How overlapping food roles and predator sizes influence kelp forest ecosystems

Keystone species often play important roles in sustaining the functioning of ecosystems and can be found occupying niches at various trophic levels (Power et al. 1996). Loss of keystone species at higher trophic levels can ultimately affect lower trophic levels through cascading forces that alter ecosystem structure and function (Paine 1966, Power et al. 1996, Estes et al. 2004). In both terrestrial and marine systems, overharvesting and climate change have driven dramatic changes in the populations of foundational and keystone species, thus resulting in abrupt and sudden phase shifts (Estes and Palmisano 1974, Daskalov et al. 2007, DeYoung et al. 2008, Ling et al. 2015, McCary et al. 2016). This trend has led to great concern about losses of biodiversity and valuable ecosystem services and spurs the need to better understand and manage for ecosystem traits that confer resilience to communities and ecosystems (Folke et al. 2004).

There is a growing body of evidence that high‐diversity systems may have enhanced ecosystem stability and resilience potential. These benefits can arise due to the number and strength of species interactions (McCann et al. 1998) and redundancy in traits and trophic roles (Menge 1983, Peterson et al. 1998, Steneck et al. 2002, Snyder et al. 2008). Ecosystems that exhibit redundancy in important functions have been shown to be more resistant to the cascading effects of disturbance and extinction (Borrvall et al. 2000, Finke and Denno 2004, Dunne and Williams 2009, Sanders et al. 2018). However, much of our understanding of the relationships between diversity, trophic redundancy, and ecosystem resilience comes from theoretical models, controlled experiments, and hypotheses derived from large‐scale correlative observations. There are few clear demonstrations of this concept in complex natural systems (but see Burt et al. 2018, McLean et al. 2019).

The Northern Channel Islands (NCI), in the Southern California Bight, provide a unique setting to explore how trophic redundancy and predator diversity strengthen resilience in kelp forest ecosystems by examining patterns occurring at the intersection of several features: historical overfishing, a historically diverse predator guild, the establishment of a large network of marine protected areas (MPAs), and a recent disturbance event. Here we explore how the disturbance, the loss of a single predator due to a widespread disease epidemic, affects populations of its preferred prey, a barrens‐forming, herbivorous sea urchin. The presence of two additional fished predators in the system along with spatial variation in fishing pressure due to the MPAs allows us to test whether functional redundancy in the predator guild has buffered the system from the cascading effects of the loss of an important predator.

Kelp forests are iconic and highly dynamic coastal ecosystems that provide a multitude of ecological functions and valuable ecosystem services (summarized in Krumhansl et al. 2016). One of the major threats to kelp forests is degradation due to overgrazing by herbivorous sea urchins, which can rapidly shift a healthy kelp forest to an urchin barren when present in high densities (Harrold and Reed 1985, Steneck et al. 2002, Filbee‐Dexter and Scheibling 2014, Ling et al. 2015). Once established, urchins in barrens may prevent the recolonization of kelps by consumption of spores and recruits and thereby stabilize the system in a degraded state (Chapman 1981). In a global analysis, Ling et al. (2015) found that such discontinuous phase shifts with hysteresis effects have been recorded in a number of macroalgae‐dominated ecosystems. Notably, across studies it was found that the minimum urchin density to initiate phase shifts from healthy kelp forests to unproductive urchin barrens occurs at ~11–14 urchins/m² but that once barrens are established urchin densities have to be reduced by an order of magnitude for kelp to recover (Filbee‐Dexter and Scheibling 2014, Ling et al. 2015, Dunn and Hovel 2019). Phase shifts with hysteresis can often have catastrophic effects on fisheries and other ecosystem services when systems are shifted into unproductive alternative states. In such situations, it is important for wildlife managers to manage for resilience, where possible, through understanding the thresholds and feedback mechanism that can foster ecosystem stability and prevent phase shifts.

