Macroalgae maintain growth outside their observed distributions: Implications for biodiversity‐ecosystem functioning at landscape scales

The spatial insurance hypothesis states that at landscape scales with environmental variability between different places, biodiversity increases ecosystem functioning if species respond asynchronously to environmental variation, and they are highest functioning in places where they dominate the relative abundance. Under this hypothesis, observed species turnover between places in a landscape with environmental variation might suggest that species diversity is an important driver of landscape‐scale ecosystem functioning. However, the spatial insurance implicitly assumes that species found in one place in the landscape will not be able to maintain high functioning in other places. Given this assumption of the spatial insurance hypothesis, we would predict that species' functioning in monoculture should decline if transplanted to a different place on an environmental gradient, away from where they naturally dominate. If this is the case, we would expect that the loss of one species in one place of the environmental gradient could not be compensated through the establishment of another species in that place. We tested this prediction using a model system of marine macroalgae on intertidal rocky shores on the Swedish west coast. We performed a reciprocal transplant experiment with adult individuals of four fucoid seaweeds that dominate the standing stock biomass at different depths on these shores and monitored their relative growth rate over 2 months. Counter to the assumptions made by the spatial insurance hypothesis, growth rates for three of the four species showed limited responses to being transplanted to different depth zones. Spatial insurance may, hence, play a minor role in sustaining landscape ecosystem functioning in this and other systems. Synthesis. The functional consequences of species loss at landscape scales may not be as obvious as observational studies and ecological models suggest. Consequently, natural patterns of species turnover should not be used directly to argue for the role of biodiversity at landscape scales. Instead, how species may or may not be able to compensate for the loss of other species is a critical aspect if we are to understand how changes in biodiversity at the landscape scale affect ecosystem functioning.


| INTRODUC TI ON
Ecologists have extensively studied how species loss affects ecosystem functioning at small spatial scales with limited environmental heterogeneity (Cardinale et al., 2012;Tilman et al., 2014). At these scales, biodiversity tends to increase ecosystem functioning in a local place either by increasing the probability that the place includes a single high functioning species or by positive interactions among species outweighing negative interactions (Cardinale et al., 2011;Hagan et al., 2021). However, we still know relatively little about the effects of biodiversity on ecosystem functioning at landscape scales with high environmental variability (Gonzalez et al., 2020).
At landscape scales, high biodiversity may be required to ensure maximal rates of ecosystem functioning because species that are high functioning in one place are low functioning in another. This general idea that biodiversity increases and maintains ecosystem functioning in landscapes with environmental variability has been formalised as the insurance hypothesis (Loreau et al., 2003). Specific models of the insurance hypothesis tend to make two central assumptions: (i) species respond asynchronously to environmental variation in space and time in the absence of interspecific competition (i.e. in monoculture) and (ii) these asynchronous species responses allow different species to dominate (i.e. attain high relative abundance) in the places and/or times where they are highest functioning in monoculture (Loreau et al., 2003;Shanafelt et al., 2015). This leads to species turnover among places and times, as different species drive functioning in different places and at different times. In addition, species loss is predicted to decrease landscape-scale ecosystem functioning because no species can fully compensate for the loss of another because species differ in their responses to environmental variation.
If we translate the ideas from insurance models into patterns in natural systems and only consider environmental variation in space, we can make two predictions. First, we expect that dissimilarity in species composition between places (i.e. spatial beta diversity) should positively correlate with ecosystem functioning averaged across places (Hautier et al., 2018). In general, the effect of beta diversity on functioning in studies that test this prediction is relatively weak and context-dependent (Hautier et al., 2018;Omidipour et al., 2021;Reu et al., 2022). However, testing if beta diversity correlates with functioning averaged across places provides a relatively weak test for at least two reasons. First, models of the insurance hypothesis make predictions about species turnover (i.e. species replacement) between times or places (Loreau et al., 2003;Shanafelt et al., 2015;Yachi & Loreau, 1999), but most empirical studies have focused on overall beta diversity. This focus is inappropriate because beta diversity consists of both species turnover and nestedness (i.e. differences in species richness), and these components of beta diversity correspond to different biological phenomena (Baselga, 2010;Omidipour et al., 2021). Second, the models of the insurance hypothesis refer to biodiversity effects on ecosystem functioning at landscape scales. However, some studies have instead tested how average beta diversity between a focal place and all other places in the landscape affects functioning, which is a local-scale comparison (Lefcheck et al., 2019;Maureaud et al., 2019).
