Recovery of macrobenthic communities in tidal flats following the Great East Japan Earthquake

The Great East Japan Earthquake, with a moment magnitude scale of 9.0M on March 11, 2011, caused huge tsunamis that altered aquatic communities in the nearshore ecosystems along the Pacific coastline of Sendai Bay, Japan. However, to understand the ecological implications of the tsunamis' impact, an assessment of the successive changes of the communities and their fate over time is essential. Therefore, we examined eight tidal flat communities of Sendai Bay with the aid of citizen volunteers over 9 years, and compared the microbenthic compositions to those before the earthquake. The results revealed that the differences in the community compositions caused by the tsunamis decreased from year to year and finally disappeared after 7 years except for one flat where the tsunamis largely changed the environmental conditions. The present study indicates that these tidal flat communities are regulated mainly by environmental conditions and are highly resilient to tsunami disturbance.

one flat where the tsunamis largely changed the environmental conditions. The present study indicates that these tidal flat communities are regulated mainly by environmental conditions and are highly resilient to tsunami disturbance.
Unusual and extreme geo-climatic events such as heatwaves, drought, hurricanes, and tsunamis are unavoidable large spatial scale disturbances affecting ecosystems (Orr and Ogden 1992;Hanson and Weltzin 2000;Anderegg et al. 2013;Ponce-Campos et al. 2013;Ummenhofer and Meehl 2017;Maxwell et al. 2019). The Great East Japan Earthquake, with a moment magnitude scale of 9.0M on March 11, 2011, the fourth largest earthquake since 1900 (United States Geological Survey 2019), was one of these disturbances (Mimura et al. 2011;Mori et al. 2012;Ranghieri and Ishiwatari 2014;Urabe and Nakashizuka 2016). The earthquake caused huge tsunamis that struck along a large area of the Pacific coastline of eastern Japan. Inundation heights of the tsunamis were often larger than 10 m, with the highest run-up wave recorded at 40 m (Mori et al. 2012). Accordingly, more than 24,000 people were reported missing or killed (Mimura et al. 2011). Studies have described how the tsunamis suddenly changed the community composition of aquatic organisms both in offshore and nearshore ecosystems on the Pacific coast of eastern Japan (Seike et al. 2013;Urabe et al. 2013;Kanaya et al. 2015, see review in Urabe and Nakashizuka 2016;Urabe 2021) and found that the changes in the community composition were greater at sites that experienced tsunamis with higher inundation heights (Urabe et al. 2013).
To understand the ecological implications of geo-climatic events, however, the fate of the successive changes in a community over time is essential (Mittlebach et al. 2006;Lake et al. 2007;Borja et al. 2010;Smith 2012;Matthews et al. 2013). For example, suppose that a community disturbed by tsunamis eventually recovers to its original state. In that case, the community has high resilience, and the event was less critical in the ecological context. However, if the community moves to a state that was not previously observed, the event is judged to be critical in pushing the community to an alternative state (Scheffer et al. 2001;Ratajczak et al. 2018;Pausas and Bond 2020). These possibilities imply that a longterm monitoring of aquatic communities is essential to assess the ecological implications of the tsunamis caused by the Great East Japan Earthquake.
In the present study, therefore we analyzed annual changes in the taxonomic composition of macrobenthic communities for eight tidal flats in Sendai Bay located along the Pacific coast of east Japan over 9 years, utilizing citizen volunteers to aid in the examination. The results were then compared to the compositions that were present before the earthquake. We focused microbenthic animals since these organisms play functionally pivotal roles in the energy and material flows of intertidal ecosystems (Kuipers et al. 1981;Otani et al. 2010), are prey for migratory birds and seabirds (Iwamatsu et al. 2007;Mu and Wilcove 2020) and represent key organisms that reflect the ecosystem health (Borja and Tunberg 2011;Liu et al. 2019).

