Can Antarctica's shallow zoobenthos ‘bounce back’ from iceberg scouring impacts driven by climate change?

All coastal systems experience disturbances and many across the planet are under unprecedented threat from an intensification of a variety of stressors. The West Antarctic Peninsula is a hotspot of physical climate change and has experienced a dramatic loss of sea‐ice and glaciers in recent years. Among other things, sea‐ice immobilizes icebergs, reducing collisions between icebergs and the seabed, thus decreasing ice‐scouring. Ice disturbance drives patchiness in successional stages across seabed assemblages in Antarctica's shallows, making this an ideal system to understand the ecosystem resilience to increasing disturbance with climate change. We monitored a shallow benthic ecosystem before, during and after a 3‐year pulse of catastrophic ice‐scouring events and show that such systems can return, or bounce back, to previous states within 10 years. Our long‐term data series show that recovery can happen more rapidly than expected, when disturbances abate, even in highly sensitive cold, polar environments.


| INTRODUC TI ON
Ecosystems globally are experiencing an intensification and diversification in the quantity and range of stressors (AghaKouchak et al., 2018;Turner et al., 2020). Anthropogenically introduced stressors including pollution, microplastics and habitat destruction are often highlighted, but of greater, and progressively increasing, concern are climate change intensified natural disturbances such as wildfires, drought or storm events (AghaKouchak et al., 2018).
It is unlikely that the climatic drivers causing the acceleration and intensifying of natural disasters will be reversed, or even mitigated, in the foreseeable future (IPCC, 2014). Although some ecosystems are resilient to disturbance and recover quickly, others are more vulnerable and thought unlikely to revert back to their previous states (Lotze et al., 2011). These vulnerable systems, such as coral reefs, rainforests and polar regions, are often high in diversity but dependent on species characterized by slow growth and low reproduction rates as well as longevity (Lotze et al., 2011). Most ecosystems affected by anthropogenically induced disturbance can recover when conditions improve, for example, if appropriate conservation measures are enacted and anthropogenic pressure is reduced (Jones & Schmitz, 2009). Antarctica is a special case as it remains protected from many direct human impacts by its isolation but is on the forefront of climate-change-driven disturbance, which cannot be mediated by simple measures (Kennicutt et al., 2019;Turner & Comiso, 2017;United Nations, 1959). As such, it is home to some of the least impacted ecosystems-many of which, for example, sub-ice-shelf ecosystems, are already being exposed to open water under the influence of climate warming (Gutt et al., 2020;Ingels et al., 2020).
West Antarctica has sustained sea-ice losses for two decades while the Southern Ocean as a whole has showed unprecedented sea-ice losses since 2015 (Cook et al., 2016;J. Turner & Comiso, 2017). The West Antarctic Peninsula (WAP) has experienced the most severe consequences of this physical change (Gutt et al., 2020;Rogers et al., 2020). These range from increased air and sea temperature and break up of ice shelves to decreases in annual winter sea-ice duration and accelerated glacial retreat. Consequences of reductions in sea-ice are multifaceted and might range from strongly detrimental community destruction via increased ice-scouring events to changes in community composition owing to increased sedimentation as well as be beneficial since reductions in sea-ice are coincident with extended duration of plankton availability to suspension feeders (Henley et al., 2020;Sahade et al., 2015). Because meal processing times in Antarctic suspension feeders are slow, extended blooms result in more meals per season and are more important than peak concentrations of phytoplankton (Barnes, 2017). This is potentially an important mechanism opposing the destructive force of ice-scouring on benthic communities (Barnes, 2017;Sahade et al., 2015).
