Climate‐driven substitution of foundation species causes breakdown of a facilitation cascade with potential implications for higher trophic levels

Climate change can alter ecological communities both directly, by driving shifts in species distributions and abundances, and indirectly by influencing the strength and direction of species interactions. Within benthic marine ecosystems, foundation species such as canopy‐forming macro‐algae often underpin important cascades of facilitative interactions. We examined the wider impacts of climate‐driven shifts in the relative abundances of foundation species within a temperate reef system, with particular focus on a habitat cascade whereby kelp facilitate epiphytic algae that, in turn, facilitate mobile invertebrates. Specifically, we tested whether the warm‐water kelp Laminaria ochroleuca, which has proliferated in response to recent warming trends, facilitated a secondary habitat‐former (epiphytic algae on stipes) and associated mobile invertebrates, to the same degree as the cold‐water kelp Laminaria hyperborea. The facilitative interaction between kelp and stipe‐associated epiphytic algae was dramatically weaker for the warm‐water foundation species, leading to breakdown of a habitat cascade and impoverished associated faunal assemblages. On average, the warm‐water kelp supported >250 times less epiphytic algae (by biomass) and >50 times fewer mobile invertebrates (by abundance) than the cold‐water kelp. Moreover, by comparing regions of pre‐ and post‐range expansion by L. ochroleuca, we found that warming‐impacted kelp forests supported around half the biomass of epiphytic algae and one‐fifth of the abundance of mobile invertebrates, per unit area, compared with unimpacted forests. We suggest that disruption to this facilitation cascade has the potential to impact upon higher trophic levels, specifically kelp forest fishes, through lower prey availability. Synthesis. Climate‐driven shifts in species' distributions and the relative abundances of foundation organisms will restructure communities and alter ecological interactions, with consequences for ecosystem functioning. We show that climate‐driven substitutions of seemingly similar foundation species can alter local biodiversity and trophic processes in temperate marine ecosystems.


| INTRODUC TI ON
Ecological interactions, such as competition, predation and facilitation, play a key role in structuring communities and ecosystems (Kordas et al., 2011;Stachowicz, 2001;Tylianakis et al., 2008). The disproportionate role of foundation species (sensu Dayton, 1972) that alter local environmental conditions and resource availability, often underpinning positive facilitative interactions with other species, has been increasingly appreciated (Bruno & Bertness, 2001;Thomsen et al., 2010). Moreover, such foundation species can also have indirect positive effects on other organisms via cascading interactions (Thomsen et al., 2010). That is, the presence of a secondary habitat-former may elevate biodiversity, but this is itself dependent on the provision or modification of habitat by the primary foundation species (Altieri et al., 2007;Thomsen et al., 2022). These facilitative habitat cascades are prominent across a range of biogeographical contexts, in both terrestrial (Angelini & Silliman, 2014;Cruz-Angón & Greenberg, 2005;Ellwood & Foster, 2004;Stuntz et al., 2002) and marine ecosystems (Bologna & Heck Jr, 1999;Hall & Bell, 1988;Thomsen et al., 2016), but remain poorly described and understood.
Recent rapid anthropogenic warming has led to a global redistribution of ectothermic species as their thermally favourable habitats shift (Poloczanska et al., 2013;Sunday et al., 2012;Thomas, 2010).
Range shifts of foundation species can have disproportionate impacts on communities and ecosystems, given that many associated species are reliant on them directly for habitat and/or food, or indirectly via modification of the local environment (Ellison et al., 2005;Thomson et al., 2015). In some cases, however, vacating foundation species may be replaced by functionally similar species that provide comparable primary habitat to support secondary habitat-formers and, as such, wider ecosystem functioning and biodiversity may be maintained (Bulleri et al., 2018). Conversely, where shifting foundation species are not replaced or replaced by dissimilar species that do not facilitate secondary habitat-formers, habitat cascades may be disrupted with consequences for associated communities and local biodiversity. Evidence of climate-driven disruption to habitat cascades is lacking, despite the recognised importance of facilitative interactions in a warming world (Stachowicz, 2001).
In the North East Atlantic, the cold-adapted boreal species, Laminaria hyperborea, is a dominant habitat-former throughout its distribution from northern Norway to northern Portugal (Figure 1a), particularly on waveexposed shallow subtidal reefs (Kain, 1979;. Atypically among kelp species, L. hyperborea supports abundant epiphytic algal assemblages on its stipe, which serve as secondary habitat-formers (King et al., 2021;Whittick, 1983). This habitat cascade increases living space and complexity for abundant and diverse assemblages of mobile invertebrate fauna, which represent an important food resource for higher trophic levels (King et al., 2021;Norderhaug et al., 2005).
In the United Kingdom (UK), the morphologically similar warmadapted species, Laminaria ochroleuca, is found towards its leading range edge where it competes with L. hyperborea for space and resources. L. ochroleuca, which is distributed from Morocco and the Mediterranean northwards to the UK and Ireland, has proliferated in southwest UK in recent decades (Teagle & Smale, 2018), and is predicted to expand polewards with continued warming (Assis et al., 2017). Unlike L. hyperborea, however, L. ochroleuca lacks the characteristic secondary habitat provided by stipe-associated epiphytic algae . Intuitively, a climate-driven substitution of the cold-adapted L. hyperborea with the warm-adapted L. ochroleuca at some locations could disrupt an important habitat cascade, with implications for local biodiversity and higher trophic levels, but evidence is currently lacking. Crucially, relative to facilitation cascades underpinned by other primary foundation species (i.e. seagrass, mangroves, bivalves), little is known about facilitation cascades in kelp forest ecosystems (Gribben et al., 2019).
Here, we adopted mixed kelp forests in the UK as a model system to test the following hypotheses: (1) climate-driven substitution impact upon higher trophic levels, specifically kelp forest fishes, through lower prey availability.