The second prediction about patterns in natural systems that we can make is that the identity of the species that contributes most to functioning differs between places . Analyses of both experimental (Isbell et al., 2011) and observational data (Genung et al., 2020(Genung et al., , 2023Winfree et al., 2018) have found support for this pattern. However, just because different species contribute to functioning in different places does not mean they are doing so where or when they are high functioning in monoculture. Moreover, it does not mean that the decrease in landscape-scale functioning due to species loss in one place could not be compensated by a different species occupying that place. Intuitively, we would expect different species to maximise functioning in different places (relative to other species), but we cannot simply make this assumption.
For example, traits that lead to high competitive ability may cause low monoculture functioning (e.g. negative selection effects ;Jiang et al., 2008). Moreover, in stochastic environments, species dominance can be decoupled from species traits (Hubbell, 2001), and for some ecosystem functions or services, species dominance may not be a strong determinant of magnitude (Dee et al., 2019).
Rather than studying observed patterns in natural systems, a stronger test of models of the insurance hypothesis would be to directly test whether species naturally dominate functioning in certain places along an environmental gradient where they are high functioning in monoculture. We performed such a test using a model intertidal marine macroalgal system from sheltered rocky shores on the Swedish west coast. The intertidal in this region is often dominated by four fucoid brown algae species that tend to occupy four relatively distinct depth zones from high to low shore (Cervin et al., 2004; see Figure 1; Table S1). The upper limits of these Consequently, natural patterns of species turnover should not be used directly to argue for the role of biodiversity at landscape scales. Instead, how species may or may not be able to compensate for the loss of other species is a critical aspect if we are to understand how changes in biodiversity at the landscape scale affect ecosystem functioning.

K E Y W O R D S
algae, biodiversity, ecosystem functioning, insurance hypothesis, spatial insurance, transplant experiment four species are often assigned to their tolerance to abiotic factors such as desiccation, whereas biotic factors such as competition and predation often explain the lower limits (Connell, 1961;Schonbeck & Norton, 1978;Southward, 1958;Zaneveld, 1937). Specifically, we tested the prediction that species' monoculture functioning declines if transplanted to a different place on the environmental gradient, away from where they naturally dominate (Figure 2a-d, solid lines).
If this is the case, we would then expect that the loss of one species in one place of the environmental gradient (where it also dominates functioning) could not be compensated through the establishment of another species in that place. An alternative expectation would be that a species can compensate for the loss of other species in places where they do not naturally dominate functioning dashed lines). To test this prediction, we transplanted monocultures of adults from the four fucoid species into each of the four zones and monitored the relative growth rate as a proxy of functioning.
Despite substantial turnover along the intertidal gradient observed in a field survey, monoculture functioning of three of the four species showed limited responses to being transplanted to different depth zones.

| MATERIAL S AND ME THODS
The study was carried out on the Swedish west coast near the Tjärnö Marine Laboratory (Latitude: 58.875 DD; Longitude: 11.145 DD) from June to September 2021. The cyclical tidal range is approximately 0.3 m, but the water level can fluctuate up to 2 m due to winds and changes in atmospheric pressure (Johannesson, 1989).
Salinity varies between 15 and 30 psu (University of Gothenburg, https://www.weath er.mi.gu.se/). The shallow rocky shores of the region are characterised by dense stands of macroalgae that provide a habitat for diverse communities of associated organisms. On sheltered shores, the intertidal zone is dominated by fucoids (brown algae belonging to the order Fucales). As in many parts of the world (e.g. Baker, 2011;Lubchenco, 1980), intertidal fucoids on sheltered shores on the Swedish west coast have a relatively clear zonation pattern (Cervin et al., 2004). The zonation typically changes from high to low shore as follows: Fucus spiralis, Fucus vesiculosus, Ascophyllum nodosum and Fucus serratus ( Figure 1).

| Field surveys to confirm zonation
We used two sets of field surveys of local fucoid communities to (i) confirm the putative depth zonation of the four species, and (ii) determine the depths at which the four fucoid species occur, which we used to set up our field experiment (see next section). First, we measured the depth distribution of the four fucoid species using transects at site Y (see Figure 3a). To do this, we laid measuring tape parallel to the shoreline where macroalgae stands were present.
