There is increasing interest in macroalgae farming in European waters for a range of applications, including food, chemical extraction for biofuel production. This study uses a 3-D numerical model of hydrodynamics and biogeochemistry to investigate potential production and environmental effects of macroalgae farming in UK and Dutch coastal waters. The model included four experimental farms in different coastal settings in Strangford Lough (Northern Ireland), in Sound of Kerrera and Lynn of Lorne (north-west Scotland) and in the Rhine plume (the Netherlands), as well as a hypothetical large-scale farm off the UK north Norfolk coast. The model could not detect significant changes in biogeochemistry and plankton dynamics at any of the farm sites averaged over the farming season. The results showed a range of macroalgae growth behaviours in response to simulated environmental conditions. These were then compared with in situ observations where available, showing good correspondence for some farms and less good correspondence for others. At the most basic level, macroalgae production depended on prevailing nutrient concentrations and light conditions, with higher levels of both resulting in higher macroalgae production. It is shown that under non-elevated and interannually varying winter nutrient conditions, farming success was modulated by the timings of the onset of increasing nutrient concentrations in autumn and nutrient drawdown in spring. Macroalgae carbohydrate content also depended on nutrient concentrations, with higher nutrient concentrations leading to lower carbohydrate content at harvest. This will reduce the energy density of the crop and thus affect its suitability for conversion into biofuel. For the hypothetical large-scale macroalgae farm off the UK north Norfolk coast, the model suggested high, stable farm yields of macroalgae from year to year with substantial carbohydrate content and limited environmental effects.
Worldwide macroalgae (seaweed) production is in excess of 28 million tons per
year and has doubled between 2000 and 2014 (FAO, 2014). The majority of this
production (
There has been increasing interest in the potential of macroalgae cultivation across the Northern Hemisphere and Europe (Van der Burg et al., 2016), partially driven by research on biofuel technologies (Kerrison et al., 2015). The characteristics of Phaeophyta macroalgae, in particular high productivity, fast growth rate and high polysaccharide content, make them a suitable biomass for biofuel production (Hughes et al., 2012; Kerrison et al., 2015; Schiener et al., 2017; Fernand et al., 2017). A further advantage is that such third-generation biofuels do not need additional freshwater and do not compete for agricultural land like many existing biofuel sources.
Marine macroalgae fix CO
Kelp species, such as
Kelp naturally occurs in sublittoral coastal waters in temperate and polar regions. These macroalgae aggregations have been shown to modify the surrounding environment by reducing water velocity and attenuating waves (Jackson, 1997; Gaylord et al., 2007), and by modifying sedimentation rates of suspended particles (Eckman et al., 1989). They are also associated with high biodiversity (Burrows, 2012), providing numerous ecosystem services including habitat, shelter and food for many species including fish (Hartney, 1996), benthic organisms (lobster, crabs; Bologna and Steneck, 1993; Daly and Konar, 2008), herbivorous organisms (Kang et al., 2008) and birds (Fredriksen, 2003); see also Walls et al. (2017).
While a large-scale kelp farm might replicate some of the ecosystem services of a natural kelp forest, assumptions as to the extent of the similarity should be considered with caution (Wood et al., 2017). Since kelp farms are monocultures suspended within the water column and are likely to undergo a yearly cycle of growth and harvesting, they are not synonymous with mature kelp beds which contain multiple species of different ages attached to the benthos (Wood et al., 2017).
Studies on the potential environmental effects of macroalgae farms are limited. This lack of information, in combination with limited knowledge on expected farm yields, results in uncertainty for potential investors, developers and macroalgae farmers, as well as legislators, who provide the relevant farming licence (Wood et al., 2017).
The aim of this modelling study was to investigate environmental effects and potential yield of macroalgae farms, at different locations in UK and Dutch coastal waters, using the European Regional Seas Ecosystem Model – Biogeochemical Flux Model (ERSEM-BFM). In particular, five farms were simulated: four experimental farms (Sound of Kerrera and Lynn of Lorne, Scotland; Strangford Lough, Northern Ireland; the Rhine region of freshwater influence – ROFI, the Netherlands) and a hypothetical farm (Norfolk, UK) (Fig. 1). Observations from the experimental farms in Scotland and Northern Ireland were used to ground truth the model.