Synthesis.
Climate-driven shifts in species' distributions and the relative abundances of foundation organisms will restructure communities and alter ecological interactions, with consequences for ecosystem functioning. We show that climate-driven substitutions of seemingly similar foundation species can alter local biodiversity and trophic processes in temperate marine ecosystems.

K E Y W O R D S
climate change, macroalgae, species interactions, temperate reefs, warming of foundation species leads to a weakening of a facilitative interaction; (2) disruption to an important habitat cascade results in impoverished, distinct mobile invertebrate assemblages; and (3) lower abundances of mobile invertebrates potentially reduce prey availability for higher trophic levels (e.g. predatory fish). We thereby explored the consequences of intra-generic replacements of morphologically similar species, differing only in some superficially small traits, but with likely major consequences for community structure and ecosystem functioning.

| Study area
The study was primarily conducted in SW England (UK), which is characterised by shallow subtidal reefs that support extensive kelp forests . Surveys and collections were conducted at four sites within and just outside of Plymouth Sound ( Figure 1c); Drakes Island (site 'A'), Ramscliff Point (B), the northwest Mewstone (C) and Stoke Point (D). The sites spanned a wave exposure gradient, whereby site A was the most sheltered site and site D the most exposed site.
All sites were characterised by extensive gently sloping subtidal rocky reef at depths of 0 to >5 m (below chart datum). Additional survey work was conducted in western Scotland (see Section 2.5 below).

| Kelp canopy structure
The density of the two kelp species was quantified through in situ surveys by scuba diving. At each site, 10 replicate 1 m 2 quadrats were haphazardly placed within dense kelp stands and the densities (inds.

| Sample collection and processing
Five replicate stipe samples were collected from each species in early autumn (i.e. September/October) at each of the four sites (20 samples of each species). Mature, individual canopy-forming plants were selected haphazardly from within well-established kelp stands at depths of 2-4 m (below chart datum). Individual kelp plants were collected from at least two metres apart in areas where the two species were intermixed. Divers removed the blade from selected kelp plants, before immediately enclosing the stipe, its epiphytic assemblage and any associated invertebrates within a fine-mesh cotton bag. The stipe was then removed by cutting immediately above the holdfast and the bag sealed with a cable tie.
Samples were immediately processed on return to the laboratory.
Stipes were carefully removed from the bag and rinsed thoroughly under freshwater to remove all associated mobile fauna; these were collected in a 1-mm sieve and preserved in 70% industrial methylated spirit (IMS) solution prior to identification. All epiphytic algae were then removed from the stipe (any additional animals were removed and added to the sample), and weighed (wet weight) to quantify biomass. Animals were subsequently identified to species level where possible (~74% of taxa) and counted. The age of each kelp plant was estimated by counting the number of annual growth rings in the basal section of the stipe sample (Kain, 1963).