Monitoring sites
We conducted biological monitoring in the tidal flats at Sokanzan, Matsukawaura-Unoo, Gamo A and B, Torinoumi A and B, Sokanzan, Hitsugaura, and Katsura Is. (Fig. 1) where there were data on the taxonomic composition of macrobenthic community in years before the earthquake (Urabe et al. 2013). In Gamo, a 10-ha lagoon where two separate tidal flats occur, we examined each of these flats separately (Gamo A and B). Similarly, in the Torinoumi lagoon of ca. 60 ha, we examined two different tidal flats (Torinoumi A and B). The exposed areas of these flats ranged from 1 to 5 ha, respectively (Supplementary Table S1). These flats received tsunami impacts of varying degrees (Supplementary Table S1, Urabe et al. 2013).

Monitoring method
We used taxon richness and composition of tidal flats communities at each flat before the Great East Japan Earthquake from data recorded in a previous study (Urabe et al. 2013). Since the monitoring surveys before the tsunami event were originally done with different objectives, the census methods differed in some places from those done after the tsunami, as described below. However, total taxon richness did not significantly differ between the methods as shown in a previous study (Urabe et al. 2013).
The post-tsunami surveys were conducted once a year (between May and August) at the low tide from 2011 to 2019, accomplished with the aid of citizen volunteers. Since the biological monitoring in large areas was laborious and needed staffing, we performed the monitoring surveys using a previously described citizen-science method (Suzuki and Sasaki 2010;Nishita et al. 2016). For 9 years after the earthquake (i.e., 2011-2019), a total of ca. 500 citizen volunteers joined in the monitoring surveys. Details of this monitoring method utilizing citizens and its validity have been shown elsewhere (Suzuki and Sasaki 2010). In short, a total of 12 citizen volunteers joined a monitoring survey for each year at each site. Since we did not know a priori their experiences and skills in field works, we randomly assigned the volunteers to each survey. In the monitoring survey, the 12 volunteers walked randomly for 15 min over the entire area of a flat and collected any epibenthic animals (epifauna) they encountered. Then, using a shovel, each volunteer dug 15 holes, approximately 15 cm in diameter and 20 cm in-depth, and visually searched and collected all endobenthic animals (infauna) found. We examined macrobenthic animals that could be identified by the naked eye. The animals were Yuhara et al.
Tsunamis' impacts on tidal flat communities identified at the finest taxon level possible. Since one fixed researcher, a trained expert in taxonomy, made all the taxon identifications throughout the study, classification consistency was the same regardless of sites or years. Although the volunteers differed among the monitoring surveys, any difference in the overall research performance was minimized by integrating the efforts of 12 volunteers (Suzuki and Sasaki 2010) as follows.
In each survey, data obtained by the 12 volunteers were pooled to determine the taxon richness and composition of the community. We did not use to evaluate the relative abundance of each animal taxon. Instead, we analyzed binary data of the taxa observed in the flats. All of the data used in this study have been deposited in the Dryad database (Urabe et al. 2022).

Statistical analysis
To examine significant difference in the taxon richness (TR) between the pre-tsunami period and yearly for 9 years after the tsunamis, we performed a generalized linear mixed model (GLMM). Assuming a Poisson distribution, we constructed the model with a log-link function as follows: where Y and F are the observation year and tidal flat, respectively, and are used as categorical variables. The observation year was treated as a fixed effect and the tidal flat as a random effect to exclude the potential influences of site-specific conditions. In this analysis, data in the pre-tsunami period were pooled as a single year and were reflected by the intercept. Then, we examined whether the coefficient for Y in the model differed significantly from zero by Wald's test (Venables and Dichmont 2004). This analysis was done using the "glmmML" package in R ver.3.6.2 (R Core Team 2019).
To quantify the difference in the community's taxon composition among different sites and years, we estimated the Jaccard index among these. The results were visualized on a two-dimensional plane, with the aid of nonmetric multidimensional scaling (nMDS), in which observations with smaller dissimilarity indices were plotted closer to each other and vice versa (Legendre and Legendre 1998). Stress value, which represents the divergence of the real value from the ordination output, was calculated to assess the statistical validity of the analysis. The analysis is statistically meaningful if the mean stress value is lower than 0.2 (Clarke 1993). According to Graffelman and Tuft (2004) and Borcard et al. (2011), the contributions of each taxon to the ordinations were assessed using the weighted average (WA) of the species scores, which were estimated using the "wascores" function in vegan package (Oksanen et al. 2017) for R ver.3.6.2 (R Core Team 2019). Among taxa showing the WA score < À0.5 or > 0.5, we plotted those that appeared > 20 times across all years and sites on the nMDS plane.
Significant differences in the taxon composition at each tidal flat before and after the 2011 tsunamis were examined