Kelp forests are complex, disturbance‐driven systems that respond to a wide variety of top‐down and bottom‐up controls (Dayton 1985, Steneck et al. 2002, Reed et al. 2011, Byrnes et al. 2013, Pérez‐Matus et al. 2017), however, in many kelp systems there is good evidence for strong cascading effects of urchin predators on foundational kelps through their control of urchin populations (Estes and Palmisano 1974, Shears and Babcock 2002, Behrens and Lafferty 2004, Hamilton and Caselle 2015). Diverse suites of predatory species can further increase top‐down control of urchin populations by differentially targeting various size classes (Burt et al. 2018), and by modifying urchin behaviors (Byrnes et al. 2006), ultimately stabilizing and enhancing ecosystem resilience through the increased top‐down control (Borrvall et al. 2000, Finke and Denno 2004, Dunne and Williams 2009, Sanders et al. 2018).

The urchin predator guild along the west coast of North America is made up of four dominant species; sea otters (Enhydra lutris), sunflower sea stars (Pycnopodia helianthoides), California (CA) sheephead (Semicossyphus pulcher, a labrid fish), and California (CA) spiny lobsters (Panulirus interruptus) (Fig. 1a). These four species vary both temporally and spatially in their distributions across the region. Sunflower sea stars, until recently, occurred from southern Alaska to the Southern California Bight (Estes and Duggins 1995, Pondella et al. 2015) and were important urchin predators across this region (Moitoza and Phillips 1979, Duggins 1983, Schultz et al. 2016). Recently, a sea star wasting disease epidemic has affected more than 20 sea star species ranging from Alaska to Mexico (Hewson et al. 2018). Sunflower sea stars were particularly susceptible to the effects of the disease, and by 2014 they were functionally extirpated from most of their historic range (Schultz et al. 2016, Harvell 2019). Sea otters, a well‐known keystone predator of urchins in North Pacific kelp forests, were historically present along much of Northeast Pacific Coast until populations were decimated by intense overharvesting for the fur trade in the early 1800s (Estes and Palmisano 1974, Estes et al. 2004). Today, sea otters only persist in high densities in a fraction of their historic range (Bodkin 2015). In British Columbia, where otters and sunflower sea stars historically co‐occurred, Burt et al. (2018) showed that sea otters and sunflower sea stars display niche complementarity in their predation on sea urchins, resulting in strong top‐down control when both predators were present. Using natural variability in the presence/absence of sea otters, they showed that after the extirpation of sunflower sea stars from disease, urchin densities increased and kelp declined, but that this result was much stronger in the absence of sea otters. This demonstrates that trophic redundancy in urchin predators had allowed kelp forests in the North Pacific to persist after the loss of a single top predator.

The Northern Channel Islands (NCI) are outside of the present‐day range of sea otters, but host two additional major urchin predator species, which, along with sunflower sea stars, comprised a diverse predator guild that likely fostered robust kelp forests in the region (Steneck et al. 2002). CA sheephead and CA spiny lobsters are both generalist predators that prey upon purple urchins (Stronglyocentrotus purpuratus), with both species increasing their effectiveness as urchin predators with increasing body size (Behrens and Lafferty 2004, Eurich et al. 2014, Hamilton et al. 2014, Hamilton and Caselle 2015, Selden et al. 2017). The waters surrounding the western NCI, where the present study was conducted, have a mosaic of protection status through a network of MPAs that was established in 2003. Within the MPAs, the fished predators (CA sheephead and CA spiny lobsters) are released from commercial and sportfishing pressures. Their numbers and sizes are significantly greater in the MPAs relative to nearby fished areas (Lafferty 2004, Hamilton et al. 2010aa, Caselle et al. 2015, Selden et al. 2017).

Using the extirpation of the unfished predator—the sunflower sea star—and the presence of a mosaic of protected areas for the other two predators as a “natural” experiment, we test the effects of predator identity, abundance, and size structure on the density of purple urchins and the consequent patterns of kelp abundance. We utilize data from a long‐term monitoring study that spans the periods before and after the loss of the sunflower sea star. Specifically, we ask how MPAs might facilitate trophic redundancy through targeted predator protection and influence patterns of size‐specific foraging on purple urchins. We then ask how purple urchin abundance affects both canopy‐forming giant kelp and understory kelp. We conduct this work in the western portion of the Northern Channel Islands in order to constrain variability in some of the other drivers of kelp abundance including productivity, upwelling, and wave exposure. We hypothesize that in places with greater redundancy of predators, trophic effects will be stronger, resulting in fewer urchins and more kelp.

