Carbon assimilation and transfer through kelp forests in the NE Atlantic is diminished under a warmer ocean climate

Abstract Global climate change is affecting carbon cycling by driving changes in primary productivity and rates of carbon fixation, release and storage within Earth's vegetated systems. There is, however, limited understanding of how carbon flow between donor and recipient habitats will respond to climatic changes. Macroalgal‐dominated habitats, such as kelp forests, are gaining recognition as important carbon donors within coastal carbon cycles, yet rates of carbon assimilation and transfer through these habitats are poorly resolved. Here, we investigated the likely impacts of ocean warming on coastal carbon cycling by quantifying rates of carbon assimilation and transfer in Laminaria hyperborea kelp forests—one of the most extensive coastal vegetated habitat types in the NE Atlantic—along a latitudinal temperature gradient. Kelp forests within warm climatic regimes assimilated, on average, more than three times less carbon and donated less than half the amount of particulate carbon compared to those from cold regimes. These patterns were not related to variability in other environmental parameters. Across their wider geographical distribution, plants exhibited reduced sizes toward their warm‐water equatorward range edge, further suggesting that carbon flow is reduced under warmer climates. Overall, we estimated that Laminaria hyperborea forests stored ~11.49 Tg C in living biomass and released particulate carbon at a rate of ~5.71 Tg C year−1. This estimated flow of carbon was markedly higher than reported values for most other marine and terrestrial vegetated habitat types in Europe. Together, our observations suggest that continued warming will diminish the amount of carbon that is assimilated and transported through temperate kelp forests in NE Atlantic, with potential consequences for the coastal carbon cycle. Our findings underline the need to consider climate‐driven changes in the capacity of ecosystems to fix and donate carbon when assessing the impacts of climate change on carbon cycling.


| INTRODUCTION
Anthropogenic climate change is disrupting the global carbon cycle, which can further amplify warming through climate-carbon cycle feedbacks (Friedlingstein, 2015;Raddatz et al., 2007). Climate change can affect the carbon cycling through the biosphere by altering the stocks of carbon held within ecosystems, as well as influencing the efficiency by which it is transferred between different compartments. Although efforts have been made to incorporate climate-carbon cycle feedbacks into climate projections (Friedlingstein, 2015), they have primarily considered climate-driven changes in the carbon storage capacity of ecosystems in isolation. For instance, increased tree productivity might increase above-ground carbon storage in tropical forests (a negative climate carbon feedback; Lewis et al., 2009), while warming might accelerate the release of carbon stored in permafrost soils (a positive carbon feedback; Schuur et al., 2015). Meanwhile, the influence of climate on rates of transfer between compartments of the carbon cycle has been largely overlooked, something that may lead to erroneous predictions of the future carbon sequestration capacity of ecosystems (Sayer, Heard, Grant, Marthews, & Tanner, 2011).
Coastal marine environments exhibit high rates of carbon fixation, export and burial, and in doing so constitute a key component of the global carbon cycle (Bauer et al., 2013). Coastal vegetated habitats (e.g. mangrove forests, seagrass meadows, kelp forests etc.) are some of the most productive ecosystems on Earth and are gaining recognition as important contributors to the oceanic carbon budget (Duarte, Middelburg, & Caraco, 2005;McLeod et al., 2011;Nellemann et al., 2009). A significant fraction of carbon fixed by coastal vegetation flows through detrital pathways, as grazers typically consume a small fraction of total primary productivity (Mann, 1988). Accumulations of detrital material within these ecosystems can form deep organic-rich soils that represent globally important carbon repositories (Donato et al., 2011;Fourqurean et al., 2012). In addition, the transport of detrital material between habitats represents an important vector of carbon transfer in the coastal marine environment (Hyndes et al., 2014;Smale, Moore, Queiros, Higgs, & Burrows, 2018). Indeed, due to the highly dynamic and open nature of the marine environment, carbon may be buried within depositional habitats great distances from the source, thereby contributing to the total amount of carbon that is buried (Duarte & Krause-Jensen, 2017).
Macroalgae-dominated habitats, such as kelp forests, are among the most extensive and productive coastal vegetated habitat types globally (Duarte, Losada, Hendriks, Mazarrasa, & Marb a, 2013), but have been considered to play a secondary role in coastal carbon cycling and storage. This is because (a) macroalgal-derived matter is assumed to decompose too quickly to allow for long-range export and burial (Howard et al., 2017); (b) most macroalgae grow on rocks where in situ burial of organic carbon into sediments is precluded (Hill et al., 2015) and; (c) reliable estimates of the amount of carbon fixed and released by macroalgae, as well as their spatial extent, are lacking for most species and regions (Reed & Brzezinski, 2009). A growing body of evidence however, suggests that macroalgaederived carbon may be transported to habitats hundreds of kilometers away from source and to depths below thousands of meters (Hobday, 2000). This transfer of carbon constitutes a key trophic subsidy for habitats with low autochthonous productivity, such as offshore sedimentary habitats (Krumhansl & Scheibling, 2012). In addition, macroalgae carbon exports can contribute to carbon storage if they accumulate within habitats with long-term carbon burial capacity, such as seagrass meadows or offshore depositional sediments (Hill et al., 2015). Furthermore, recent investigations have shown that macroalgal tissues contain refractory carbon compounds (Trevathan-Tackett et al., 2015), which may represent important organic carbon reservoirs in the ocean (Wada et al., 2008). In light of these recent advances, macroalgal-dominated habitats are emerging as important donors of carbon within the coastal carbon cycle (Chung, Beardall, Mehta, Sahoo, & Stojkovic, 2011;Hill et al., 2015;Krause-Jensen & Duarte, 2016). Although climate and other anthropogenic stressors have been shown to alter the carbon stocks contained within marine habitats (Duarte et al., 2013;Fourqurean et al., 2012;Yando et al., 2016), how the flow of carbon between compartments of the coastal carbon cycle will respond to persistent climatic changes remains poorly resolved. Here, we used kelp forests dominated by Laminaria hyperborea-which constitute one of the most extensive coastal vegetated habitat types in the NE Atlantic Ocean-as model systems to examine the likely effects of continued ocean warming on carbon stores and fluxes in vegetated coastal ecosystems. Specifically, we quantified (a) the amount of carbon assimilated and stored in living biomass, and (b) the amount of carbon that is donated as particulate detritus through kelp forests persisting under two contrasting thermal regimes.