Then, we used a random number generator to select five transect positions (to the nearest cm). We placed a perpendicular transect into the water at these five randomly selected positions. Starting from 10 cm perpendicular to the uppermost algae, we recorded the identity of all fucoid algae within 15 cm left and right of the transect.
In addition, we measured the depth as the distance from the water surface to the substrate where each individual was attached.
We also recorded the concurrent water level relative to the RH2000 standard from the nearest water level monitoring station (Kungsvik, Latitude: 58.996 DD; Longitude: 11.127 DD) run by the Swedish Meteorological and Hydrological Institute (SMHI; https://www.smhi.se/), accessed through the Viva mobile application (https://www.sjofa rtsve rket.se/en/). Using the measured depth of each individual and the water level, we calculated the depth relative to the RH2000 as water level relative to the RH2000 F I G U R E 1 (a) Photo of the vertical turnover observed on the rocky shores of the study site. Icons and vertical lines correspond roughly to the species seen in the photo, but Ascophyllum nodosum is only visible on the right of the photo. (b) Field surveys revealed the depth distribution (relative to the RH2000 standard) of Fucus spiralis, Fucus vesiculosus, A. nodosum, and Fucus serratus. Points and error bars: mean ± SD. All data points are overlaid. The four thick horizontal ticks along the y-axis correspond to the depths at which the experimental tiles were placed to represent the four depth zones. standard + measured depth. This provided detailed data on the depths at which the transitions between the fucoid species occur.
The second survey was based on opportunistic algae collections.
Specifically, we collected eight individuals of A. nodosum, six individuals of F. serratus, 25 individuals of F. spiralis and 20 individuals of F. vesiculosus prior to the experiment. These individuals were collected at sites 1, 2, 3 and 4 ( Figure 3a; Table S2), where we recorded their depth. Depths were corrected to the RH2000 standard as previously described. We combined these opportunistic depth data with the depth data from our transects to generate empirical depth distributions for each species. These empirical depth distributions ( Figure 1b) confirmed the putative zonation based on previous literature (e.g. Cervin et al., 2004) and our own personal observations. We used the empirical depth distributions of each species to inform our field experiment.

| Transplant experiment
We tested our prediction that species' monoculture functioning declines if transplanted to a different place on the environmental gradient, away from where they naturally dominate, using a reciprocal transplant design. Specifically, we transplanted monocultures of each of our four focal fucoid species from the zone in which they are most common to each of the four zones. We used the relative change in dry weight as a proxy of primary production (i.e. functioning).
The experiment was spread over five intertidal sites ( Figure 3a; Table S2). At each site, we deployed 16 separate 30 × 30 cm granite tiles on gravel and sand beaches in a 4 × 4 matrix ( Figure 3b). The tiles were less than 2 m from a rocky substrate with macroalgae canopies containing the four focal species. Moreover, we observed gastropod herbivores and other macroalgae-associated organisms on our experimental tiles, which are commonly found in intact macroalgae canopies. This suggests that the conditions on our experimental tiles were similar to intact macroalgae canopies on the nearby rocky shores. The granite tiles were placed 1 m apart (centre to centre, see Figure 3b). In the 16-tile matrix, the four tiles were placed at four different depths (−5, −12, −28, and −40 cm) below the Swedish RH2000, which were chosen based on the observed depth distribution of these four species (Figure 1b). Since those depths were measured at the top corner of each tile, the algae were slightly below this depth due to the vertical angle of the shore. These depths are within one standard deviation of the mean depths at which the four species were naturally observed in our field surveys ( Figure 2b; Table S1).
Given that our field surveys showed that F. vesiculosus and A. nodosum were occurring at similar depths, we chose experimental depths to represent the dominant depth zones for F. vesiculosus and A. nodosum above and below the mean depth that they were observed in, respectively. However, the chosen depths were still within 6.5 cm depth of the mean and were not close to the tail of the depth distributions of the species (Figure 1b). To place the tiles at the correct F I G U R E 2 Expected growth rates of (a) Fucus spiralis, (b) Fucus vesiculosus, (c) Ascophyllum nodosum and (d) Fucus serratus in each of four experimental depth zones ( Figure 1; Section 2) relative to the depth zone in which each species naturally dominates (grey vertical rectangles). The solid line resembles the expectation based on the observed dominance of the four species at each depth (see Figure 1), and the dashed line represents our expectation if growth is independent of depth and species dominance. These lines represent rough theoretical expectations and not directly testable quantitative relationships.
