The behavior of sympatric sea urchin species across an ecosystem state gradient

Study site

The study was conducted at Arikawa Bay (Fig. 1), located in northeastern Nakadori Island of the Goto Islands, Nagasaki Prefecture, Japan (129.11°E, 32.99°N). The study site presents a case of contrasting ecosystem states where a vegetated and an isoyake habitat are spatially adjacent but physically separated by a concrete structure (Fig. 1, inset).

Arikawa Bay is exposed to wind and waves from the north. In general, the coastline is rocky while some parts are armored with concrete walls and wave-breakers. We examined two distinct ecosystems: (A) a vegetated habitat, hereafter known as VH and (B) an isoyake habitat, hereafter known as IH. Within each ecosystem, a 20-m stretch of rocky reef was selected for monitoring in deep (4.27 ± 0.60 m mean ± SD at IH and 4.62 ± 0.62 m at VH) and shallow areas (2.07 ± 0.60 m at IH and 3.06 ± 0.67 m at VH).

The VH is open to the north (i.e., windward side of a concrete jetty) and has a mean wind fetch of 2,342 m (number of vectors reaching maximum limit: 13) (Sato et al., 2022) (see Fig. S1). The substrate is mostly consolidated rocks and small to medium-sized boulders. The VH is dominated by small algae (Corallina spp., Gelidium spp.) and turf algae present year round and sparse patches of perennial Sargassum macrocarpum. Blooms of seasonal macroalgae (i.e., Sargassum horneri, Asparagopsis taxiformis, Dictyopteris spp., Codium fragile, Colpomenia sinuosa and Hydroclathrus clathratus) occur in mid- to late- spring and persist until early to mid-summer. The IH is located behind a concrete jetty and is sheltered from waves approaching from the north (Fig. 1, see inset) and has a mean wind fetch of 1,889 m (12). The IH substrate is also rocky, however silty patches occur in the deeper areas. The rocky surfaces are bare, except for a few patches of encrusting coralline algae apart from the seasonal macroalgal bloom (i.e., S. horneri, C. sinuosa, H. clathratus) that occurs from spring to early summer.

Field monitoring

To determine changes in the benthic environment, we conducted monitoring activities by skin diving in the daytime (ca. 07:00–08:00), along marked sections in the deep and shallow reefs of VH and IH. A 20-m rope was secured to the substrate on both ends and set parallel to shore to guide the installation of a transect tape every monitoring period. Photos of the substrate were taken using a 1 m tall, 1 m² polyvinyl chloride (PVC) photoquadrat (Preskitt, Vroom & Smith, 2004) with a GoPro Hero 8 camera at 1 m intervals along each belt transect. A total of 10 photos were taken for each belt transect every monitoring activity. For the sea urchin monitoring, all sea urchins observed within a 2-m-wide swath at 1 m intervals (2 $\times$ 1 m) along the transect were recorded for their species and whether they occupied pits, crevices, or were free-living (Yusa & Yamamoto, 1994; Gravem & Adams, 2012; Frey & Gagnon, 2016). A total of 10, 2 $\times$ 1 m swaths were examined for each belt transect every survey period and the body diameter of all sea urchin species found were measured using a caliper in situ. Sea urchin sizes were classified as small (<3.9 cm), medium (4.0–5.9 cm) or large (>6 cm). Five sea urchin species occur at the study site, which include Hemicentrotus pulcherrimus, Toxopnuestes pileolus, D. savignyi, D. setosum, and H. crassispina. Of these, H. pulcherrimus was cryptic, usually found under boulders and was observed along the transect only once, whereas T. pileolus were scattered widely, and only thirteen individuals were found during the monitoring period. Given the scarcity of T. pileolus and H. pulcherrimus, they were not included in the formal analysis. Sea urchin biomass was estimated from sea urchin size-weight relationship determined by fitting a generalized linear model (GLM) on the weight and size from 10 randomly collected individuals of D. savignyi, D. setosum and H. crassispina from the shallow and deep transects.