| Study sites
We quantified the amount of organic carbon held within, and donated by, Laminaria hyperborea kelp forests at multiple subtidal rocky reef sites situated along a gradient of~9°of latitude in the NE Atlantic. We sampled two sites within four locations (Figure 1a), which were comparable in terms of key environmental variables (e.g. wave fetch, salinity, nutrients, etc.), but differed with regards to thermal regime (Supporting Information Tables S1, S2 and S3; see also Smale et al., 2016). Sea temperatures within the "cold" locations (hereafter location C1 and C2) were, on average,~2°C lower compared with the "warm" (W1 and W2) locations (Figure 1b, Supporting Information Table S3); this regional variability in seawater temperature was most evident in summer, when the maximum temperature variability between the coldest and warmest location was 4°C (Figure 1b, Supporting Information Table S3). The surveyed forests extended from the low intertidal to~10 m depth and were located on wave-exposed rocky reefs that were similar in terms of PESSARRODONA ET AL. | 4387 geomorphology and topography. Forests were characterized by dense stands of L. hyperborea, which was the dominant kelp species (see  for details on kelp forest structure).

| Carbon assimilation and storage
To characterize the carbon held within kelp forests (i.e. their carbon standing stock), SCUBA divers carried out surveys at~3-5 m depth (below Chart Datum) in spring (April/May) and summer (August) yearly between 2014 and 2016 at each study site. During each sampling event, the density of L. hyperborea was quantified by haphazardly placing eight replicate 1 m 2 quadrats on hard bedrock and recording the density of mature canopy-forming plants (plants defined sensu Bolton, 2016 . To obtain the carbon biomass of each plant, fresh weight (FW) was first converted to dry weight (DW) and then to carbon biomass using an additional conversion factor. To calculate average site-specific FW:DW ratios, the fresh weight of the 15 complete plants was recorded in 2015, and individual sections of stipe (~10 cm length) and lamina (5 cm strips of both basal and distal material) were removed and dried at~60°C for at least 48 hr to determine FW:DW ratios for each section. The stipe and basal and distal parts of the lamina were dried separately as the relationship can vary between different parts of the plant (Smale et al., 2016).
The FW:DW ratios varied between sites and between parts of the plant (Supporting Information Table S4).
Dry weights were subsequently converted to carbon content using a conversion factor of 0.3125 AE 0.005 (mean AE standard error; Supporting Information Table S5). This factor was a yearly average obtained from routinely sampling two independent kelp populations within the W2 location (50°21 0 45″N, 4°08 0 32″W; 50°21 0 28″N, 4°07 0 42″W). Sampling was conducted approximately every 2 months to account for seasonal variability in carbon content.
During each sampling event, three individual mature L. hyperborea plants from each population were collected; kelp tissue from each collected plant was then obtained by sectioning a strip of each kelp lamina along its length (~4 cm width). The samples were freeze-dried and ground to a fine powder, before quantifying carbon content with a standard elemental analyser (CHN Analyser, EA1110, CE Instruments Ltd, Wigan). The same carbon conversion factor was used to convert all the dry weights from this study (see below).