| Manipulative experiment
In July 2019, a field-based manipulative experiment further examined the nature and strength of facilitative interactions between the foundation species (i.e. kelp), the secondary habitat-former (i.e. red epiphytes) and the focal assemblage (i.e. mobile invertebrates). At two sites ('B' and 'C' described above), three treatments were established on bare L. hyperborea stipes: no epiphytes, low density of epiphytes and high density of epiphytes. Initially, an abundant stipeassociated epiphyte, Palmaria palmata, was collected from the field and returned to the laboratory, where it was thoroughly rinsed to remove any associated fauna. Epiphytes were then weighed and sorted into low-density (mean ± SD wet weight biomass = 15.7 ± 1.4 g) and high-density (59.8 ± 2.4 g) treatments, with five replicates for each site placed in separate bags. P. palmata is a common and abundant epiphyte on L. hyperborea stipes and densities (i.e. biomass values) were representative of naturally occurring epiphytic assemblages (King et al., 2021;Teagle & Smale, 2018). For each replicate, multiple P. palmata thalli were carefully tied together at the base, secured with tape and affixed to an open cable tie for easy attachment onto stipes in the field. At the site, divers randomly selected and tagged 15 L. hyperborea plants, which were rigorously cleaned to remove any flora and fauna on stipes. Epiphytic algae were then carefully removed from each bag and cable tied around the bare stipe (towards it base), with five stipes randomly assigned to each of the lowand high-density treatments and five stipes left bare of epiphytes (see Figure S1 for example). The experiment was set up in summer and after 9 days the plants were relocated; a 10 cm section of stipe (which included the epiphytes where present) was then excised and placed in individual labelled bags. On return to the laboratory, the stipe and epiphytes were weighed, and all fauna was passed through a 1-mm sieve, sorted to a coarse taxonomic level and counted.

| Latitudinal comparison between pre-and post-range expansion
To examine the likely impacts of continued range expansion of the warm-water kelp L. ochroleuca into kelp forests dominated by the coldwater L. hyperborea, a space-for-time substitution approach was used by comparing across latitudes a region which has not yet experienced an incursion of the range expanding kelp ('pre-expanded') with the current study region ('post-expanded'). The current northern distribution limit of L. ochroleuca in the UK is in the southwest of England (n.b. an isolated marginal population has been recorded at a higher latitude in Ireland-see Schoenrock et al., 2019), but it is predicted to expand polewards to the north of Scotland by the end of the century (Assis et al., 2017, Figure 1b). As such, the west of Scotland was selected as a 'pre-expanded' region in which to examine and compare facilitative interactions and assemblage structure within kelp forests. Subtidal kelp forests in west Scotland were characterised by similar conditions, reef topography and wave exposure to those in sampled near Plymouth . However, these kelp for- ests are yet to be influenced by range expansion of L. ochroleuca and are dominated by L. hyperborea; they therefore offer an opportunity to compare current pre-warming (i.e. monospecific stands of cold-adapted kelp species) with likely future conditions (i.e. mixed kelp stands following the expansion of a warm-adapted species). The overall density of kelps (L. ochroleuca and L. hyperborea, m −2 ), the total biomass of stipe associated epiphytic algae (m −2 ) and the total abundance of epiphyte associated mobile fauna (m −2 ) were compared between regions, using identical approaches. Two survey sites from within each region were selected for the comparison; all sites were broadly comparable in wave exposure, substrate type, nutrients and light availability, and grazing pressure, but the regions differed in mean sea temperature (by ~2.5°C) and the presence of L. ochroleuca ).

| Fish stomach contents
During the late summer of 2015, fish were sampled from within dense kelp forests in southwest England (at sites C&D) using a combination of fyke nets (set by divers) and rod and line fishing. In total, 21 fish were caught, these were weighed, measured and identified to species, before immediately removing the stomach and digestive tract and transferring them to 70% IMS solution for preservation. Stomach contents were later analysed; the poor condition of many prey organisms allowed for identification only to a high taxonomic level (i.e. order or class) or to a morphological group. The wet weight and abundance of each group was then quantified (abundances of partly digested prey items were estimated from shell fragments, number of antennae, gnathopods, jaws and other conspicuous remains). The guts of two individuals (both Pollachius pollachius, pollack) were entirely empty and were excluded from further analysis. Fish were sampled under a dispensation granted by the Devon and Severn Inshore Fisheries and Conservation Authority (IFCA) and dispatched under a UK Home Office licence granted to MBA; no other permits were required for fieldwork.