Reductions in winter sea-ice extent and duration, coupled with glacial retreat, lead to greater mobility of icebergs (Barnes & Souster, 2011). In shallow polar seas, such as the WAP coastal zone this means that icebergs, which can extend hundreds of meters below sea level, scour the seabed more frequently, thereby decimating benthic communities ( Figure 1; Barnes & Souster, 2011;Dowdeswell & Bamber, 2007). Conditions in Antarctica's seas are unique and include glacial cycling, constant low temperatures, extreme seasonality of primary production and high oxygen concentrations among others (Clarke, 2008). So far, data are limited on the life histories of the vast majority of benthic species. The facts we do know seem to confirm slow growth rates, longevity and a low reproductive output as well as prolonged gametogenic cycles, which are heavily dependent on seasonal primary production (Clarke, 1979;Peck, 2018). Infrequent rates of disturbance by iceberg scour, however, have been part of the evolutionary history of shallow polar benthos for millennia.
Intermediate disturbance rates by icebergs, thus, have helped maintain high levels of β-diversity by contributing to patchiness in polar benthic ecosystems (Connell & Keough, 1985;Gutt & Piepenburg, 2003). This means, assemblages in areas generally affected by ice-scour (all coastal habitat lacking permanent ice cover) can be expected to be in various successional stages and to rarely reach maturity. Thus, we ask the questions of how resilient communities inhabiting shallow polar benthic systems are to the predicted intensification of ice-scouring impact and how sensitive near-shore coastal Antarctic systems will be to this particular consequence of climate change?
In normal years, at Ryder Bay, Adelaide Island, on the Antarctic Peninsula, approximately 24% of the shallow seafloor is impacted by ice-scouring events, but exceptionally high rates of ice-scouring (up to 50% per year) were observed along the WAP between 2007 and 2009 ( Barnes et al., 2014). Such a major pulse of large-scale disturbance has the potential to alter community structure, a shift which was observed during the time of the impact ( Figure 1; Barnes et al., 2014). Until now it was unknown whether the affected ecosystems had permanently phase-shifted into new steady states, defined by pioneers and scavengers with high mortality and low competition for space (Barnes & Souster, 2011;Sutherland, 1974) or whether, given suitable conditions, slow recovery was possible. To test this, long-term datasets linking climate-induced change to the state of benthic Antarctic ecosystems are needed. These are, however, rare, because repeated measurements over decades are costly and difficult in such challenging environments.
Here, we present data on the status of benthic assemblages in Ryder Bay (Rothera, WAP) for 23 years, prior to (1997-2006), during (2007-2009) and post (2010-2019), the period of high impact (PHI) caused by high-frequency catastrophic ice-scour disturbance events driven by exceptionally low seasonal sea-ice coverage Barnes & Sands, 2017). We estimated species richness of sessile and mobile epifauna as well as mortality rates and competitive interactions of sessile fauna at four locations within Ryder Bay ( Figure S2). We aimed to answer the question of how resilient benthic polar systems are to dramatic, sustained disturbance events, and how long recovery might take. To our knowledge, no Antarctic ecosystem has been fully monitored from before, during and after recovery, perhaps in part because of the isolation of the continent and the considerable period of sustained, in situ effort required. 2 | ME THODS  Brown et al., 2004;Smale et al., 2008). Another similar array has recently been established further north on King George Island (WAP; Deregibus et al., 2017). IBIS is a grid system laid out in concrete blocks on the seafloor ( Figure S2). Concrete blocks were deployed in nine sets of 5 × 5 m grid systems. Blocks were set 1 m apart (25 blocks per 5 × 5 m grid). Three grids were set-up at each of three different depths (5, 10 and 25 m) in South Cove, close to Rothera Research Station in 2002. Since then, the grids were monitored annually (December-January) to quantify disturbance, and manually replace blocks damaged by icebergs. Thus, annual iceberg disturbance was monitored for 18 years over an area of 225 m 2 . The extent, intensity and frequency of ice-scour at this location are described in detail in Barnes (2017). The different depths of IBIS meant that a true representation of impacts was measured for shallow assemblages. Shallower sites tend to be more often affected by ice-scour, but the impacts are smaller in size . This dataset provides an estimate of the number of how many one m 2 in a 25 m 2 grid have been hit at least once during 1 year in Ryder Bay, as such it is representative of the minimum mean disturbance for each 25 m 2 area or seabed each year.