| Carbon donation via particulate detritus
To examine rates of carbon donation to potential receiver habitats, we quantified the release of organic matter as particulate detritus. In L. hyperborea, detritus is produced through two main mechanisms: Where T is the number of tagged kelps (i.e. 15), L is the number of lost kelps at each plot, D is the mean L. hyperborea density at each site and w is the mean dry weight per plant at each site obtained from our yearly surveys. The results from each plot from the two sampling periods were combined to obtain an annual estimate and then converted to carbon using the conversion factors mentioned above.
In Laminaria hyperborea, loss of lamina tissue occurs through two discrete processes: "May cast" and "chronic erosion" (L€ uning, 1969;Kain, 1971; Supporting Information Figure S1). May cast is the popular name given to the major detrital pulse arising from the shedding of the previous-season's lamina growth, which remains attached to newly growing lamina until it is lost entirely as a "growth collar" Chronic erosion relates to gradual lamina loss throughout the year and, as it requires more frequent monitoring (i.e. at least monthly surveys), it is far more difficult to quantify across large spatial gradients. Logistical constraints prevented us from conducting monthly sampling at our eight sites simultaneously. To address this issue, we quantified lamina loss rates (through both May cast and chronic erosion) and the relative contribution of each mechanism at two independent, regularly sampled populations; this information was then used to model the annual chronic erosion rates along the latitudinal gradient, based on May cast measurements obtained at the principal study sites. The two independent study populations The growth of each digit was in turn calculated as where H f denotes the final position of the holes punched at the central (holes 1 and 2) and outer (hole 3) digits. To convert the loss of distal tissue (cm) to biomass (g), three 5 cm segments from the most distal part of each retrieved lamina were cut, and then weighed (FW). We determined the relationship between fresh and dry weight by drying the outermost segment at 60°C for 48 hr. All FW:DW relationships were highly significant and had an R 2 ≥ 0.89 (Supporting Information Table S6). We then estimated the dry weight of the rest of the 5 cm segments for each plant using the stated PESSARRODONA ET AL.

| 4389
relationship. Finally, the measured and estimated dry biomass per unit length was averaged between all three segments to give the dry biomass per unit length (g/cm) of the distal part of the lamina (B distal ). The daily erosion rate (E, g d À1 plant À1 ) of lamina tissue from each plant was calculated as: Where M is the mean lamina loss and t denotes the days between the initial and final measurements. The monthly erosion rates were then determined by multiplying the mean daily erosion rates by the total days within a given month. Finally, the relative contribution of the May cast to the total annual production was estimated by dividing the mean detrital production recorded in March and April by the total annual loss of lamina tissue.
Applying our observations of lamina loss from the independent year-long study, we estimated chronic erosion at each of the sampling sites along the gradient using Monte Carlo simulations. For each sampling site, 1,000 values of May cast production (g DW m À2 ) were generated by sampling randomly from a normal distribution with the obtained mean and standard deviation. Each May cast production value was then randomly assigned to a percentage contribution to the total detritus production (1,000 randomly generated percentages, p), which ranged from 56% to 70%, as per our observations from the two independent populations. These values agree with the observations of L€ uning (1969), who observed that May cast usually surpasses 50% of the total lamina loss. Chronic lamina erosion (g DW m À2 ) at each site was then calculated as follows: where p is the randomly generated percentage contribution and MC denotes the randomly generated May cast production obtained from randomly sampling the normal distribution. We then retrieved the mean, standard deviation and standard error from the 1,000 erosion estimates.
Annual estimates of carbon transfer via each mechanism of detritus production (i.e. dislodgement, May cast and chronic erosion; g C m À2 year À1 ) were summed for each site to obtain an estimate of total annual carbon flux. We used simple linear regression to examine the relationships between mean sea temperature and sitelevel values of carbon standing stock and carbon transfer via particulate detritus production.