Quadrat photos were downloaded from the camera and 50 points were randomly plotted on each photo across a 10 $\times$ 10 grid via R (R Core Team, 2022) using the magick (Ooms, 2021) and imager (Barthelme et al., 2021) packages. Each point was identified and classified whether they were macroalgae, coralline algae (i.e., encrusting, geniculate), turf algae, substrate (i.e., rock, sand, silt), or other (i.e., debris, benthic animals). Benthic elements not associated with benthic cover (i.e., other) were excluded from the analysis.

Environmental conditions were recorded using data loggers. For each habitat, one wave height logger (Infinity-WH AWH-USB; Alec Electronics Co., Kobe, Japan) was deployed at a point midway between the deep and shallow transects. For each deep and shallow belt transect, one depth logger (HOBO U20; Onset Computer Corp., Bourne, MA, USA) and one temperature/light logger (HOBO Pendant MX2202; Onset Computer Corp., Bourne, MA, USA) was installed. The instruments recorded environmental data at 10-min intervals for 20–25 days each month before retrieval and data offloading. The monthly daily averages for wave height (m) and temperature (°C) were calculated for each month. The monthly average of the integrated photosynthetic photon flux density (PPFD, mol m⁻² day⁻¹) was also calculated for each month.

Sea urchin mark-recapture experiment

To measure sea urchin movement patterns within a 24 h period, mark-recapture experiments were conducted in late August to early September 2020 on the two most abundant sea urchin species at the study sites; the long-spined black sea urchin, D. setosum and the purple sea urchin, H. crassispina. Whereas D. setosum occurred in groups of 2–30 individuals, H. crassispina was solitary. Twenty individuals each of D. setosum and H. crassispina from each habitat were selected, while ensuring that the individuals were separated by at least 5 m. The initial location of all individuals was marked by placing numbered buoys. A numbered plastic T-bar tag was inserted into the inter-ambulacral region using a tagging gun with a 0.5 mm diameter needle (Rodríguez-Barreras & Sabat, 2015), and each tagged individual was treated as one trial. All tagged sea urchins were returned to their original position and the number of individuals which composed the group at the initial location, were also recorded. The initial capture and tagging activity was conducted between 06:00 to 07:00, labelled as “start” and assigned a linear distance of 0 m. The first recapture occurred after 12 h (i.e., 18:00) and the second recapture was after another 12 h (i.e., 6:00–7:00 on the next day). About 5–10 min was allocated for recapturing the tagged sea urchins via a circular search pattern by snorkeling. Once a tagged individual was found, the marked buoy was moved to the new location and the linear displacement was measured. Those not found during the recapture were labelled as lost and excluded from the analysis.

Benthic rugosity measurements

Rugosity can be used as a proxy to describe substrate complexity, where a flat surface would have a value of one and higher values can describe a more complex benthic structure. Benthic rugosity was measured by laying a 5 m length of stainless-steel chain with 1 cm links along the contour of the rocky reef. A transect tape was laid beside the chain to measure the straight-line length the chain end-points (Risk, 1972; Trebilco et al., 2015). The straight-line chain length (5 m) was divided by the overlaid length of the chain to estimate rugosity. A total of 35 and 25 measurements were done at the deep and shallow areas, respectively in the IH and 24 and 36 were done at the deep and shallow areas, respectively of the VH.

Data analysis

Sea urchin movement patterns

The linear displacement by D. setosum and H. crassispina over a 24 h period during the mark-recapture experiment was analyzed using a GLM, assuming a hurdle-gamma distribution (Lewin et al., 2010) with a log link-function (Eq. (3)).

(3) $$\begin{array}{rl}& y=\left(1-\pi \right)\mathrm{\Gamma }\left(0,\theta \right)+\pi \mathrm{\Gamma }\left(\mu ,\theta \right)\\ & \mu =x\beta \\ & log{\displaystyle \frac{\pi }{1-\pi }=x\alpha }\end{array}$$

$$\begin{array}{rl}& \beta \sim Normal\left(0,1\right)\\ & \theta \sim Studen{t}_3\left(0,1\right)\\ & {\theta }_{\theta ,trial}\sim Studen{t}_3\left(0,2\right)\end{array}$$

Here, $y$ is the sea urchin displacement over the mark-recapture period; $\pi$ is the probability of a non-zero value; $\alpha$ and $\beta$ are the model coefficients; $x$ is the predictor variable. The parameters $\mu$ and $\theta$ are the location and scale of the Gamma distribution, respectively. The predictor variables were urchin species (two factor levels: D. setosum and H. crassispina), habitat (two factor levels: IH, VH), recapture period (two factor levels: first recapture, and second recapture), and the experimental trial was treated as a random intercept. The $\beta$ coefficients were assigned a normal prior with a location of 0 and a scale of 0.5 while $\theta$ was assigned a Student’s t prior with three degrees-of-freedom, a location of 0, and scale of 2.