Sunflower sea stars were completely extirpated by disease from the rocky reef/kelp forest communities in the western NCI by 2014, which corresponded with strong temperature increases associated with an ENSO event (Fig. 2a). Prior to the extirpation of sunflower sea stars, the magnitude of difference between CA sheephead density inside and outside of reserves was relatively stable. Postextirpation, those differences increased (Fig. 2b). CA spiny lobster densities remained extremely low in the western NCI, pre‐ and postextirpation in both the reference and MPAs; however, CA spiny lobster densities were slightly higher in the MPAs and continued to increase postextirpation of sunflower sea stars (Fig. 2c). In general, purple urchin densities were similar in the MPAs and reference sites prior to the extirpation of sunflower sea stars, but coinciding with the loss of sea stars, purple urchin densities increased dramatically in the reference areas (Fig. 2d). Inside MPAs purple urchin density were variable following the loss of the sea stars (Fig. 2d). Both kelp stipe and understory algal densities varied over time (Fig. 2e, f). However, in the reference areas, algae showed a steady decline resulting in very low density, especially in recent years. Inside the MPAs, despite interannual variation, kelp and understory algal densities were greater than the reference areas (Fig. 2e, f).

We note that although our field observations of urchin barrens formation following the loss of sunflower sea stars indicated that barrens became more common and widespread, particularly in the reference areas (motivating the current analysis), we did not necessarily observe wholesale, site‐level phase shifts at all sites such as reported elsewhere in the world. Reference‐area sites showed overall declines in canopy forming and understory kelps, but retained some patches of kelp in places (Fig. 2e, f). Similarly, we did observe some high‐density patches of urchins developing in the MPAs but to a lesser extent than in the reference areas and without the resulting site‐level declines in macroalgae (Fig. 2d–f). When analyzed at finer spatial scales, at the transect level, we see supporting evidence for small‐scale patch dynamics, with barrens being more common in the reference areas than in the MPAs after the extirpation of sunflower sea stars, but with a high degree of variability in the abundance of urchins and algae within and across sites (Appendix S1: Fig. S1).

Prior to the extirpation of sunflower sea stars, the top model (lowest AICc and highest AICc weight [0.51]) for purple urchin density included fixed effects of temperature, depth, and sunflower sea star densities (Appendix S1: Table S1a) and a random effect of site (variance partition coefficient VPC = 0.615). Examining residual plots indicated linearity between the predictors and response, and homoscedastic variance; we also did not find any outliers or overly influential points (Appendix S1: Fig. S2). Interestingly, model predictions indicate purple urchin populations both in the MPAs, and reference areas were low when sunflower sea stars were present at any density (Fig. 3).

Postextirpation of sunflower sea stars, the top model (lowest AICc and highest AICc weight [0.99]) included fixed effects of temperature, depth, CA sheephead abundance, CA sheephead total length, and protection status (Appendix S1: Table S2a) and a random effect of site (VPC = 0.158). Again, examining residual plots indicated linearity between the predictors and response, homoscedastic variance, and no outliers nor overly influential points (Appendix S1: Fig. S3). Empirically measured mean CA sheephead total length (TL) and abundance were larger inside the MPAs (X¯_TL = 33.45 cm [S_TL = 9.03 cm], X¯_Abundance = 98.35 [SD_Abundance = 77.06]) compared to reference areas (X¯_TL = 30.34 cm [SD_TL = 7.89 cm], X¯_Abundance = 42.91 (SD_Abundance = 34.81), indicating that the urchin predator demographics are different, with larger more numerous individuals within the MPAs. We plotted the predicted urchin densities from this model against CA sheephead density for three different size classes of CA sheephead (small = 24.56 cm, medium = 31.50 cm, and large = 38.15 cm; sizes were calculated based on the first, second, and third quartiles of CA sheephead sizes in the empirical field data set). Predictions from the fitted model suggest effects of CA sheephead abundance and size on purple urchin densities are stronger in MPAs (Fig. 4a). Large CA sheephead are more effective at suppressing purple urchins at lower abundances than other size classes, yet even smaller CA sheephead can also suppress purple urchin populations when present at high abundances (Fig. 4a). Outside of MPAs, the effects that CA sheephead size and abundance have on purple urchin populations are similar to the effects in MPAs, yet of smaller magnitude (Fig. 4b).