| Carbon storage and donation across the geographical extent of Laminaria hyperborea forests
To assess whether the patterns observed across our study were representative of the wider geographical extent of Laminaria hyperborea forests, we compiled kelp biometrics data from study locations across a gradient of 28°of latitude. We used the largest average stipe length recorded for a given age class as a proxy for the maximum biomass (and carbon) accumulation attainable within a given location, as carbon assimilation and storage rates have not yet been measured across the geographical range of L. hyperborea. While stipe length is not a direct measure of biomass or carbon assimilation, it is a robust proxy for biomass accumulation for this species, given that: (a) there is a positive relationship between stipe length and plant biomass production (Kain, 1963), (b) stipes are perennial and longlived, reaching a maximum length at~6 years of age (Kain, 1963); (c) mature stipes exhibit minimal seasonal or annual variability in length or biomass (Sjøtun & Fredriksen, 1995); and (d) while stipe length is influenced by a range of factors such as wave exposure and competition for light at local scales (Smale et al., 2016), maximum attainable length is strongly influenced by environmental conditions at regional scales, of which light and temperature are critically important (Rinde & Sjøtun, 2005 Tables S9 and S10.

| Carbon assimilation and storage
The structure of kelp forests was spatially variable along the latitudinal gradient.  (Figure 2a). On average, populations in the warm climatic regime stored 68% less carbon than in the cold regime (Figure 3a).  Table S12). The three mechanisms of detritus production were combined to quantify the total annual flux of carbon via kelp detritus production. Although between-site variability was high, the greatest estimated values of carbon transfer were recorded at a site within the C1 location and the lowest a site within the W1 location, with a sixfold difference in total detritus production (Supporting Information Figure S3) Figure 2d), with the forests in the warmer locations releasing 54% less particulate carbon than those in colder waters, on average ( Figure 3a).