The change in sea urchin group size was analyzed using a Bayesian GLM which assumed a negative binomial distribution with a log-link function (Eq. (4)). For this analysis, H. crassispina was excluded since it was solitary throughout the mark-recapture period.

(4) $$\begin{array}{rl}& y=NegBin\left(\mu ,\theta \right)\\ & \mu =exp\left({\beta }_1+{\beta }_2\right)\end{array}$$

$\begin{array}{l}{\beta }_1~Studen{t}_3\left(0,{\theta }_{{\beta }_1}\right)\\ {\beta }_2~Studen{t}_3\left(0,{\theta }_{{\beta }_2}\right)\\ \theta ~Studen{t}_3\left(0,{\theta }_{\theta }\right)\\ {\theta }_{{\beta }_{1,trial}}~Studen{t}_3\left(0,2\right)\\ {\theta }_{{\beta }_{2,trial}}~Studen{t}_3\left(0,2\right)\\ {\theta }_{\theta ,trial}~Studen{t}_3\left(0,2\right)\end{array}$where, $y$ is the number of D. setosum individuals together with the tagged sea urchin during the start, first recapture, or second recapture period; The parameters $\mu$ and $\theta$ are the location and scale of the negative binomial distribution. The $\theta$ and $\beta$ coefficients were assigned Student’s t priors with three degrees-of-freedom, a location of 0 and a scale of 2. The predictor variables were the recapture period (three factor levels: start, first recapture, second recapture), and habitat (two factor levels: IH, VH), while the period and habitat were nested in the varying intercept, experimental trial.

All analyses were conducted in R version 4.1.2 (R Core Team, 2022). Bayesian methods were used for all models through the brms package (Bürkner, 2017, 2018) and assigned weakly informative priors (Gelman, Simpson & Betancourt, 2017), where each model was ran with four Markov chains with 4,000 iterations per chain. The chains and posterior distributions were assessed visually for convergence (see model validations in Figs. S3–S17). All models were compared with a similar model having the same distributional parameters but with no explanatory variables (null model) using the difference in the expected log pointwise predictive density (ELPD) of the leave-one-out cross validation (LOO) (Table 1) (Vehtari, Gelman & Gabry, 2017). As the data were not normally distributed, we reported the conditional means that were estimated by the Bayesian models.

Sea urchin size-class and microhabitat preference

The sea urchins were distributed widely across the two study sites and had distinct microhabitat preferences during the study period (Fig. 6 and Table S9). At the IH, greater than 50% of medium-sized D. setosum occurred as free-living during the autumn and early winter months of 2020 and during the summer-autumn months of 2021 (Figs. 6A and 6B). However, the preference for crevices among D. setosum increased from the winter months of 2020 until mid-spring of 2021. The same patterns were observed for the deep and shallow transects of IH. For D. savignyi, few sea urchins were recorded at IH. In the deep area, an increasing trend was found for small individuals preferring crevices from spring until summer 2021, while less than 50% were free-living. In the shallow IH, the occurrence of D. savignyi in crevices and as free-living was below 25%. For H. crassispina, sea urchins preferred mostly crevices in the deep IH. Free-living H. crassispina in the deep IH occurred only in September of 2020 and 2021 (13.94% and 10.86%, respectively). The H. crassispina found in the shallow IH was found in all microhabitat types and almost exclusively composed of medium-sized individuals. However, free-living individuals were not recorded from August to December 2021.