Depth treatment (cm) depth, we measured the depth below or above the current water level using a ruler and corrected it to the RH2000 in situ based on accessing water level data from the Kungsvik station of SMHI using the Viva mobile application (as described previously).
The four depths vary in several environmental dimensions. The different depths are differentially exposed to periods above and below the water surface, leading to a gradient in the strength of the desiccation. To quantify this, we used almost 2 million observations of water level relative to the RH2000 standard data from the Kungsvik station (described previously) covering 6 years: 2015-01-01 to 2021-10-06 (one measurement every 1.7 min on average).
Given that we know the depth of our tiles relative to the RH2000 standard (i.e. −5, −12, −28, and −40 cm corresponding to the depth treatments), we can calculate the actual depth of the water at any time point in the water level dataset as depth = depth relative to the RH2000 − RH2000 water level. Using these calculated depths for each tile, we generated a series of summary variables in the data during the study period and over 6 years ( Figure S1a).
During the study period, the deepest treatment (i.e. −40 cm) was almost always submerged, while the two shallowest treatments (−5 and −12 cm) spent two or more hours outside the water more than 10 times per week, on average. The shallowest treatment (−5 cm) spent 19 h continuously outside of water during one period in the study ( Figure S1b). These differences in submersion also created substantial differences in temperature and light conditions ( Figure S2a,b, respectively). Here, temperature corresponds to a combination of water temperature because there were periods when the experimental tiles were not submerged due to water level fluctuations.
Although the average temperature does not vary substantially among depth treatments ( Figure S2a), the shallower depths are exposed to much higher temperature variability. For example, the shallowest treatment (−5 cm) experienced temperatures above 27°C, the upper tolerance for F. vesiculosus (Graiff et al., 2015) for an average of 94 h (±17.5 SD across sites) across sites during the study period.
In contrast, the deepest treatment (−40 cm) experienced these temperatures for an average of 4.6 h (±5.9 SD across sites).
At each depth level, we transplanted each of the four focal species onto one of the four tiles at that depth. Therefore, the four species were represented at each depth. We attached nine individual algae to nine 1 cm diameter drilled holes placed to reflect the natural densities of the species from field measurements.
To do this, algae were sampled from four different sites ( Figure 3a; Table S2) by cutting and kept in an outdoor flow-through tank until deployment (the time from sampling to deployment was a maximum of 4 days). The algae from these four sites were randomly assigned to each tile so that the four collection sites were represented on each tile. Before attachment, the algae were cut to an average length of 31 cm (min 12.5 cm, max 50.5 cm) and an average  (Table S3). After cutting, we patted the thallus dry using a paper towel before measuring the maximum length (to 1 mm accuracy), the wet weight (balance: VWR F I G U R E 3 (a) Map of the study area with transplant sites coloured red (letters) and sampling sites coloured purple (digits); see also Table S2. (b) Image of the tile matrix (site V) when the water level was low and a close-up of a tile with Fucus vesiculosus attached.

(a) (b)
PBP5201I-1S-FCECN 611-4828; to 0.1 g accuracy) and taking a flattened photograph on a lightboard (ARTOGRAPH LightPad 950) using a digital camera mounted on a tripod (Panasonic, Waterproof 4 k Compact Camera LUMIX DC-FT7). We used ImageJ 1.53r (Schneider et al., 2012) to measure the surface area (Cappelatti et al., 2019). After all measurements were taken, we tied a floating rope of 5-10 cm through each hole in each experimental tile and attached the nine individual algae to each piece of rope with cable ties (see Figure 3b).
The experimental tiles were installed at the five sites for a mean of 59 days (min = 54, max = 65) from June 29 to September 2, 2021.
We checked the tiles every 2-3 days and adjusted the depth, if necessary, as the stone tiles slowly sunk into the sand. As sandy beaches are prone to accumulating large amounts of algae debris, especially during stormy conditions, we continuously removed algae debris to prevent tiles from being smothered. At the end of the experiment, all tiles were collected over the course of a week. We placed the tiles in a flow-through outdoor tank for at least 24 h before any measurements were taken. Many algae had considerable epiphytic growth.