Yuhara et al.
Tsunamis' impacts on tidal flat communities using a null model analysis (Gotelli and Graves 1996). For this analysis, we first calculated the Euclidean distance of the nMDS plots between a pre-tsunami year and each of the nine post-tsunami years, and denoted it as observed similarity distance at each post-tsunami year. Then, we generated a data set for the null model by randomly replacing data using a "Fixed-Equiprobable model" (Gotelli 2000), in which row sums (i.e., total appearance number of given species) were held constant, and the column sums (total number of species per given sites) were allowed to vary with equiprobability. This procedure ensured that all of the monitoring sites shared a common property such that we had the same chance of discovering a species with a fixed probability of appearance. Using the generated data, we performed nMDS and estimated the Euclidean distance of the plots between the pre-tsunami year and each of the nine post-tsunami years as above. We repeated this procedure 1999 times. Then, the observed Euclidean distances between pre-tsunami and posttsunami years were compared against the generated distances. If the observed similarity distance was larger than 95% of the generated distances, we concluded that the community structure observed in a given post-tsunami year was significantly different from that in the pre-tsunami year. The alpha level was corrected according to Holm's sequential Bonferroni procedure in this analysis. We performed these statistical tests with the statistical package R ver.3.6.2 (R Core Team 2019).

Results and discussion
The taxonomic composition before the tsunamis Before the tsunami struck in 2011, we found 107 taxa of macrobenthic animals present across the entire eight flats. The taxonomic richness varied from 27 to 47 depending on the flats (Fig. 2). The taxonomic compositions also differed largely among the tidal flats (Nishita et al. 2016). To visualize these differences, we plotted the taxonomic composition of different sites in a two-dimensional ordination by nonmetric multidimensional scaling using the Jaccard similarity (Fig. 3). In the pre-tsunami period, the taxonomic compositions of Gamo A and B and Torinoumi A and B fell on the negative end of the first dimension, while those of Katsura Is., Sokanzan, Hitsugaura, and Matsukawaura-Unoo were on the positive end (Fig. 3a). The macrobenthic animals that prefer relatively low salinity and salt marshes, such as the snail Assiminea spp., the polychaete Hediste spp., and the marsh crabs Helice tridens and Deiratonotus cristatus, gathered on the negative end of the first dimension, while animals that prefer relatively high salinity, such as the snails Monodonta confusa, and Batillaria multiformis were clustered on the positive end of the first and second dimensions (Fig. 3b,c). The results indicate that each tidal flat was occupied by a taxonomically different microbenthic community.