Variables of the macroalgae model.
Study area with SmartBuoy stations (black circles: 1 is Warp
Anchorage, 2 is Liverpool Bay, 3 is west Gabbard;
The growth of
In coastal waters around the UK, kelp species show high growth rates from late autumn to early summer. This is then followed by a slower growth phase between July and December (Parke, 1948; Kain, 1963). Maximum length developments are also associated with maximum fresh weights (Parke, 1948; Black, 1950).
Kelp plants show effective uptake of nutrients (ammonium, nitrate and
phosphate) from seawater (Birkett et al., 1998; Kregting et al., 2014;
Kregting et al., 2016). When nutrients are abundant and exceed metabolic
requirements, these plants have the ability to store nutrients in the plant
tissues (Birkett et al., 1998). For example,
The presence of these carbohydrate reserves and a fast growth rate make
The 3-D hydrodynamic General Estuarine Transport Model (GETM;
The ERSEM-BFM version used here (1 June 2016) is a development of the model
ERSEM III (see Baretta et al., 1995; Ruardij and Van Raaphorst, 1995; Ruardij
et al., 1997, 2005; Vichi et al., 2003, 2004, 2007; Van der Molen et al.,
2013, 2014, 2016;
Parameters of the macroalgae model.
Equations of the macroalgae model. The last column lists the numbers of the corresponding equations used by Broch and Slagstad (2012).
A macroalgae functional type representing
Schematic representation of the farmed macroalgae in ERSEM-BFM, modified and expanded after Broch and Slagstad (2012).
The Northern Irish farm site run by Queen's University, Belfast, is located
at 54.4
The Strangford Lough research farm is located near the south-western shore of
the lough (Fig. 1, Site A). The farm cultivated a mixture of
The mean biomass per line (kg wet weight m
The Sound of Kerrera and Lynn of Lorne farms (Fig. 1, Sites B and C) are located in the Firth of Lorne and operated by the Scottish Association for Marine Science (SAMS).
The first Scottish farm site was located at 56.38
The farm at Sound of Kerrera (Fig. 1, Site B) consists of 180 m of double-headed long line buoyed by mussel floats, with growing lines suspended at 1.5 m depth. For the Sound of Kerrera farm, observations of nutrient concentrations, light and temperature are available from a 17-month period in 2013–2014. Nutrient concentrations were collected in triplicates at 1.5 m depth, whereas light and temperature were collected at half-hourly intervals at 1.5 m depth, using HOBO Pendant data loggers (Onset Computer Corp, MA, USA). Here, we use the means of the triplicates for nutrients and monthly means for light and temperature. The nutrient data showed a typical seasonal cycle with high winter concentrations and low concentrations following the spring bloom, but with surprisingly high summer concentrations in 2013, which are unexplained. Early summer concentrations in 2014 were substantially lower.
Farm parameters as used in the model.
The second Scottish farm is located at 56.49
The Lynn of Lorne farm (Fig. 1, Site C) consists of a 100
The southern coast of the Netherlands is characterised by shallow water
depths (
Potential areas for a commercial farm off the north Norfolk coast.
Yellow to brown shading: suitability index. Black: moderately high to high
shipping intensity (derived from marine vessel automatic identification
system ping data obtained from exactEarth Ltd.,
Another experimental farm, run by the North Sea farm foundation (Stichting Noordzeeboerderij), was deployed for the first time in the autumn of 2016 within the nutrient-rich Rhine ROFI off the port of Scheveningen, the Netherlands (Fig. 1, Site D). The farm consists of a single line of 100 m, undulating between 0 and 4 m below the surface. Data from this farm will only become available in the summer of 2017. The farm was included in the model to obtain predictions of potential performance.