Duration of annual nano-phytoplankton bloom (chlorophyll concentrations >200 µg L −1 ) and fast-ice duration in Ryder Bay were extracted from data collected for the Rothera Time Series (RaTS) environmental monitoring programme (Venables et al., 2020). Nanophytoplankton was used as a proxy, since many suspension feeders in Antarctica feed on this size group (see Barnes, 2017 and references therein).
Sampling to monitor the proxies for the status of the ecosystem was carried out opportunistically; thus, samples were not taken every year and at a subset of different sites dependent on accessibility (see Table 1). Generally, three random locations within Ryder Bay and one random IBIS grid in South Cove were chosen (N = 4) during each sampling event (Table 1; Figure S1). To prevent any bias of depth stratification on community assemblages, all sites were sampled at ~10 m depth (Barnes & Brockington, 2003). Substratum at all sites was similar and was comprised of a steeply sloping seafloor made of boulders and cobbles interspersed with shallow TA B L E 1 Monitoring a shallow Antarctic location for disturbance and its biological consequences. Data availability varied across years for physical and biological parameters: The duration of fast ice (and phytoplankton above threshold of 200 µg L −1 ) was measured in days year −1 (left) and ice-scour is represented as total ice-scours per total seabed markers (i.e. total/225 m −2 ). Years when the biological assemblage status was sampled are marked by 'x' patches of coarse sediment and mud (Clarke, 1996;Smith et al., 2015; Figure S2). Sampling was carried out during the Antarctic

| Statistical analysis
To To determine whether any other explanatory variables, such as fast-ice duration or extent of phytoplankton bloom, could explain variability in the data we carried out correlations between all variables. We used Pearsons and Spearmans correlation coefficients to account for non-normally distributed data. We found no difference between both coefficients; thus, results represented here have been calculated using the Pearsons correlation coefficient (Table S1). All data analyses were carried out in R version 3.5.2 (R Core Team, 2020).

| Fast-ice cover and frequency of ice-scouring in Ryder Bay
Between 2007
As found previously , the incidence of competition declined during and after the PHI (Figure 2; GLM: F 6,37 = 5.54, p = 0.0004). This continued for several years after, up to 2013. At which point competition was significantly below pre-impact levels.
Hereafter, competition rates started to increase, though they remained statistically similar to levels prior to and during the PHI until 2019. Only then were they different to competition levels during (and even prior to) PHI ( Figure 2).
Throughout the survey, we found 21 taxa of megafauna and 19 taxa of epifauna (Table S2). Mega-and epifaunal species richness were affected very differently by the PHI. Although decreasing, megafaunal species richness was not significantly changed by the PHI. After 2009, megafaunal species richness increased steadily until, by 2019, megafaunal species richness was significantly higher than that observed during the PHI (GLM: F 7,37 = 2.73, p = 0.022; Figure 2).

| DISCUSS ION
Our results highlight a concerning trend across the WAP. Although extreme low sea-ice cover, as those observed between 2007 and 2009, might still remain anomalous, the general trend within the last decade indicates that sea-ice cover in the coastal WAP is likely to decline in extent and duration over the coming years (Ducklow et al., 2013;Rogers et al., 2020). As part of this study, we specifically quantified the impact of this ice loss for benthic ecosystems in Ryder Bay.
The similar set-up at King George Island will allow us to validate the generality of our findings over time . As far as we know, the dominant species and the physical environment of near-shore systems in Ryder Bay seem representative of most high energy shallow subtidal coastal systems along the WAP. Sea-ice duration is spatially complex but the consequences of a reduction in coastal fast ice are likely to be experienced along the complete WAP (Turner & Comiso, 2017).