Impacts of the tsunamis on the taxonomic composition
Immediately after the tsunamis, > 50% of taxa observed before the tsunamis disappeared in Matsukawaura-Unoo, Gamo A and B, and Trinoumi A and B, where tsunamis were reported with inundation heights > 6 m (Fig. 2). Accordingly, the taxon richness was, on average, significantly decreased in 2011 (Supplementary Table S2) as reported previously (Urabe et a. 2013). The taxonomic composition also changed significantly from those in the pre-tsunami years in these flats except for Gamo B (Fig. 3d,e; Supplementary Table S3). For example, at Matsukawaura-Unoo, the polychaete Scoletoma nipponicas and the barnacle Fistulobalanus albicostatu were found before the tsunami but not observed in 2011. At Gamo B, which received tsunamis with inundation highs > 6 m, no significant difference was detected in the taxonomic composition between the pre-tsunami year and 2011 (Fig. 3d,e). Contrastingly, at Sokanzan, Hitsugaura and Katsura Is. where tsunamis were reported to have inundation heights < 3 m, the taxon richness increased immediately after the tsunamis because of the appearance of taxa typically living in subtidal zones, such as the polychaete Marphysa sp. and the swimming crab Charybdis japonica (Urabe et al. 2013). In these flats, however, several species found before the tsunamis, such as the snail Littorina brevicula and the bivalve Nuttallia japonica were rarely found for several years after the tsunamis.
For the 9 years after the tsunamis, we found 102 out of the 107 taxa that appeared in the pre-tsunami years. In addition, we found 110 other taxa through the flats after the tsunamis. However, the appearances of most taxa that were found only in the post-tsunami years were temporally limited (Urabe et al. 2022). Accordingly, although the taxon richness oscillated somewhat depending on the years, it was within the level of the pre-tsunami years for several years after 2014 in all the flats (Fig. 2). However, in some flats such as Katsura Is. and Matsukawaura-Unoo, the taxon richness tended to exceed 60 after 2017. Due to this increase, the mean taxon richness among the flats in 2018 and 2019 was slightly but significantly higher than in the pre-tsunami year (Supplementary Table S2). The taxonomic compositions also differed significantly from pre-tsunami years for several years after the tsunamis. However, these compositions approached the original states slowly, year by year, in most of the flats (Fig. 3d,e). According to the randomization test, a significant difference was no longer detected in taxonomic composition between the pre-tsunami years and 2019 in all the flats except for Gamo B (Supplementary Table S3). The results indicate that the microbenthic communities disturbed by the tsunamis returned to their original states within 9 years in most tidal flats.

Recovery and resilience of the tidal flat communities
The present results imply that the ecological impacts of the tsunamis were not so large as to shift the communities into different stable states. Rather, the tidal flat communities of Yuhara et al.
Tsunamis' impacts on tidal flat communities the coastal ecosystems were shown to be highly resilient to tsunami disturbance. Since most marine macrobenthic invertebrates have planktonic larval stages, they can disperse spatially between remote habitats (Günther 1992;Van Colen et al. 2008;Rao et al. 2020). Thus, if a taxon reminds in at least some habitats within or near Sendai Bay after the tsunami disturbances, it had the possibility of migrating to and recolonizing appropriate tidal flats within the Sendai Bay study sites. By utilizing the dispersal processes made possible by the planktonic stages, the macrobenthic community could recover rapidly from the large-scale disturbances caused by the tsunamis in the tidal flats examined. Interestingly, although the taxonomic composition returned to the original state, the taxonomic richness in Katsura Is. and Matsukawaura-Unoo increased after 2016 (Fig. 2). This trend suggests that the habitat conditions of microbenthic animals in these flats may have been rather improved relative to the pre-tsunami conditions over the several years after the tsunamis.
The directional return to the original state found in most communities suggests that the taxonomic composition was determined mainly by niche-based processes, i.e., regulated by the environmental condition in each flat. This inference was supported by the community trajectory at Gamo B. In this tidal flat, the taxonomic composition did not differ significantly between the pre-tsunami year and the successive 2 years after the tsunamis but changed significantly after 2013. Although dense reed marshes surrounded this site before 2011, almost all marshes were lost when the tsunami struck (Kanaya et al. 2015). Accordingly, in Gamo B, the snail Nozeba ziczac, the bivalve Limecola contabulata, the polychaete Tylorrhynchus osawai, and the sesarmid crab Chiromantes dehaani, common in the reed mashes, were never found after 2013, although these animals were abundant before the tsunamis and often found in 2011 and 2012 when the dead branches of reeds lingered sporadically. Instead, the bivalve Ruditapes philippinarum, the crustacean Ligia cinerascens, and the shore crab Hemigrapsus spp. that occurred commonly in Gamo A, a sandy mudflat without reed mashes, were also observed in Gamo B after the tsunamis. Due to such changes, the taxonomic composition of the macrobenthic communities in Gamo B gradually becomes similar to that in Gamo A. A similar result was found in 2004 on the Andaman Sea coast of Thailand, where the community composition of the macrobenthos fundamentally changed after tsunamis caused by the Indian Ocean earthquake, especially where the seagrass beds were swept away on a large scale by the tsunamis (Whanpetch et al. 2010). These results support the idea that the macrobenthic community in the tidal flat is mainly determined by the environmental conditions that include sediments, bottom substrate, and surrounding vegetation (Okuda et al. 2010;Moritz et al. 2013).
However, the environmental conditions are probably not the sole factor regulating the taxon composition of the macrobenthic community. Similar to Gamo B, the tsunamis had no immediate ecological impact on the macrobenthic community in Sokanzan that received tsunamis with inundation heights < 3 m. However, in this flat, the taxon composition changed significantly from 2013 to 2016. This fact suggests that changes in the taxon composition of macrobenthic communities in this flat cannot be attributable to environmental conditions. This flat is placed at the inner most region of a bay, and stochasticity, such as changes in the balance of immigration and extinction rates of animal taxa, may have affected the community composition, as has often been