The north Norfolk coast of the UK is also characterised by shallow water depths, high winter nutrient concentrations (Hydes et al., 1999; Proctor et al., 2003; Foden et al., 2011) and high turbidity (Dyer and Moffatt, 1998; Bristow et al., 2013). Turbidity is higher than off the coast of the Netherlands, resulting in comparatively lower primary production by phytoplankton.
Limiting environmental variables for macroalgae cultivation. Ranges in bold were satisfied within the selected farm area (green rectangle in Fig. 3). Between brackets: values suggested by Capuzzo et al. (2014) if different.
The hypothetical commercial farm off north Norfolk (Fig. 1, Sites E–G) was selected based on the method of Capuzzo et al. (2014), with minor modifications. The method consisted of overlaying maps of suitability scores (optimal, sub-optimal, unsuitable) of key limiting environmental variables (temperature, light, tidal velocity, wave height and nutrient concentrations; Table 5) and spatial-use data (shipping, structures, Marine Protected Areas, wind farms, etc.) in a GIS system. The modifications applied here consist of slight variations to the threshold levels of certain environmental variables and the adoption of a farming area based on the suitability data rather than rectangles of predefined size.
The area selected by this method was nearly rectangular (53.0545
It was assumed that within each grid cell of 25 km
GETM-ERSEM-BFM was run without macroalgae farms from 1990 to 2011, using initial conditions from an earlier model version. The first 10 years of this simulation were considered as spinup time to enable the biogeochemistry of the model to adjust. The years 2001–2011 constituted the reference conditions (absence of farms). Farming scenarios were run for five consecutive seasons, starting on 1 October in 2006–2010 and running until the end of July of the following year in accordance with potential farming practice. The scenario runs were hot-started for each year from the corresponding conditions of the reference run on 1 October. To detect potential environmental effects, differences with the reference run were calculated from farm-season-averaged, depth-averaged model scenario output for all routinely stored variables (covering nutrient concentrations, functional type biomass and a selection of fluxes for both pelagic and benthic systems), filtered for model variability using the method of Van der Molen et al. (2016) and plotted as maps. The filtering method discarded differences between the reference run and the scenario run that were smaller than similarly calculated differences between the reference run and a duplicate of the reference run. Time series consisting of daily values were extracted for pelagic nutrients, light conditions and macroalgae conditions at each model grid cell containing a macroalgae farm to assess farm performance and functioning.
SmartBuoys, which are instrumented moorings (Mills et al., 2005), have been deployed in UK and Dutch waters as components of monitoring programmes and were configured to determine turbidity, chlorophyll fluorescence, salinity, temperature and dissolved oxygen and data processed according to Greenwood et al. (2010). Concentrations of suspended particulate matter and chlorophyll were derived from measurements of turbidity and chlorophyll fluorescence, respectively (Greenwood et al., 2010). Discrete samples were collected using an automated Aquamonitor and subsequently analysed for TOxN (total oxidisable nitrogen) and silicate according to Gowen et al. (2008). In addition, on most buoys, TOxN was determined using an automated in situ NAS-2E or NAS-3X nutrient analyser. Daily mean values were calculated from all data which passed the quality assurance process.
Daily spatial distributions of chlorophyll concentrations were derived from
the MODIS satellite (
Comparison of winter chlorophyll-
Modelled
Comparison of summer chlorophyll-
Surface chlorophyll concentrations from the reference run were compared with satellite observations for 2007–2008 (see also Van der Molen et al., 2016). Winter concentrations (October 2007 to March 2008; Fig. 4a, b) were low in both the model and the satellite data. For a better comparison, the model output was subsampled for each grid cell (Fig. 4c) using the available clear-sky satellite observations (Fig. 4d). Subsequently, the relative offset (Fig. 4e) and correlation coefficient (Fig. 4f) were calculated. The resulting plots show that the model over-predicted along the northern UK coast and in the coastal areas in the Celtic Sea. Correlations showed a patchy pattern, with typically better correlations along the Dutch and Belgian coasts.
In summer (April 2008 to September 2008; Fig. 5), the model had a small bias in offshore waters but tended to over-estimate coastal chlorophyll concentrations. It achieved good correlations in large parts of the North Sea and in parts of the Celtic Sea.