| Statistical analysis
All analysis was conducted using univariate/multivariate permutational analyses using the PERMANOVA add on (Anderson et al., 2008) for Primer v7 software (Clarke & Gorley, 2015). Variability in faunal taxon richness and abundance were examined with univariate PERMANOVA using a two-factor design, with kelp 'Species' (2 levels) and 'Site' (4 levels, representing a gradient in wave exposure) as fixed factors. To examine correlations with habitat size, epiphyte biomass was included as a covariable in the analysis. Univariate habitat metrics (i.e. epiphyte biomass and kelp age) were examined using the same model, but without the covariable.
Kelp density data were converted to the abundance of Laminaria ochroleuca relative to the kelp assemblage as a whole (i.e. L. ochroleuca + L. hyperborea), and was subsequently analysed using a one-factor design with 'Site' (4 levels) as a fixed factor. All permutations (4999 under a reduced model) were based on a similarity matrix of Euclidean distances between untransformed data. Pairwise tests were conducted wherever significant main effects of interactions were observed (p < 0.05).
Variability in faunal assemblage structure was examined with multivariate PERMANOVA using a Bray-Curtis similarity matrix constructed from fourth-root transformed abundance data. The model used included the Species and Site fixed factor, with the epiphyte biomass co-variable. Where significant differences in assemblage structure were detected, SIMPER analysis was performed to establish which taxa contributed most to the observed dissimilarity. For all analyses, differences in within-factor dispersion were examined with the PERMDISP routine.
For the manipulative experiment, differences in faunal abundance and richness between treatments were also examined with one-way univariate PERMANOVA (with treatment as fixed factor, each site analysed separately); permutations were based on Euclidian distances between untransformed data. from one another). The surface area of stipes was similar between species, but differed across sites (F 3,39 = 4.14, p = 0.013; Table S1), with stipes of host kelps being generally larger in more wave exposed conditions (site A was statistically different to all others; Figure 2c). The kelp species were broadly similar in age structure, with study-wide mean age of collected L. ochroleuca and L. hyperborea plants being 4.6 ± 0.4 (SE) and 5.7 ± 0.3 (SE), respectively. The total biomass (wet weight) of epiphytic algae differed greatly between kelp species (Figure 2d), as the average biomass of epiphytes associated with L. hyperborea was >50 g per plant at all study sites. In stark contrast, L. ochroleuca was devoid of any epiphytes at all but the most exposed site. The epiphytic biomass associated with L. hyperborea was significantly greater, this pattern being consistent across sites (F 1,39 = 22.51, p = 0.001; Table S1).
The total abundance of mobile invertebrates ranged from 19 to 375 per stipe and 0 to 25 per stipe for L. hyperborea and L. ochroleuca, respectively. Mean abundance values were consistently and markedly higher for L. hyperborea assemblages compared with L. ochroleuca ( Figure 2e). Univariate PERMANOVA detected a significant species × site interaction term (F 3,39 = 3.52, p = 0.021; Table S3), as the magnitude of difference between the kelp species was lower at site D (but still significant, p < 0.01). The main effect of kelp species was highly significant (F 1,39 = 6.89, p = 0.013), as was the epiphyte biomass covariable (Table S3). Taxon richness varied from 8 taxa to 29 taxa per stipe for L. hyperborea stipes, and from 0 taxa (9 stipes) to 9 taxa for L. ochroleuca stipes. Again, mean taxon richness values were notably higher for L. hyperborea assemblages compared with L. ochroleuca (Figure 2f). At site D, for example, L. hyperborea supported ~5.5 times as many taxa as L. ochroleuca. Taxon richness associated with L. hyperborea was significantly higher than L. ochroleuca, which was consistent across sites (F 1,39 = 66.86, p = 0.001; Table S3). Across the survey sites, significant positive relations were recorded between epiphyte biomass and faunal abundance and richness ( Figure S2, Table S4).