Yuhara et al.
Tsunamis' impacts on tidal flat communities shown in other communities (Hubbell 1997;Heino et al. 2015). However, such a stochastic effect on the taxonomic composition seems to be relatively small when compared to niche limitation. Indeed, after 2017, the macrobenthic community in the Sokanzan flat no longer showed taxonomic composition differences from that of the pre-tsunami composition.
A recent review showed that if a disturbance is a physical event, a coastal community could generally recover in less than 10 years after the termination of the disturbance, although recovery can often be much slower if the disturbance is of a chemical nature, such as a wastewater discharge (Borja et al. 2010). The present result shows that the taxonomic compositions of macrobenthic animals in the tidal flats returned to their original states within 9 years after the tsunamis disturbance if the environmental conditions were unchanged or quickly recovered. Thus, recovery time of the tidal flat communities was on the same level as those disturbed by local physical events, such as hydrological alternations (Borja et al. 2010). These results imply that the macrobenthic community has high resilience to physical disturbances and that the ecological impact of the Great East Japan Earthquake was not large enough to push the community to an alternative stable state. In this study, we examined the taxonomic composition of microbenthic animals in the tidal flats. If we had considered the relative and absolute abundance of these animals, the recovery time of the tidal flat community may have been longer than what was found in this study.
After the tsunami disaster, a total of 432 km of seawalls varying in height from 7 to 14 m have been constructed along the coastline of the Pacific Ocean in the northeast areas of Japan in an effort to reduce the social damages that may occur with future tsunami disasters. However, it is unclear whether and how the constructions of these seawalls will affect the macrobenthic communities in the coastal ecosystems. Although the tidal flat communities have almost all returned to their original pre-tsunami states, will these results be the same in the future or will there be a seawall effect? To assess the effects of the constructed seawalls on the coastal communities, continual monitoring studies will still be needed. . The trajectory of annual changes in the community compositions before and after the tsunamis for Katsura Is., Hitsugaura, Matsukawaura-Unoo, and Torinoumi A and Torinoumi B, Sokanzan, Gamo A and Gamo B (d) are presented in panels (d) and (e), respectively. The number in the circle symbols in panels (d) and (e) represent the last two digits of the monitoring year after the tsunamis. In each flat, monitoring years showing significant differences in the community composition from that in the pre-tsunami year (square symbol), examined by a randomization test, are denoted by the filled symbols.