall, implying ~10.5% per annum expansion.
• Projection: If unmitigated, barrens cover in eastern Tasmania (Sites 1–9) is projected to approach ~50% by ~2029, consistent with local observations at St. Helens and patterns in NSW. Incipient barrens currently occur on ~40% of reefs, signaling widespread risk of imminent collapse.

Results demonstrate a climate-driven, poleward-propagating invasion by C. rodgersii coupled with nonlinear collapse from kelp forests to urchin barrens. The marked rise in incipient barrens—especially small patches that expand and coalesce—provides a practical, spatial early-warning indicator preceding catastrophic phase shifts. Steepening kelp–urchin negative relationships reflect local exceedance of the overgrazing tipping point, with hysteresis implying difficult recovery once barrens form. Depth and substratum patterns (deeper, large boulder reefs most vulnerable) and trophic associations (negative with lobsters) clarify mechanisms of invasion success and grazing impact. The study shows that looking equatorward along invasion fronts and using space-for-time substitutions can reveal early-warning spatial pattern formation in systems lacking high-frequency time series, enabling adaptation ahead of collapse. Towed video offers scalable, cost-effective monitoring to detect these patterns across depth and large spatial extents. Findings support management that rebuilds predator populations and targets urchin removal to reduce densities below overgrazing thresholds.

The study identifies incipient barrens as robust spatial early-warning signals of kelp ecosystem collapse driven by a poleward range extension of C. rodgersii in an ocean-warming hotspot. Over 15 years, urchin density and barrens cover increased markedly, with projections indicating potential collapse of roughly half of eastern Tasmanian kelp reefs by ~2030 if trends continue. Mechanistic insights highlight vulnerability on deeper, large boulder reefs and the importance of predator (spiny lobster) recovery. Practically, the work advances a space-for-time, equatorward-looking framework for detecting impending regime shifts and guiding timely adaptation—via predator rebuilding, targeted urchin culling, and development of urchin harvest incentives focused on high-risk reefs. Future efforts should expand high-frequency, spatially extensive monitoring (e.g., towed video), refine predictive models of barrens expansion, and evaluate the efficacy and scalability of management interventions to maintain urchin densities below critical thresholds.

• Temporal resolution is limited to two main survey periods (2001/02 and 2016/17) for the primary comparisons, constraining detection of interannual variability and transient dynamics.
• Projections of barrens reaching ~50% cover by ~2029 assume continuation of observed unmitigated expansion rates and may be sensitive to future environmental variability and management actions.
• Spatial heterogeneity was substantial (significant subsite-level variability; uneven site-scale urchin abundance), which may affect generalizability across the coastline.
• The study relies on observational surveys (with experimental evidence cited from prior work) and space-for-time substitution to infer early warnings; causal attribution is informed but not exclusively experimental within this dataset.
• Methodological constraints include depth limits for diver surveys (to 18 m) and potential positional offset of towed-camera relative to GPS (camera 20–30 m behind the boat), although standardized protocols were applied across periods.