In the deep VH, D. setosum was most abundant (Fig. 6C), composed mainly of small and large-sized individuals and followed similar seasonal patterns of microhabitat preference as those in IH. In the shallow VH, all size classes of D. setosum mostly preferred crevices while few medium and large-sized individuals were free-living from spring to late summer of 2021. For D. savignyi, small and large individuals preferred crevices while small individuals were free-living in the deep VH. In the shallow VH, medium and large individuals preferred crevices while large individuals were free-living. For H. crassispina, small sea urchins mostly preferred crevices together with a few large individuals in the deep VH throughout the study period. Although free-living individuals were found in the spring of 2021, H. crassispina in the shallow VH generally preferred either pits or crevices throughout the study period.

Sea urchin movement patterns

There was a difference between the movement patterns of H. crassispina and D. setosum across the IH and VH (Fig. 7A), with a clear pattern that shows increasing linear distance after the second recapture compared to the first recapture. On average, both species displaced further in the IH than at the VH in the second recapture (Table S10).

Both species also had distinct social behaviors because H. crassispina was always solitary in its crevice or burrow and was not found together with other conspecifics when it was outside. In contrast, D. setosum was always observed in groups. The mark-recapture experiment shows that D. setosum group composition varied widely in the VH compared to the IH (Fig. 7B). The group-size GLM (Table S11) shows that there was a decrease in the average number of individuals in the first recapture in both IH (from 4.38 to 2.16 indiv.) and VH (from 7.73 to 5.55 indiv.) while an increase was observed for the group size in IH (from 2.16 to 5.29 indiv.) but not for VH (from 5.55 to 4.80 indiv.), after the second recapture.

The relatively large displacement and the changing group size by D. setosum caused difficulty in locating the tagged individuals during the recapture periods leading to losses (Table 2). Following the second recapture in VH, 25% (n = 5 indiv.) of tagged D. setosum were lost. At IH, 30% (n = 6 indiv.) were lost during the first recapture and 50% (n = 10 indiv.) were subsequently lost after the second recapture period. For H. crassispina, eighteen individuals in VH remained in their initial positions throughout the experiment while two were lost after the second recapture period. All H. crassispina were located in the IH.

Sea urchin size-class and microhabitat preference

Sea urchin microhabitat preference may have been associated with changes in environmental conditions because the gradual decline in the occurrence of free-living D. setosum from autumn until spring and the increase in preference for crevices during the same period (Figs. 6A–6C) closely coincides with high wave action around those months (Fig. 2A). Shelter-seeking behaviors may be a response to strong hydrodynamic forces because during the calmer months in summer, the occurrence of free-living sea urchins gradually increased. It is known that high water motion reduces urchin movement and foraging behavior due to the threat of dislodgement (Kawamata, 1998; Siddon & Witman, 2003; Cohen-Rengifo et al., 2018). Urchins respond to increasing water motion by escaping from exposed areas and changing their outline to a more streamlined shape (Cohen-Rengifo et al., 2018) or remain sheltered until conditions improve (Yusa & Yamamoto, 1994; Tamaki, Muraoka & Inoue, 2018). In addition, D. savignyi, a closely related Diadematid species shows a similar pattern in preferring crevices in the deep IH but only few individuals were found. In the entire VH, D. savignyi were mostly crevice dwellers throughout the study period. We speculate that although Diadematid urchins (i.e., D. setosum and D. savignyi) occupy the same guild, D. setosum was a better competitor for resources resulting in a skewed population in favor of D. setosum. For H. crassispina, this species generally prefers crevices throughout both habitats except in calm conditions where they may leave their crevices as in the shallow IH but also to a lesser extent in the shallow VH. Very few sea urchins of all species were found in pits along the deep VH and IH (Figs. 6A and 6C) because pit microhabitats were scarce in these areas. Overall, it was clear that the IH was mostly composed of small and medium sized urchins while there was a higher chance of observing larger individuals in the VH, especially for D. savignyi and D. setosum, probably due to the food availability. A study on interactions between Echinometra mathaei, D. savignyi and D. setosum in a coral reef shows that among Diadematid urchins, the body size characteristics and sedentary habits of D. savignyi enabled it to outcompete D. setosum for microhabitat refuge while E. mathaei was the top competitor for crevice microhabitats due to settlement success while benefiting greatly from living in crevices because of high predation pressure compared to Diadematid urchins (McClanahan, 1988). In our study, we suspected that the sedentary habit of D. savignyi was a disadvantage particularly in areas where resources were scarce. In these conditions, active foraging is better suited to maximize encountering food and refuge. Since our surveys were done in the daytime, we were not able to observe urchin foraging at night when they are most active (Tuya, Martin & Luque, 2004). Among the three species, H. crassispina was the exception, because it primarily preferred crevices while rates of being free-living occurred only in the shallow IH where wave action was less. A good example of this scenario is in Hong Kong where H. crassispina form destructive feeding fronts in sheltered bays (Urriago Suarez et al., 2021).