Principal component ordination (PCO) plots indicated clear par-
titioning in multivariate assemblage structure between host species, with some small degree of convergence only evident at the most exposed site (site D; Figure 3a). PCO also indicated partitioning between sites, with a gradient evident between the most sheltered and most exposed sites (Figure 5a). PERMANOVA detected a significant site × species interaction (F 3,39 = 2.75, p = 0.001; Table S3), suggesting that the magnitude of difference between species was not consistent between sites. Pairwise tests within the interaction term showed that assemblages associated with L. hyperborea were statistically distinct from those associated with L. ochroleuca at all sites; but the magnitude of dissimilarity was lower at site D. The main effects of species (F 1,39 = 26.73, p = 0.001) and site (F 3,39 = 3.90, p = 0.001) were highly significant (Table S3). The covariable, epiphyte biomass, was significantly correlated with assemblage structure (Table S3), and bubble plots overlaid onto the PCO plot showed how partitioning in assemblage structure was correlated with epiphytic biomass (Figure 3b).
SIMPER analysis was used to identify which taxa were the principal contributors to the observed dissimilarity in assemblage structure between the kelp species. The dissimilarity in mobile invertebrate assemblages between kelp species was consistently related to lower abundances of dominant amphipods and polychaetes on L. ochroleuca stipes compared with L. hyperborea stipes (Table S5).

| Manipulative experiment
The presence and density of epiphytes associated with L. hyperborea stipes had a significant influence on the abundance and richness of mobile invertebrates (Figure 4).  Table S4).

| Latitudinal comparison
The density of kelp plants per square metre was comparable across regions (Plymouth: 10 plants m −2 ; Scotland: 9 plants m −2 ; Figure 5a). L. ochroleuca comprised ~45% of the plants recorded across both sites in southwest England. The total biomass of stipe-associated epiphytes differed markedly between F I G U R E 2 Habitat structure of mixed Laminaria ochroleuca (red bars) and Laminaria hyperborea (blue bars) kelp stands at each site, showing both (a) absolute densities of mature plats and (b) relative densities. Density measurements are means derived from 10 m 2 replicate quadrats at each site. Lower plots show (c) stipe surface area, (d) epiphyte biomass, (e) mobile invertebrate abundance and (f) taxon richness for each kelp species at each site. All values are means derived from five replicate stipes per species, per site (mean ± SE). Sites A-D were situated along a wave exposure gradient, from least to most exposed. regions, with almost twice the amount of algae (per unit area) recorded in Scotland (Figure 5b). Differences in the abundance of mobile faunal assemblages were even more pronounced, with total abundances (per unit area) in the pre-expansion region ~5 times greater than that in the post-expansion region ( Figure 5c).

F I G U R E 3
Principal component ordinations depicting the structure of stipe assemblages (a) and overlaid with bubbles representing epiphyte biomass (b). Red points indicate Laminaria ochroleuca assemblages, and blue points indicate L. hyperborea assemblages. Labels indicate site; sites A-D were situated along a wave exposure gradient, from least to most exposed. Circles enclose samples devoid of any faunal assemblage.

F I G U R E 4
Results of manipulative experiment conducted at two sites to examine influence of stipe epiphyte density of mobile invertebrate (a) abundance and (b) taxon richness. All values are means derived from five replicates per treatment, per site (mean ± SE).

| Fish stomach contents
From the 19 fish stomachs analysed, 10 higher taxonomic groups (class or order) and 6 morphological groups were identified (Table 1). Prey items were largely comparable across fish species. The exception was Labrus bergylta (Ballan wrasse), which fed more heavily on gastropods compared with the other species (Table 1). Generally, the groups observed in the fish stomachs reflected the dominant taxa sampled in the stipe-associated assemblages, with high abundances of decapods, gastropods, isopods and gammarid and caprellid amphipods observed in all instances. Even so, comparisons of the relative abundances of major taxonomic groups between stipe assemblages and fish gut contents showed a higher representation of gastropods and decapods and a lower representation of nemertean and polychaete worms in fish guts, compared with stipe assemblages ( Figure S3).