Filbee‐Dexter and Scheibling (2014) summarized the threshold urchin densities needed to reverse phase shifts from barrens to kelp in California using data from Dean et al. (1984) and Dayton et al. (1992). The threshold densities were 2–3 individual urchins/m². We asked how many CA sheephead of different size classes (TL) would be required to produce an urchin density of 2/m². Using our model predictions, we find that inside the MPAs 11 “large‐sized” individuals (average TL 38.2 cm) per 60‐m² transect corresponds with 2 urchins/m², compared with 91 “small‐sized” individuals per transect (average TL 24.6 cm) to result in that same urchin density. To obtain that same urchin density in the reference areas would require a density of 24 large‐sized or 210 small‐sized CA sheephead per transect.

The top model (lowest AICc and highest AICc weight [1.00]) for macroalgal density included effects of temperature, depth, and purple urchin densities (Appendix: Table S3a, Appendix: Fig. S4), showing a strong negative association between purple urchin density and macroalgal density. S1 S1

We found that healthy kelp forest ecosystems harbored trophic redundancy in the urchin predator guild. Additionally, we found relationships between the observed size class distributions of certain predators and urchin populations. Populations that included large predators were more effective at controlling urchins. These differences in predator demographics were related to levels of protection across an MPA network (Lafferty 2004, Hamilton et al. 2010aa, Caselle et al. 2015, Selden et al. 2017). Using the NCI as a model ecosystem for kelp forest dynamics, we showed that the effects CA sheephead, CA spiny lobster, and sunflower sea stars had on purple urchin densities varied as a function of the presence of the other urchin predators, ultimately leading to the differences in the persistence of canopy forming giant kelp and understory macroalgae. In MPAs, prior to the disease outbreak, purple urchins were exposed to predation from a three‐pronged predator guild that effectively maintained low urchin densities. With the onset of the sea star disease and widespread extirpation of sunflower sea stars, the predation pressure by CA spiny lobsters and CA sheephead on purple urchins was still strong enough to maintain low purple urchin densities inside the MPAs allowing macroalgae to persist (Figs. 2 and 4). In addition to protection from fishing afforded by the MPAs, the loss of the sunflower sea star might also have benefited the other two predators by releasing them from competitive pressures. However, outside of the MPAs, these fished predators were incapable of compensating for the loss of sunflower sea stars, and urchin barrens became common. The greater predation pressures exerted by CA sheephead and CA spiny lobsters in MPAs are likely driven by the intact size structures and higher abundances resulting from protection from fishing (Hamilton et al. 2010a). However, the predictions from the models indicate that CA sheephead at the reference sites might exert weaker predation pressures on purple urchins than CA sheephead of the same size and abundance inside the MPA. Although this intriguing result requires further research, we hypothesize that CA sheephead may preferentially avoid eating urchins from barrens because of a lower quality and lack of gonads in barren urchins, as was found for CA spiny lobster by Eurich et al. (2014). This pattern may also be explained in part by the additive predation pressure resulting from higher densities of CA spiny lobster in the MPAs (Kay et al. 2012). One particularly interesting result of our modeling of relationships between different size classes of CA sheephead and urchin abundance was the large disparity in the density required to maintain urchins at low densities across size classes of CA sheephead. Both in and out of the MPAs, the number of small sheephead required is nearly an order of magnitude higher than the number of large sheephead required to maintain the same density of urchins. Given the difficulty for small predators to compensate for the functional role of large predators, this result further emphasizes the importance of MPAs in restoring size distributions of fished species. Furthermore, in addition to the direct consumptive effects of predators on urchin populations, higher densities and diversity of predators in MPAs can potentially magnify trophic cascades indirectly because of effects on prey behavior (Byrnes et al. 2006, Spyksma et al. 2017). It is also important to note that CA sheephead are diurnal predators and CA spiny lobster are nocturnal predators, resulting in temporal variation in predation risk which could also have impacts on urchin grazing behavior and degree of urchin predation. Lobsters may have been underestimated by our daytime, visual surveys, but clear MPA effects on their abundance and size have been shown in the NCI (Kay et al. 2012). Our study lends support to the effectiveness of MPAs not only for restoring the biomass of fished species (Hamilton et al. 2010aa), but also for enhancing ecosystem resilience through trophic redundancy and restoration of biodiversity and size class structure.