According to historical records, D. setosum was implicated in the loss of seaweed beds in Japan since the late 1800’s (Fujita, 2010) and are regularly culled as a first step in the rehabilitation of barren areas (Nanri et al., 2011; Ohmura, Watanabe & Fujita, 2011). Unlike other species, Diadematid urchins have lesser commercial value in Japan due to the higher ratios of bitter tasting compounds in their gonads (Kaneko et al., 2009) while their optimal season occurs only in June (Kaneko et al., 2012). In our study, sea urchin densities in both habitats were low relative to thresholds necessary to trigger a forward phase shift (Fig. 8), but were enough in some months to maintain the barren state (Fig. 8, shallow VH, August–October) or even support recovery (Fig. 8, deep IH) (Ishikawa, Maegawa & Kurashima, 2016; Kriegisch et al., 2016; Ishikawa & Kurashima, 2020). Sea urchin barrens are known for their inherent stability since thresholds for macroalgal recovery (reverse shift) is different from the one which initiated the shift (forward shift) (Filbee-Dexter & Scheibling, 2014). The environmental conditions in the IH may be contributing to unfavorable conditions for algal growth despite having lower urchin densities compared to the VH.

A meta-analysis of studies on macroalgal bed-to-barren regime shifts found that sea urchin biomass over 668 g m⁻² led to overgrazing while about 34–71 g m⁻² allows for macroalgal recovery (Ling et al., 2015). Based on these estimates, the sea urchin biomass in our transects were at the lower limits of those that were reported. However, the sea urchin biomass in our study could be underestimating the actual values because sea urchin weights were inferred by sampling a few individuals in the spring. Sea urchin reproductive cycles are marked by distinct events that affect their overall biomass. For example, gonadal development and maturation and the high availability of algal food lead to increasing body mass in the winter and spring, while spawning and the low macroalgal diversity in the summer to autumn lead to lower body weights (Kaehler & Kennish, 1996; Horii, 1997; Bronstein, Kroh & Loya, 2016; Urriago et al., 2016). Hence, we suggest considering sea urchin density as an indicator for estimating the risk of phase shifts as biomass is dependent on season. The Arikawa Bay Fisheries Cooperative that manages the fisheries rights of the study area has indicated that sea urchin culling and harvesting has not occurred at our study sites prior to or during this study.

Sea urchin movement patterns

In terms of sea urchin movement patterns, we demonstrated how sympatric sea urchin species behaved differently in adjacent habitats. Generally, both taxa were more active and displaced farther over a 24 h period in the IH where wave action and benthic rugosity was low, compared to the VH. The displacement GLM showed a higher displacement occurring after the second recapture compared to the first recapture period (Fig. 7A), indicating both species were nocturnal. A study on Diadema antillarum, show nocturnal foraging behaviors and site fidelity among the tagged urchins (Tuya, Martin & Luque, 2004). However, we were not able to determine whether D. setosum displayed homing behaviors during our tagging study. We interpret the sea urchin movements as a response to wave action and as foraging behavior. The lower wave action and the relative food scarcity prompted urchins to forage more in the IH, whereas the necessity to forage further was less at VH where waves were stronger and food was relatively abundant. A number of studies document the flexibility of sea urchin diets to include detritus and animal matter when algal food is low (Freeman, 2003; Wangensteen et al., 2011; Rodríguez-Barreras et al., 2015; Umezu et al., 2017; Camps-Castellà, Romero & Prado, 2020). When food was abundant, sea urchins have been known to subsist on drift algae (Kelly, Krumhansl & Scheibling, 2012; Kriegisch et al., 2019; Rennick et al., 2022), reducing the need to forage far from their refuge which indirectly reduces feeding pressure on the existing seaweed beds. It is also known that the threat of predation induces alarm responses among several species of sea urchins (Snyder & Snyder, 1970; Campbell et al., 2001; Morishita & Barreto, 2011). Recent studies have found that sea urchin response to predator cues may vary depending on species. For example, when a dead conspecific was present, Strongylocentrotus intermedius formed aggregations while Mesocentrotus nudus responded by leaving the vicinity (Zhadan & Vaschenko, 2019). There is also some evidence that for species like Echinometra mathaei (Hart & Chia, 1990) and H. crassispina (Belleza et al., 2021), starvation weakens the anti-predator response and prioritizes foraging behaviors, but not in M. nudus (Chi et al., 2021).