We removed all epiphytes before squeezing and patting them dry with a paper towel and measuring the epiphyte wet weight (balance: VWR PBP5201I-1S-FCECN 611-4828, with 0.1 g accuracy). As with the previous initial measurements, we dried all the experimental individuals with a paper towel before we measured the maximum length and wet weight and took a flattened photo to measure the surface area (as described previously

| Data analysis and statistics
Of the initial 720 individuals that we deployed, 52% were lost throughout the field experiment. In addition, two data points were discarded due to obvious data entry errors. This meant that we were left with 345 usable data points. The loss of individuals was mainly due to cable ties that severed the attachment point on the stem in combination with a storm at the end of the experimental period. Such high losses could affect our results if there were strong biases in loss rates based on site, depth or species (see Table S4 for an overview). However, despite these high losses, 77 out of 80 tiles had at least one individual that was intact, and 66 of those 77 tiles had more than one intact individual. There was some minor site bias as 9 of the 11 tiles with only one intact macroalga individual leftover were at sites W and X. Thus, at the tile level, the bias in lost data was relatively minor, and almost all tiles (96%) had some usable data. sites (χ 2 (4) = 46.78, p < 0.001) and depth (χ 2 (3) = 8.7, p = 0.033), the number of individuals lost per tile did not differ between species (χ 2 (3) = 1.70, p = 0.637) and we did not find an interaction between species and depth treatment (χ 2 (9) = 9.19, p = 0.420). Despite these potential biases, we focused on the 345 intact algae that were recovered for further analysis. However, we used a sensitivity analysis to make sure our results were robust to these biases in lost individuals (see below).
We chose to measure growth rate as the change in dry weight as it is the most direct and accurate measure of biomass (Bickel & Perrett, 2016). Given that we could only measure dry weight after the experiment, we used a linear model to predict the initial, preexperiment dry weight of each individual. To do this, we first fit a linear model to the post-experiment data that related post-experiment dry weight to the post-experiment surface area, and wet weight using the lm() function in R. A model selection based on AIC revealed dry weight ~ area + wet weight + species + area:species + wet weight:species + area:species + wet weight:area as the best model (Table S5). The model had an r 2 value of 0.98 ( Figure S3). We then used this model to predict the expected initial dry weight values based on the initial surface area and the initial wet weight data. Using these predicted initial dry weight values (t 0 ) and the measured, post-experiment dry weight values (t 1 ), we quantified the relative growth rate as follows: We tested whether the relative growth rate (i.e. dry weight change per day, g g −1 % day −1 ) depended on depth using a linear mixed effects model (LMM) for each species. We fit the models using the lmer() function from the lme4 package (Bates et al., 2015) in R (formula: relative growth rate ~ depth + (1|origin site) + (1|site/ tile)) with restricted maximum likelihood and used the lmerTest package (Kuznetsova et al., 2017) for subsequent significance testing.
For each model, we calculated the variance explained by the fixed effects depth, species, and/or epiphyte wet weight (i.e. marginal r 2 ) and including the random effects (i.e. conditional r 2 ) using the R package MuMIn (Barton, 2022). The significance values were derived from a Type III analysis of variance using Satterthwaite's method and implemented using the anova() function from the stats package. Subsequently, we calculated the estimated marginal means, confidence intervals and conducted a Tukey post-hoc analysis using the emmeans package in R (Lenth, 2022). In all analyses, depth was coded as a nominal variable because we did not want to assume an equivalent linear change in environmental variables with the depth zones. Moreover, given that we found a large amount of epiphytes relative dry weight change % day −1 = dry weight t1 (g) dry weight t0 (g) days −1 × 100 % .
on the algae, we tested whether the epiphyte wet weight (per individual and per thallus unit area) affected the relative growth rate using a similar linear mixed effect model that we fit with restricted maximum likelihood (formula: relative growth rate ~ epiphyte wet weight + species + depth + species:depth + (1|origin site) + (1|site/ tile)) and conducted the same analysis as mentioned above.
Given that we detected some bias in the number of lost algae individuals due to depth and site, we performed a sensitivity analysis to determine whether this bias may have affected our results. To do this, we equalised the sample size per experimental tile by drawing one random individual from each tile that had at least one individual and redid the analysis, testing whether relative growth rate (i.e. dry weight change per day, % day −1 ) depended on depth using a linear mixed effect model for each species. We repeated this procedure 200 times and compared the results to the results using the 345 intact individuals.