In southern California, there is evidence that kelp forest/urchin barren dynamics operate at finer spatial scales (<500 m) than in other sections of the Pacific coast, where physical forces such as wave disturbance have been shown to influence regional trends in urchin and kelp abundance (Dayton et al. 1984, Reed et al. 2011, Cavanaugh et al. 2013, Karatayev et al. 2019). Our findings are consistent with these fine‐scale patch dynamics in urchin and algal abundances in our study region (Appendix S1: Fig. S1). Along with Southern California’s finer‐scale feedbacks between grazers and kelps, we know that abiotic (physical and chemical) factors are important for structuring macroalgal communities (Dayton 1985, Reed et al. 2011, Cavanaugh et al. 2013) and likely contribute to some of the temporal variation documented in our data sets. However, the spatial configuration of our sites is appropriate for controlling those factors to evaluate the effects of the MPAs. Sites are closely paired, in and out of MPAs at three islands, thus we would not expect abiotic factors to be driving the differences we show here. Instead, though correlative, the MPA effects we document are more consistent with trophic effects.

The NCI are situated in a biogeographic transition zone where the cold waters of the California current meet the warmer southern California countercurrent (Harms and Winant 1998). The resulting strong east–west temperature gradient is associated with variation in kelp forest communities across the islands (Hamilton et al. 2010bb, Caselle et al. 2015). Nearby Point Conception is a well‐recognized biogeographic boundary and this region is near the northern range limits for both CA sheephead and CA spiny lobsters (Fig. 1a), which are both associated with the warmer waters south of Point Conception and the eastern NCI (Murray et al. 1980, Burton 1998, Pondella et al. 2015). Prior to extirpation, sunflower sea stars were associated with the cooler waters north of Point Conception and the western NCI (Bonaviri et al. 2017). The overlapping range limits of these predators demonstrates an interesting spatial complementarity that potentially balances the weaker predator pressures exerted from CA sheephead and CA spiny lobsters at the cooler islands. North of Point Conception, sunflower sea stars and sea otters were the primary urchin predators prior to the onset of sea star disease; however, the patchy distribution of sea otters along the northern coast of North America has left large sections of the coast without a dominant urchin predator following the loss of sunflower sea stars. In some of these regions, such as northern California, the loss of macroalgae in those areas has been extensive. This mosaic of urchin predator distributions across the coast of North America (Fig. 1a) sets up interesting and testable hypotheses regarding the relationships between predator redundancy and coastwide resilience of kelp forests, which, we recommend, should guide future work.

At Anacapa Island, the easternmost and warmest of the NCI, CA spiny lobsters and CA sheephead are the primary predators of sea urchins (Caselle et al. 2018), and densities of sunflower sea stars were historically low even prior to their extirpation (Bonaviri et al. 2017). The island is zoned with varying levels of protection, which include a no‐take marine reserve, a partial‐take marine conservation area (commercial and recreational take of CA spiny lobsters permitted), and areas fully open to fishing. Caselle et al. (2018) showed that CA sheephead and CA spiny lobsters have strong direct negative effects on urchin abundance and an indirect positive effect on native algal abundance through their predation on urchins. Urchin densities were lowest and kelp densities highest in the no‐take reserve, where both urchin predators are protected. The inverse was observed in the fished area where CA sheephead and CA spiny lobsters are in low abundance. In the partial‐take MPA, where only one predator is protected, urchins and kelp were at intermediate densities, demonstrating the complementary effects of multiple predators such as we showed here. These spatial patterns are an indication that trophic redundancy enhances resilience and may help prevent ecosystems from shifting to alternative states. It also reinforces the importance of ecosystem‐based management strategies such as MPAs, which preserve functional roles at high trophic levels, thus restoring and preserving ecosystem integrity.