We expected that most response variables representing benthic ecosystem status would differ between the years before and during the PHI (Barnes, 2017;Barnes et al., 2014;Barnes & Souster, 2011), but that after 10 years the system would start to show signs of recovery (Conlan & Kvitek, 2005). Surprisingly, this was only partly the case. Megafaunal species richness, for example, was not greatly affected by the increased disturbance. This is surprising as, although, megafauna was comprised primarily of mobile species (Table S2), many taxa including anemones and sea-cucumbers move extremely slowly (Arntz et al., 1994). As such, it is unlikely that this failed response is caused by the mobile species ability to 'flee' from the iceberg, but by their capacity to invade recently impacted systems from neighbouring unaffected areas more rapidly than originally thought (Robinson et al., 2020). It seems that, unlike sessile species, even slow mobile species are less dependent on recruitment events for recolonization or dispersal. The presence of carrion and the chance to evade competition (Dunlop et al., 2014;Peck et al., 1999) may stimulate this migration into recently disturbed patches as food availability for some scavengers can also be highly seasonal and thus limiting in Antarctic benthic ecosystems (Dunlop et al., 2014). In fact, previous research has shown that Odontaster validus, a highly abundant, scavenging starfish, will move towards a feeding cue in great numbers within 30 h (Kidawa, 2005;. Further studies have shown that mobile species can recolonize a recently disturbed space within 10 days in sub-Antarctic near-shore systems and within 4 years in the deep sea (Gollner et al., 2017;Peck et al., 1999). Thus, an ability to profit from such disturbance events, for example, by F I G U R E 2 A shallow Antarctic ecosystem before, during and after impact. The high impact period is shown by the light blue box. The duration of fast ice reaches a minimum following 2006 (A). This allows more iceberg movement leading to peak ice-scour (B). The high ice-scour drives a rise in benthic mortality rates (C), depressed competition rates (D) and changes in biodiversity (E). Biodiversity is shown as mean megafaunal (dark brown) and epifaunal (light brown) species richness per m 2 between 1997 and 2019. Grey shading represents 95% confidence intervals. Different letters denote statistically different values increased food abundance in terms of carrion in conjunction with high mobility, such as shown by the majority of Antarctic scavengers including O. validus, sea urchins (Sterechinus neumayeri), a range of amphipod species (Seefeldt et al., 2017), and some primary consumers such as sea cucumbers (Heterocucumis steineni) and limpets (Nacella concinna), seem to be successful traits in areas with high but localized disturbance regimes (Robinson et al., 2020;Seefeldt et al., 2017;Smale, 2008). This means that megafauna appear to be fairly resilient to iceberg impact. However, population structure and densities in affected communities may have changed during the experimental period, which represents an important unknown and caveat to this interpretation (Jones & Schmitz, 2009).
As expected, epifaunal species richness declined significantly during the PHI, which was also reflected in increased mortality rates and a decline in competition rates. Competition for space is one of the main drivers structuring sessile benthic assemblages (Dayton, 1971;Schoener, 1983). Increased space availability, created by the repetitive removal of established attached fauna via icescouring, reduces competitive pressure for primary colonizers such as Fenestrulina rugula, an early successional bryozoan . Intraspecific competition between early colonizers, however, may be just as intense as interspecific competition between later successional stages (Lubchenco & Menge, 1978), which explains a continuing occurrence of competition in years of high ice-scour. In fact, a previous detailed analysis showed that while interspecific competition decreased during the PHI, rates of intraspecific competition, primarily among F. rugula populations, increased . This is not only typical of ecosystems in early successional stages but also a common phenomenon found in higher latitudes (Barnes & Neutel, 2016) where harsh environmental conditions and abiotic perturbations, such as ice-scour, limit the dominance of stronger competitors (Barnes, 2002).