| RE SULTS
The average (predicted) initial dry weight and maximum length for the four species of seaweed across depth treatments were as follows: F. spiralis at 3.9 g (SD = 1.2) and 27.9 cm (SD = 4.1), F. vesiculosus at 4.0 g (SD = 0.9) and 32.2 cm (SD = 4.5), A. nodosum at 4.0 g (SD = 0.9) and 33.4 cm (SD = 5.5) and F. serratus at 5.1 g (SD = 1.1) and 32.6 cm (SD = 4.2). Over the course of the experiment (i.e. 59 days) and averaged across all four experimental depth zones, the mean dry weight change for F. spiralis was 0.7 g (SD = 2.2), representing a 14% increase. F. vesiculosus grew by 2.3 grams on average (SD = 2.1), representing a 55% increase in dry weight. The mean dry weight change for A. nodosum was 2.1 grams (SD = 1.4), representing a 52% increase, and the mean dry weight change for F. serratus was −1.4 grams (SD = 3.0), representing a 28% decrease. However, in the experimental depth zones corresponding to their natural distributions, the four seaweeds grew 0.5 g (SD = 1.7; 9%), 2.4 g (SD = 1.7; 59%), 2.9 g (SD = 1.1; 68%), and 1.6 g (SD = 1.9; 30%) on average, respectively (Table S6). Our observed growth rates are on par with what has been found in previous studies on the same fucoid species (Table S3).
The relative growth rate depended on the depth for both F.  (Figure 4b,c). In contrast, the average relative growth rate for F. spiralis did not differ from zero on average in any of the depth zones ( Figure 4a; Table S6). F. serratus showed the strongest depth response and could not maintain positive relative growth in either the −12 cm or −5 cm depth treatments (Figure 4d). The sensitivity analysis showed that these results were robust to potential biases in the number of individuals lost at different depths and across species ( Figure S4).

| DISCUSS ION
In this study, we tested the prediction that the monoculture functioning of four species of fucoid macroalgae declines when transplanted to a different location on an environmental gradient. This is a central assumption of several models of the insurance hypothesis (e.g. Loreau et al., 2003;Shanafelt et al., 2015;Yachi & Loreau, 1999) for the effect of biodiversity on ecosystem function (Loreau et al., 2021). Although our field survey showed that the four species occupy distinct zones in the intertidal (Figure 1), we did not find strong evidence that monoculture functioning declined when species were transplanted to different zones ( Figure 4). Rather, two of the four species (F. vesiculosus and A. nodosum) maintained positive relative growth rates in all zones and the relative growth rates of F. spiralis and F. vesiculosus were not significantly affected by depth zone (Table 1; Table S7). Only F. serratus responded strongly to the transplant, and its relative growth rate decreased substantially when transplanted to shallower depth zones ( Figure 4) (Hawkins & Hartnoll, 1985), and parts of the lower shore if competitors are removed (Lubchenco, 1980). We also found positive average growth rates for A. nodosum across the depth gradient, although its growth rate was significantly reduced in the shallowest zone (Figure 4). Similar effects were found by Stengel and Dring (1997) on growth rate and mortality when transplanting A. nodosum from the lower shore to the upper shore. In contrast, F I G U R E 4 Relative growth rate (dry weight change per day) of the four fucoid species (a) Fucus spiralis, (b) Fucus vesiculosus, (c) Ascophyllum nodosum and (d) Fucus serratus across the four different depth treatments. Positions of the grey bars and images correspond to the species' observed depth zones based on field surveys (see Figure 1). Diamonds and error bars indicate the mean ± CI 95% derived from models for each species (Table S7). Different letters indicate significant differences, as determined by Tukey post-hoc tests. All data points are overlain as jittered points. F. serratus only showed positive average growth rates in the two deepest treatments. Thus, unlike the other three species, F. serratus is limited by long periods of emersion at the upper shore due to its high sensitivity to desiccation (e.g. Zaneveld, 1937). The growth of epiphytes per unit thallus area did not differ between the four fucoids and therefore does not explain any of the differences in growth rates that we observed ( Figure S5).
Given the short time frame of our experiment, we could not detect the influence of rare or extreme events, such as ice scouring, which can affect the community structure of fucoid species in this system (Cervin et al., 2004). In addition, other life stages, such as zygotes and germlings, may be more strongly affected by environmental factors that vary with depth (Choi & Norton, 2005), which may drive species turnover. However, an ontogenetic perspective was beyond the scope of our study.