| DISCUSS ION
Climate-driven shifts in the distribution and abundance of habitatforming macroalgae can significantly alter the structure and functioning of coastal marine ecosystems (Fernandez, 2011;Voerman et al., 2013;Wernberg et al., 2013Wernberg et al., , 2016. The vast majority of evidence to date, however, has focussed on the loss of these foundation species. Our study suggests that climatedriven substitutions of structurally and taxonomically similar species can have significant impacts on local biodiversity patterns, through weakening of a facilitative interaction leading to complete breakdown of a critical habitat cascade, with consequences for higher trophic levels. A recent increase in the abundance of Laminaria ochroleuca relative to Laminaria hyperborea at certain locations, driven by recent warming, has disrupted an important habitat cascade associated with the stipeepiphyte complex, leading to significantly reduced secondary habitat and impoverished faunal assemblages ( Figure 6). Reductions in faunal abundances may limit the quantity and composition of available prey items, with potential impacts on kelp forest fishes ( Figure 6).
Kelp forests support high biodiversity in coastal marine ecosystems (Steneck et al., 2002;Teagle et al., 2017). The secondary habitat provided by epiphytic algae on kelp stipes has been shown to facilitate the development of highly diverse and abundant faunal assemblages (Christie, 1995;Christie et al., 2003;King et al., 2021).
Epiphyte biomass varied dramatically between species, with 90% of L. ochroleuca stipes devoid of any epiphytic material. Across our southwest England survey, the total biomass of epiphytes found in association with L. hyperborea was 1783 g, over 250 times greater than those observed on L. ochroleuca, which supported just 7 g in total and was only present at the most exposed site. Marked variability in the biomass of secondary habitat-formers between host kelp species was consistent across sites, providing strong support for our first hypothesis, that a climate-driven substitution of L. ochroleuca over L. hyperborea weakens a facilitative interaction.
A number of recent studies have compared habitat provision by different species, and in general have shown that different kelps host distinct assemblages (e.g. Blight & Thompson, 2008;Teagle & Smale, 2018;Tuya et al., 2011). Here, despite offering similar habitat structure (e.g. stipe surface area, plant age), mobile invertebrate assemblages associated with L. hyperborea and L. ochroleuca stipes differed markedly. Of the 2430 individuals recorded across the study, over 98% were found in association with L. hyperborea.
On average, each L. hyperborea stipe hosted ~120 individuals representing ~17 taxa, whereas L. ochroleuca hosted just over two individuals representing ~1.5 taxa per plant. The marked difference in epiphyte biomass observed between these kelp species underpins the observed variation in associated invertebrate assemblage structure, as demonstrated by our manipulative experiment. As such, our second hypothesis, disruption of a habitat cascade leads to impoverished, distinct mobile invertebrate assemblages associated with L. ochroleuca, was strongly supported. Given that stipe assemblages associated with L. hyperborea in southwest England are comparatively lower in abundance and richness than those further north in Scotland and Norway (e.g. Christie et al., 2003;King et al., 2021), replacement by L. ochroleuca in other regions (albeit not predicted until the end of the century, see Assis et al., 2017) could lead to even greater shifts in stipe assemblage structure and diversity.
Ecological interactions are mediated by the environmental context in which they occur (Harley et al., 2012;Sanford, 1999). In the current study, the strength of the facilitative interaction between kelp plants and stipe-associated epiphytic algae differed between sites along a wave exposure gradient. With regards to L. hyperborea, the biomass of epiphytes and the taxon richness (and to a lesser extent, the total abundance) of associated invertebrate assemblages all followed a similar pattern, with higher values generally recorded at the more wave -exposed sites (i.e. C and D) compared with more sheltered sites (i.e. A and B). The importance of wave exposure in structuring epiphytic algal and invertebrate assemblages associated with L. hyperborea has been demonstrated previously (Andersen, 2007;Christie et al., 2003;Norderhaug & Christie, 2011).