Areas along the coast that lacked diverse predator guilds or redundancy in the trophic structure were affected disproportionately more from loss of predators because of the sea star disease (Schultz et al. 2016). Sea otters, where they were present, maintained the predatory niche breadth, through trophic redundancy, in the absence of sunflower sea stars (Burt et al. 2018). However, in our study this trophic redundancy was only observed in the MPAs, because the redundant predators (i.e., CA sheephead and CA spiny lobsters) are subject to fisheries. Sea otter populations are positively correlated with black abalone populations and negatively correlated with urchin populations along the central coast of California, where sea otters are locally present (Raimondi et al. 2015). The sea otters suppress urchin populations, preventing urchin barrens from forming and allowing kelp to persist. The presence of the kelp forests creates a favored environment for black abalone food sources, thus demonstrating how the presence of a keystone species in a healthy kelp forest can indirectly harbor more diversity. In northern California, the recreational red abalone fishery is valued in the tens of millions of dollars (Reid et al. 2016). The California Fish and Game Commission closed the abalone fishery in 2018 for a minimum of three seasons, citing negative effects on abalone populations due to “extreme environmental condition” which broadly included the loss of the only key urchin predator to disease, massive increases in urchin barrens, and widespread losses of kelp (Rogers‐Bennett and Catton 2019). This is an example of how cascading ecological effects can reverberate through lower trophic levels resulting in devastating environmental effects that require drastic management actions that ultimately have profound economic consequences. Such consequences could perhaps be minimized, however, by placing additional management focus on trophic redundancy when possible.

As conservation efforts work to enhance ecosystem stability and buffer against anthropogenic change, management that promotes trophic redundancy has the potential to be highly effective. In coral reef systems, trophic redundancy has been found to play an important role in macroalgal abatement. With protection, there was significantly more functional herbivore diversity and functional redundancy when compared to fished areas, with the highest functional diversity and functional redundancy found in the large roving browsers of turf and macroalgae (Micheli et al. 2014). Through protection, these grazers proliferated, thus keeping the reef in a coral‐dominated state due to the resiliency that increased diversity brings. Diversity of grazer species can maximize herbivory processes leading to increased coral recruitment due to a release from competition with algae (Lefcheck et al. 2019). However, protecting the diversity needed to promote resiliency can be especially challenging for wildlife managers in naturally lower diversity systems such as those that occur at higher latitudes (Pianka 1966). Lower natural diversity at high latitudes combined with human harvesting may present additional challenges in preventing ecosystems from shifting to alternative states once disturbed. The interlocking web of stability a predator guild creates allows for ecosystems to absorb more perturbation than single predator systems. Mathematical modeling has indicated that ecosystem stability increases when multiple weak predator–prey interactions occur when compared to one strong predator–prey interaction (McCann et al. 1998). The presence of only one predator in a system creates a scenario where a perturbation to the predator populations could drastically shift the community to an alternative state.

As ecosystems are becoming more and more disturbed through climate change and anthropogenic stressors, it is important that we understand the mechanisms that prevent community‐wide phase shifts and ecosystem attributes that enhance resiliency. Our study demonstrates the importance of trophic redundancy in stabilizing perturbed system, which lends support to the previous theories, models, and small‐scale experiments on trophic redundancy (Steneck et al. 2002, Byrnes et al. 2006, McCary et al. 2016, Sanders et al. 2018), and identifies practical management actions that help to stabilize a key marine ecosystem in the face of a large‐scale ecosystem shock. Now more than ever, we need to protect our ecosystems’ predator guilds in order to maintain the stability of our ecosystems and prevent phase shifts in hopes of combatting ecosystem stressors.

We thank all the current and previous PISCO divers who helped obtain the data. We also thank A. Parsons‐Field for the helpful comments and leading the dive teams. Temperature data were made available by the SBC‐MBON project. Funding was provided by David and Lucille Packard Foundation, the California Ocean Protection Council, California SeaGrant, and the UC Santa Barbara Coastal Fund. JHE, KD, and JEC conceived the project idea and collected the data. JHE, JME, and PC designed the analytical methods. JHE and JME analyzed the data. JHE led the manuscript. All authors contributed to drafts and editing of the manuscript and provided final approval for publication. This is publication 506 from PISCO, a long‐term ecological consortium, which is funded primarily by the David and Lucille Packard Foundation.

Eisaguirre, J. H. , Eisaguirre J. M., Davis K., Carlson P. M., Gaines S. D., and Caselle J. E.. 2020. Trophic redundancy and predator size class structure drive differences in kelp forest ecosystem dynamics. Ecology 101(5):e02993 10.1002/ecy.2993