Our findings exemplify that more species are not necessarily re- There is some empirical evidence that species' monoculture functioning can remain relatively constant when transplanted to a different location on an environmental gradient. For example, (Germain et al., 2018) found that invasion growth rates of 30 annual plant species only changed with environmental variation in the presence of interspecific competition. A recent analysis of monoculture data from 26 biodiversity-ecosystem experiments showed that in at least half of them, a single species was highest functioning in monoculture in several different environmental contexts (Gamfeldt et al., 2023). The analysis by Gamfeldt et al. (2023) includes experiments spanning a range of ecosystem types (terrestrial, freshwater, marine) and organisms (e.g. algae, plants, fungi, primary consumers).
Therefore, results similar to ours may be relatively common in other systems as well. On the other hand, some reciprocal transplant experiments have revealed mixed results regarding how species perform when they are moved from where they usually occur and dominate. A meta-analysis of 14 transplant experiments in plants along elevational gradients found that plants had lower survival in 'foreign' habitats compared to their 'home' habitat, but reproductive output and biomass production were not affected (Halbritter et al., 2018). In a study using the adults of two kelp species in two different marine habitats (a bay and an island), while one species naturally dominated entirely in the bay, the nondominant kelp species performed better (in terms of both survival and growth), when both species were transplanted into the bay (Picard et al., 2022). The authors speculated that the naturally restricted occurrence of the subdominant species in the bay habitat might be due to constraints during earlier life stages. The early life stages of the subordinate species were discussed to be sensitive to factors such as high sedimentation and ice. Similarly, the zonation of the four fucoids in our study may be partly explained by, for example, high susceptibility to desiccation or heat for recruits of the lower-occurring species (F. serratus and A. nodosum). However, several species could be able to occupy and achieve high functioning in a place if competitors are absent, even though a natural pattern shaped by environmental conditions and not by, for example, drift would intuitively suggest otherwise (Germain et al., 2018). Consequently, in our view, natural patterns of species turnover should not be used directly to argue for the effect of biodiversity at landscape spatial scales.

| CON CLUS IONS
Most biodiversity-ecosystem functioning studies have taken place at small spatial scales. Therefore, it is unclear how changes in biodiversity at landscape scales could affect ecosystem functioning. It is important to be aware of this mismatch in scale because most ecosystem management interventions take place on the scale of landscapes and not on the local scale. We have shown that the functional consequences of species loss at landscape scales may not be as obvious as observational studies (e.g. Schiettekatte et al., 2022;Winfree et al., 2018) and ecological models, which commonly assume narrow niches (Loreau et al., 2003), have suggested. Instead, how species may or may not be able to compensate for the loss of others is a criti-

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information can be found online in the Supporting Information section at the end of this article.   Figure S3. Observed versus predicted dry weight with a linear model using the data collected at the end of the experiment (formula: dry weight ~ area + wet weight + species + area:species + wet weight: species + area:wet weight) Figure S4. Sensitivity analysis of the models for each species to evaluate biases due to lost individuals. Figure S5. (a-d) Epiphyte wet weight per thallus area (g cm −2 ) shown for each species and depth treatment. (e) The effect of epiphyte wet weight per thallus area (g cm −2 ) on dry weight change is shown as slopes for each species and depth treatment with a dashed line. Table S1. Summary statistics of empirical depth distributions for the different fucoid species used in the experiment and the experimental depth chosen for each species. Table S2. The names and coordinates (latitude, longitude) of the sites used for the transplant experiment (V-Z) and sites used to sample the algae used in the transplant experiment (1-4). Table S3. Comparison of the maximum length growth observed in our experiment in the natural depth zone of each species (i.e. zone in which they naturally dominate the abundance) compared with the maximum length growth reported in the literature. Table S4. Summary of losses of individuals per tile by species and depth treatment. Table S5. Model selection to predict seaweed dry weight (g) based on the post-experimental measurement of area (cm 2 ) and wet weight (g) of each species Table S6. Summary statistics of initial values and growth per day for dry weight, wet weight, area, and maximum length for all species. Table S7. Estimated marginal mean relative dry weight change (g g −1 % day −1 ) and 95% confidence interval for each species in each depth zone based on the linear mixed effects models for each species (Table 1)