Time series comparison with Warp Anchorage SmartBuoy at the surface. Blue:
model; red: observations.
Time series comparison with Liverpool Bay SmartBuoy at the surface. Blue:
model; red: observations.
In the vicinity of the north Norfolk farm (Fig. 1, Sites E–G), the model bias for surface chlorophyll was slightly negative in winter and the correlation coefficient was low (Fig. 4e, f). In summer, chlorophyll concentrations were slightly over-estimated and correlations were moderate (Fig. 5e, f). Near the Rhine plume farm (Fig. 1, Site D), bias was slightly positive and correlations high in winter (Fig. 4e, f), and bias was slightly positive and correlations moderate in summer (Fig. 5e, f).
The model results from the reference run were compared with time series of
in situ observations from SmartBuoy for chlorophyll, nitrate, silicate,
salinity, temperature and suspended sediment. For Warp Anchorage (see Fig. 1
for location), modelled peak spring-bloom chlorophyll concentrations were
within 10 mg Chl m
Time series comparison with west Gabbard SmartBuoy at the surface. Blue:
model; red: observations.
In Liverpool Bay (see Fig. 1 for location), spring and summer chlorophyll
concentrations generally exceeded observed values from the SmartBuoy by a
factor of 2 (Fig. 7a). Nitrate concentrations were reproduced well in the
last 5 years of the simulation but were over-estimated in the first four
winters (Fig. 7b). Winter silicate concentrations were also higher in the
first few years but exceeded observed winter values for all the years in the
time series (Fig. 7c). Modelled salinities were slightly higher than observed
(Fig. 7d), and there was no apparent relationship with winter nutrient
concentrations as for Warp Anchorage. Summer temperatures were reproduced
mostly within a degree, while winter temperatures were under-estimated by up
to 2
At the more offshore location of west Gabbard (see Fig. 1), peak chlorophyll
concentrations were under-estimated for most, but not all, of the years
(Fig. 8a). Nitrate concentrations were under-estimated by a factor of 2–3
(Fig. 8b), whereas silicate concentrations were reproduced fairly closely
(Fig. 8c). Summer salinities were over-estimated by 0.8–1.2 (Fig. 8d).
Maximum summer temperatures were exceeded by up to 2
None of the maps of differences in biogeochemistry and plankton dynamics with the reference run, averaged over the farming season, showed detectable changes in the region of any of the farm sites; i.e. any differences between the run with farms and the first reference run were smaller than or of similar magnitude to differences between the two reference runs. For the experimental farms, this was to be expected because of their relatively small size. The north Norfolk farm (Fig. 1, Sites E–G) was located in a dynamic area with high tidal currents and substantial residual circulation, which may account for this result. Hence, in the following, we will focus on the performance of the macroalgae farms.
Model results for the Strangford Lough farm site.
Modelled winter nutrient concentrations at the Strangford Lough farm site
(Fig. 1, Site A; Fig. 9a, b), 1.5–5 mmol N m
Model results for the Sound of Kerrera farm site.
Model results for the Lynn of Lorne farm site.
For the Sound of Kerrera farm (Fig. 1, Site B; Fig. 10), winter nutrients
were higher but also with substantial differences between years
(7–15 mmol N m
Model results for the Rhine plume farm site.
The Lynn of Lorne farm (Fig. 1, Site C; Fig. 11) showed a very similar pattern but achieved almost twice the yields due to higher modelled nutrient concentrations. Interestingly, the carbohydrate content at harvest was lower for the high-yield years. In the year with highest yield, mortality shot up 10-fold shortly before harvest, suggesting that timing of harvesting may be critical. This latter result corresponds with the experience of the farm operators.
For the Rhine plume farm (Fig. 1, Site D; Fig. 12), the model over-predicted
winter nitrate concentrations as compared with observations from the
Noordwijk-10 station further to the north by up to a factor of 3 (up to
180 mmol N m
Simulated farm yields at harvest at the end of July
(10
The results for the Norfolk hypothetical farm (Fig. 1, Sites E–G; Fig. 13)
showed winter nitrate concentrations of 40–50 mmol N m
Model results for the western-most grid cell of the north Norfolk
farm site.