| Carbon storage and donation across the geographical extent of Laminaria hyperborea forests
Across our study region, site-averaged maximum stipe length was strongly related to total detritus production (Supporting Information Figure S4), providing support for the use of stipe length as a proxy for detritus production over larger spatial scales. Our synthesis of existing morphological data for L. hyperborea populations across its wider range indicated that maximum attainable stipe length is   Hill et al., 2015;Howard et al., 2017;Krause-Jensen & Duarte, 2016;Reed & Brzezinski, 2009). So far, reliable estimates of macroalgal carbon fluxes in the coastal ocean have been hampered by a scarcity of accurate data of the extent of macroalgal forests and the spatial variability of the amount of carbon they assimilate and release (Reed & Brzezinski, 2009 Figure 4a). However, the high productivity rates of macroalgaeand kelps in particular-underpin a considerable flux of particulate detrital carbon, which surpasses that of many other vegetated habitats ( Figure 4b). It is important to note that our study did not take into account the exudation of dissolved organic carbon, which may represent up to a quarter of the total carbon assimilated and released by L. hyperborea (Abdullah & Fredriksen, 2004). Given that to reach carbon sink habitats, then L. hyperborea forests would make a sizeable contribution to biogenic carbon sequestration, comparable to the 0.72 Tg C AE 0.12 year À1 accumulated within European saltmarshes for example (Ouyang & Lee, 2014).
Although the proportion of kelp detritus that reaches carbon sinks is currently unknown, detrital mats comprising L. hyperborea-derived material have been reported at depths in excess of 100s of meters (Freiwald, 1998 Figure S3). Although multiple environmental variables such as nutrient availability, irradiance and wave exposure influence the size, morphology and productivity of kelp plants (Pedersen, Nejrup, Fredriksen, Christie, & Norderhaug, 2012;Smale et al., 2016), they did not co-vary with temperature or differ between the climate regimes (Supporting Information   Table S2). Biological factors such as grazing pressure are also known to affect kelp standing biomass (Estes & Palmisano, 1974). However, the principal kelp-grazers in our study area (an omnivorous sea urchin and several species of gastropod mollusc) are not considered to exert strong top-down control on kelp populations, as they are small and do not form dense grazing aggregations (Hargrave, Foggo, Pessarrodona, & Smale, 2017;Smale et al., 2013). Reduced plant productivity (and therefore lower potential for carbon assimilation and potential storage and release) is frequently observed in temperature-stressed populations, such as those living at the warm edge of their distribution (Hatcher, Kirkman, & Wood, 1987), or those experiencing frequent heat and drought stress (Allen, Breshears, & McDowell, 2015). Our findings agree with previous historical (John, 1968;Whittock, 1969) and more recent (Smale et al., 2016) latitudinal surveys across the study region, which also found reduced kelp sizes and productivity in southern Great Britain compared to northern sites. We found a similar pattern across the geographical range of L. hyperborea when examining maximum stipe size, a reliable proxy for plant productivity and detritus production (Kain, 1971).
Although the quantity of morphological data from marginal populations at the trailing range edge is limited, Pereira, Engelen, Pearson, Valero, and Serrão (2017) (Belkin, 2009), with a further 1.5-5°C of warming predicted for this century (Philippart et al., 2011). L. hyperborea has already undergone a range contraction of~250 km at its warm, trailing range edge over the past 40 years (Assis, Lucas, B arbara, & Serrão, 2016), and further losses are expected as the water continues to warm and carbon assimilation as they allow for wind-driven resuspension of sediments (Bonsell & Dunton, 2018). In addition, benthic productivity in the Arctic will still nonetheless be greatly restricted by long periods of darkness (Dunton, 1985 F I G U R E 4 Per area carbon standing stock and carbon flux via detritus for dominant habitat-forming primary producers in Europe. The carbon stock contained within each habitat is partitioned into the amount stored in soils (dark blue bars) and in living plant tissues (light blue bars), which includes above and below-ground living biomass. The flow of carbon via detritus includes various kinds of litterfall and leaf shedding and detritus production. Values for Laminaria hyperborea are averages across the current study, details and references for other primary producers are provided in Supporting Information Tables S9 and S10. Values are means AE SE [Colour figure can be viewed at wileyonlinelibrary.com] positive effect on future kelp productivity and carbon assimilation (Brodie et al., 2014), which may compensate for temperature-meditated declines. In many kelp species however, photosynthesis is already carbon saturated under present conditions, and elevated CO 2 levels do not lead to increased primary production and growth rates (Iñiguez et al., 2016). In reality, elevated CO 2 levels may facilitate kelp competitors such as mat (turf)-forming algae (Connell, Kroeker, Fabricius, Kline, & Russell, 2013), which can displace kelp populations under stressful conditions, such as extreme ocean warming or reduced water quality .
The observed decline of kelp forests in the Atlantic reflects patterns in other temperate regions of the global ocean. Recent warming has been linked with loss of kelp forests and other large canopyforming macroalgae in several systems (e.g. Filbee-Dexter et al., 2016;Johnson et al., 2011;Verg es et al., 2014;Wernberg et al., 2016), leading to declines in detrital production (Krumhansl et al., 2014) and, intuitively, carbon fluxes. In a recent meta-analysis of kelp forest change over the past half-century, Krumhansl et al. (2016) reported significant declines in kelp abundances in 38% of the global ecoregions analyzed, while 28% of regions registered increases. Nevertheless, some of the regions where populations were found to be stable, or even increasing, have experienced massive declines and even extirpations of kelp populations in recent years (e.g. the South European Atlantic shelf; Fern andez, 2011; Raybaud et al., 2013;Assis et al., 2016Assis et al., , 2017. Continued declines in kelp forest extent, together with declines in productivity and shifts in species composition, could presumably diminish carbon assimilation and transfer through vegetated temperate marine ecosystems globally, with potential consequences for carbon cycling in the coastal ocean. Evidence to-date suggests that kelp losses may only be partly compensated by moderate range expansion into polar regions, with productivity still being severely restricted by sea ice dynamics (Bonsell & Dunton, 2018).
Until now, most efforts have been focused on understanding how climate change might affect the carbon sequestration and storage capacity within ecosystems. It is becoming evident, however, that climate-driven alterations in the fluxes between different compartments can also alter processes within the carbon cycle. For instance, climate change is predicted to increase riverine carbon exports (Larsen, Andersen, & Hessen, 2011), a significant component of the global carbon cycle which connects terrestrial and oceanic carbon reservoirs. Our work suggests that such alterations may also be important in the coastal ocean, where populations, habitats and trophic resources are highly interconnected because of the open and dynamic nature of the marine environment. Moreover, our results indicate that the magnitude of carbon transfer via detritus in coastal vegetated habitats is greater than previously reported, highlighting the need to incorporate this process into coastal biogeochemical models. Considering climate-driven changes in the carbon donor capacity of ecosystems will improve our understanding of how carbon cycle pathways will change in a future warmer ocean.