Greater epiphyte biomass at higher wave exposure levels was likely driven by a range of factors, including (i) increased light penetration through the overlying canopy with greater water movement (e.g. in Macrocystis pyrifera forests; Wing et al., 1993); (ii) more efficient transfer of nutrients across algal surfaces under greater water motion, leading to increased growth rates ; (iii) greater availability of biogenic habitat for colonisation at higher wave exposures, due to morphological responses of the kelps themselves (Pedersen et al., 2014;Smale et al., 2016); and reduced grazing pressure at higher wave exposures Taylor & Schiel, 2010). Only at the most exposed site did L. ochroleuca host any epiphytic algae, at low biomass, being devoid of epiphytes at other sites in contrast to L. hyperborea. It is clear, even with site-level variability in epiphyte biomass and assemblage structure for both species, that the facilitative interaction between kelp and epiphytic algae is consistently much stronger for L. hyperborea compared to L. ochroleuca, regardless of environmental context, which provides further support for our first hypothesis. This is despite both congeneric species being morphologically and functionally very similar, differing in minor traits such as stipe rugosity. Clearly climate-driven substitutions of foundation species that differ in seemingly minor ways can have consequences for community structure and ecosystem functioning.
It has long been recognised that kelp forests are important for many coastal fish species, which utilise these habitats as nursery and feeding areas and as refugia from predation (Bodkin, 1988;Norderhaug et al., 2005;Reisewitz et al., 2006). Kelp forest extent and structure have been positively linked with local fisheries production (Bertocci et al., 2015). It is clear that some fish species, despite their high mobility, rely heavily on kelp-associated fauna and feed extensively within kelp forests (Fredriksen, 2003;Leclerc et al., 2013). Despite the low taxonomic resolution of our examination of fish gut contents, the main groups recorded aligned closely with those found associated with the stipe-epiphyte complex of L. hyperborea, with a number of well represented taxa being known to constitute a high proportion of the diets of kelp forest fish (e.g. Jassa spp., Ampithoe spp., Caprellid amphipods, and Rissoa spp. gastropods, see Norderhaug et al., 2005).
Moreover, the accessible nature of the stipe-epiphyte complex to predatory fish, in contrast to holdfast or epilithic algal assemblages for example, and the high abundance of fauna within these assemblages suggests that they enhance prey availability for kelp forest fish. Indeed, previous work has shown that faunal assemblages associated with L.
hyperborea stipes tend to exhibit far higher abundances than those associated with holdfasts or understorey algae (e.g. Christie et al., 2003;King et al., 2021;Teagle et al., 2017). We cautiously infer, therefore, that fish were feeding within the kelp forests they were sampled from, if not exclusively, and that a breakdown of a facilitation cascade may reduce food availability for higher trophic levels. This provides some evidence to support our third hypothesis that shifts in the structure or abundance of mobile invertebrate assemblages have the potential to impact upon higher trophic levels. However, further work is required to reveal trophic pathways within these systems and to determine the magnitude of impacts caused by shifts in the relative abundances of dominant habitat-formers.
Shifts in the structure and functioning of kelp forests in the northeast Atlantic are predicted to occur throughout the coming century in response to ocean warming (Assis et al., 2017;Brodie et al., 2014;Smale, 2020). By comparing the current state of kelp forests across latitudes that include regions where L. ochroleuca has proliferated and regions where it does not yet occur, we can gain an understanding of the likely impacts of future range expansions. While the total densities of kelp plants were similar between both study regions, almost half of the assemblage recorded at the lower latitude comprised L. ochroleuca. It should be noted that these habitats were targeted as areas of mixed kelp stands and are not representative of wider open wave-exposed coastlines, TA B L E 1 Results of fish gut content analysis from fish caught within kelp forests in SW England. Mean values (±SE) for total length (TL) and wet weight (WW) are given below each species. Values for prey items are biomass (g) with abundance values in parentheses

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest. D.A.S. is an Associate Editor of Journal of Ecology, but took no part in the peer review and decision-making processes for this paper.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/1365-2745.13936.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data on kelp stipe assemblages can be found at https://doi. org/10.6084/m9.figsh are.19902 595.v2 (Smale, 2022).

O RCI D
Dan A. Smale https://orcid.org/0000-0003-4157-541X Nadia Frontier https://orcid.org/0000-0003-0189-1282 Pippa J. Moore https://orcid.org/0000-0002-9889-2216 F I G U R E 6 Habitat cascades in kelp forests in the Northeast Atlantic, pre-and post-expansion of the warm-water kelp Laminaria ochroleuca. An increase in the abundance of L. ochroleuca relative to L. hyperborea is likely to reduce the extent of secondary epiphytic habitat available for colonisation, and therefore reduce the abundance of faunal assemblages. Higher trophic levels, such as predatory fish, may be subsequently affected.