An increase in epifaunal species richness in combination with a rise in competition rates suggests that later successional stages in assemblage development have been reached. A decade after disturbance eased at Ryder Bay, sessile communities appear to have 'bounced back' (sensu Lotze et al., 2011). This result also highlights the sensitivity of sessile species, here primarily cheilostome bryozoa, spirorbinae and porifera, to perturbations. Recruitment of most epifaunal species in Antarctica is highly seasonal and predominantly occurs during the summer season. Bryozoans, however, which contribute one of the most speciose groups of epifauna in near-shore Antarctic systems, start recruitment during late winter beginning with weak competitors followed by stronger competitors during the early summer (Bowden, 2005). However, there seems to be a single short recruitment pulse per species per year (Bowden, 2005). Early colonizers often contribute to ameliorating environmental conditions to later successional stages (Potthoff et al., 2006;Turner, 1983). Thus, increased mortality of new early successional colonizers owing to sustained ice-scour disturbance may have far-reaching consequences for the structure and function of benthic communities (Robinson et al., 2020). Although annual recruitment of sessile species will contribute to the resilience of Antarctic near shore benthic systems, short durations and single occurrence of annual recruitment in conjunction with sessile species inability to evade disturbance make them highly vulnerable to sustained ice-scouring events.
Interestingly, increased phytoplankton duration (food availability for suspension and deposit feeders) was not directly linked with an increase in epifaunal species richness. It seems likely, though, that enhanced phytoplankton duration might aid re-establishment of sessile species by benefiting recruitment success and growth rates of sessile species (Barnes, 2017). Although we were not able to measure this during this study, the question of how far increased duration of phytoplankton abundance ameliorates detrimental effects of ice-scouring on recruitment success of epifaunal species would be a sensible priority for future research to determine drivers underpinning recovery of benthic ecosystems in Antarctica.
With the exception of mortality rate, we found evidence that three of our four descriptors of ecosystem status seem to have reverted to, or increased above, levels prior to impact. The inconsistent decrease in mortality rates after the PHI may be because frequencies of ice-scour (a key driver of shallow benthic mortality; Barnes & Conlan, 2007) still remained higher than pre-impact levels.
This relatively high level of disturbance may have also curbed the recovery of sessile components of communities (Connell, 1978;Jones & Schmitz, 2009).
Marine benthic systems, in general, seem to be very resilient to disturbance. With an average recovery time of less than 10 years globally, many of them are among the fastest ecosystems to recover from perturbations. This is in contrast to, for example, terrestrial forest system which have an average recovery time of 40 years (Jones & Schmitz, 2009). Many benthic systems, such as rocky shores or exposed shallow subtidal benthic systems, including the one stud-  (Clarke, 1979;Potthoff et al., 2006). However, recovery time also depends on the magnitude and frequency of disturbance (Lotze et al., 2011 provide a more optimistic view of potential persistence throughout the realization of these predictions, it is still unclear how far this capacity will stretch in increasingly stressed conditions. Scale dependency of disturbance recovery has already been considered at larger scale in deeper water (Gutt et al., 2011;Gutt & Piepenburg, 2003). Our design incorporated multiple spatial scales from 1 m to hundreds of metres but did not specifically test for the effect of scale on resilience. Thus, the possibility still exists that the ecosystem will shift to a state of greater biotic impoverishment once the disturbance becomes chronic and spatially more extensive (Keck-Al-Habahbeh et al., 2020). This could result in reduced time available for the recovery of the ecosystem alongside a decrease in non-disturbed patches acting as sources for recolonization (Sutherland, 1974).
The extent of this study covered a miniscule spatial area in comparison to the vast extent of the WAP. Thus, the generality of our assumptions needs to be challenged with any new information that is being revealed about WAP's near shore coastal systems in the future. However, this is the only long-term data that show the state of aspects of a physical-biological system, before and during a PHI and at the cusp of recovery in Antarctica. Thus, these data contribute significantly to our understanding of the resilience of Antarctic benthic systems to the effects of climate change in a way that was unknown previously.

ACK N OWLED G EM ENTS
We would like to thank the current and all previous Rothera winter and summer marine teams and the Rothera 2019 wintering team. Team of the British Antarctic Survey.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interest.