As for the patterns observed with the changes in urchin aggregations, the group-size GLM shows that all H. crassispina remained solitary throughout the study. There was a net decrease in D. setosum group-size in both habitats after the first recapture and a subsequent increase in group-size after the second recapture in the IH but not in the VH (Fig. 7B). Furthermore, it is important to note that D. setosum group sizes in the IH were similar while there were large variations in group composition among experimental trials in the VH. We suspect that the high availability of food and refuge in the VH supported higher recruitment survivability and safety from predation, although the level of predation pressure on sea urchins in both habitats remains unclear. It is known that aggregation behavior is one strategy to deter predators (Garnick, 1978) and are maintained by close spine contact between members of the group while increased spine movement indicates defensive activity against a potential predator (Morishita & Barreto, 2011). Nevertheless, tagging activities tend to disrupt aggregations when an individual is taken from the group. The tagging procedure was invasive and involved puncturing a 0.5 mm hole on the sea urchin’s test at the inter-ambulacral region to insert the tag. Even though the tagging injury is small, urchin coelomic fluids may still leak out and evoke a flee response from the surrounding urchins when the tagged urchin is returned to the group (Campbell et al., 2001; Morishita & Barreto, 2011; Spyksma, Shears & Taylor, 2020). This may help explain the reduction in group size in the first recapture. After the second recapture, the tagged D. setosum may have partially healed and retained some group members or have encountered other aggregations while foraging at night. This phenomenon has been observed in the Solomon Islands where D. setosum and D. savignyi aggregations are constantly changing in size, and aggregations of both species often combine (Pearse & Arch, 1969). In the IH, we expect sea urchin aggregations to encounter other groups easily in the flat terrain but not in the VH where abundant crevices and wave action results in more dynamic group interactions. Furthermore, the relative abundance of food in the VH reduces the necessity to forage further.

Past records and future implications

While the importance of sea urchins as key drivers of ecosystem states is generally recognized, it is equally important to note environmental changes that occurred in recent years. A fisheries survey in 1991 shows that VH was once a vibrant algal bed that included five species of Sargassum, three subtropical kelp species and about four species of abalone and several commercially important fish species (Nagasaki Prefecture Fisheries Development Association, 1991). There were no official records of surveys done at IH but remote sensing maps available through the Geospatial Information Authority of Japan (https://www.gsi.go.jp/johofukyu/johofukyu210322.html), shows that IH was once possibly vegetated in the late 1960’s prior to coastal armoring and the construction of the concrete jetty in the early 1970’s (Fig. 9), given its proximity to areas previously surveyed (Nagasaki Prefecture Fisheries Development Association, 1991). Urbanization and coastal development are known anthropogenic stressors that affect shallow subtidal ecosystems (Kevekordes, 2001; Coleman et al., 2008; Coleman & Wernberg, 2017). In particular, coastal armoring in Japan has been used extensively as a response to coastal erosion, sea level rise, and storm surge (Koike, 1996). Ironically, a trade-off occurs as natural habitats are lost for coastal protection. In this case, we speculate that coastal development has contributed to the loss of a seaweed habitat by altering the area’s hydrodynamic conditions leading to increased sedimentation (Airoldi, 1998, 2003). Furthermore, the average winter temperature in Arikawa Bay appear to have increased and remained elevated since 2018 (Fig. 10, GN Nishihara, unpublished data; Supplemental files: Raw table 11). Higher winter temperatures mean sea urchins (Ishikawa & Kurashima, 2020) and herbivorous fish (Yamaguchi et al., 2006; Noda et al., 2016) are able to remain active and grazing pressure may persist throughout the year.