Consequences of kelp forest ecosystem shifts and predictors of persistence through multiple stressors

Abstract

Ecological communities can be stable over multiple generations, or rapidly shift into structurally and functionally different configurations. In kelp forest ecosystems, overgrazing by sea urchins can abruptly shift forests into alternative states that are void of macroalgae and primarily dominated by actively grazing sea urchins. Beginning in 2014, a sea urchin outbreak along the central coast of California resulted in a patchy mosaic of remnant forests interspersed with sea urchin barrens. In this study, we used a 14-year subtidal monitoring dataset of invertebrates, algae, and fishes to explore changes in community structure associated with the loss of forests. We found that the spatial mosaic of barrens and forests resulted in a region-wide shift in community structure. However, the magnitude of kelp forest loss and taxonomic-level consequences were spatially heterogeneous. Taxonomic diversity declined across the region, but there were no declines in richness for any group, suggesting compositional redistribution. Baseline ecological and environmental conditions, and sea urchin behaviour, explained the persistence of forests through multiple stressors. These results indicate that spatial heterogeneity in preexisting ecological and environmental conditions can explain patterns of community change.

Keywords: alternative stable states; climate change; community structure; marine heatwave; regime shift; sea urchin.

Figure 1.

Study area along the Monterey Peninsula, California, USA. Black points on the map depict the locations of 24 long-term subtidal monitoring sites [31]. The inset figure depicts temporal trends of kelp (Macrocystis pyrifera) stipe density (left vertical axis) and purple sea urchin (Strongylocentrotus purpuratus) density (right vertical axis) across all 24 sites. Each point in the inset figure represents the mean stipe (green) or sea urchin (purple) density at a site. Splines (λ = 0.05) were fitted across interannual means with 95% confidence intervals. The vertical dotted line represents the timing of the 2013 sea star wasting (SSW) event, and the 2014–2016 marine heatwave (MHW) is shaded in red.

Figure 2.

Regional community metrics over time. (a) Changes in community structure using a 2D nMDS plot of centroids across all 24 sampled locations. Each point represents the centroid for a given year connected with interannual arrows. The green and purple ellipses depict significantly different clusters as determined by a k-means cluster analysis (k = 2). (b) Community stability over time as measured by the site-level multivariate distances (blue lines) from their 2007–2013 centroid. The median multivariate distance across sites is shown as the black line with a 95% confidence region shown in grey. Blue lines depict individual site trajectories. The vertical dotted line represents the timing of the 2013 sea star wasting (SSW) event, and the 2014–2016 marine heatwave (MHW) is shaded in red. (c–e) Community taxonomic metrics showing (c) diversity, (d) richness and (e) evenness. Each point represents a single site and splines (λ = 0.05) were fitted across interannual means with 95% confidence intervals for sessile inverts and macroalgae (purple), mobile and conspicuous inverts (orange), kelps (green), and fishes (pink).

Figure 3.

Contribution of interactions between marine heatwave period (before, 2007–2013; during, 2014–2016; after, 2017–2020) and taxonomic traits on variation in the relative abundance of fishes (top panel) and invertebrates and macroalgae (bottom panel). Colours depict the standardized coefficients for all period-trait interaction terms, based on the results of a multivariate abundance model. Coefficients were scaled to unit variance to make them visually comparable. Blue indicates a positive interaction between a given trait and period on species abundance, and red indicates a negative association. Black x's indicate interaction terms that were dropped from the multivariate model using a LASSO penalty.

Figure 4.

Changes in abundance of taxa that explained community shifts after (2014–2020) versus before (2007–2013) the sea urchin outbreak for (a) persistent and sites (b) transitioned. Only species that significantly contributed (as determined by simultaneous generalized linear models) to observed community structure differences between periods (after versus before) are shown. Each point connected with a horizontal line represents the percentage change in absolute abundance. Points to the left of the vertical dashed line indicate a decline in abundance and points to the right of the line indicate an increase in abundance. Line colour indicates the trophic function of a given species. Observed mean abundances before versus after the sea urchin outbreak are parenthetically included next to each species († = density, no. individuals per 60 m² transect; * = percentage cover). The number of taxa that significantly contributed to community shifts are indicated in the upper left and right corners (left = number that declined, right = number that increased). Finally, sea star species were removed because of the wasting event that occurred in 2013.

Figure 5.

Environmental and ecological correlates of kelp forest dynamics. (a) Partial effects of significant model predictors on kelp density from a generalized additive model. Partial effects represent the smoothing term while holding all other variables constant. Solid black lines depict the shape of the relationship between each predictor and kelp density, and dashed lines represent 95% confidence intervals. Residuals are included as blue points. (b) Boxplots of each predictor variable for persistent and transitioned sites (see electronic supplementary material, figure S3). Persistent sites are those where kelp density was not significantly different (as determined by a t-test) from before (2007–2013) versus after (2017–2020) the marine heatwave and sea urchin outbreak. Transitioned sites are those where kelp density significantly declined (p < 0.05) during the 2017–2020 period. Also included are p-values from a t-test on a given variable between persistent and transitioned sites.