In addition to the per unit performance of the farms presented in the
previous section, it is, from the point of view of biomass production, useful
to list the total predicted yield of the farms at their current size. Total
modelled farm yields are summarised in Table 6 for both carbon and wet
biomass. In terms of wet biomass, yields were in the range of
2–3 t yr
Logarithm of nitrogen uptake as a function of irradiance and nitrate
concentrations in the model:
The model results suggested that macroalgae growth was dictated by combined
availability of nutrients and a sufficient level of light. To illustrate
this, nutrient uptake was plotted as a function of irradiance and nutrient
concentration. The resulting graphs for nitrogen (Fig. 14) show that this was
indeed the case: the Strangford Lough farm (Fig. 1, Site A; Fig. 14a)
experienced low uptake. For the Sound of Kerrera farm (Fig. 1, Site B;
Fig. 14b), there was only limited opportunity for higher uptake, as under most
conditions either light or nutrients were lacking. For the Rhine plume farm
(Fig. 1, Site D; Fig. 14c), very high uptake occurred, starting at high nitrate
concentrations in winter, and for light levels over
100
Logarithm of structural biomass as a function of irradiance and
nitrate concentrations in the model for the Sound of Kerrera farm:
Plotting modelled macroalgae biomass for the Sound of Kerrera farm in a
similar way and for the individual years (Fig. 15) elucidates the mechanism
behind the variability in farm yield in the model (Fig. 10f). The final
biomass appeared to be correlated not only with the winter nutrient
concentration but also with structural biomass in spring, when nutrient
concentrations were still elevated and light levels exceeded
50
The modelled production of macroalgae showed a range of responses that may illustrate the actual production that can be expected from commercially operated farms in these locations. Having said this, the model results were not highly accurate for all sites. At Strangford Lough (Fig. 1, Site A), the modelled winter nutrient concentrations were likely too low, leading to low macroalgae production in the model. This result provides an analogue for potential lack of farming success at sites with naturally low winter nutrient concentrations.
The model results at the Sound of Kerrera site (Fig. 1, Site B) were realistic, comparing in range with observed winter nutrient concentration levels (higher than at the Strangford Lough site) and also comparing in range with the observed variation in macroalgae production. Modelled production at the Lynn of Lorne site (Fig. 1, Site C) was higher than at Sound of Kerrera, coinciding with higher modelled winter nutrient concentrations. However, as a side effect of this, macroalgae carbohydrate content was lower.
At the Rhine plume site (Fig. 1, Site D), modelled nitrate levels were substantially higher than observed. Despite this, macroalgae production per metre was higher than at Sound of Kerrera and the last 3 years at Lynn of Lorne despite less favourable light conditions caused by higher concentrations of suspended solids and the line being deeper below the surface. This is most likely the result of more favourable nutrient concentrations. The modelled macroalgae contained low concentrations of carbohydrates, as they had continuous access to nutrients. The observed nitrate concentrations at a nearby location suggest limiting conditions in summer, and hence the real farm may yield macroalgae with a higher carbohydrate content.
The Norfolk farm (Fig. 1, Sites E–G), after compensating for the
over-estimated modelled suspended particulate matter concentrations by
reducing the depth of the lines below the surface, produced modelled
macroalgae biomass per metre of line higher than those simulated for the
Sound of Kerrera farm. This production showed good interannual stability and
contained up to 60 % carbohydrates. Simulated winter nutrient
concentrations were comparable with observed concentrations. There was a
slight variation in macroalgae production between the three model grid cells
occupied by the farm, in line with a slight gradient in suspended particulate
matter concentrations. Even for this farm, which was the largest that was
modelled, we did not find significant changes in temporal averages of other
model environmental variables over the period of simulated farming. This is
presumably because nutrient requirements of
The model results suggested that, in areas where winter nutrient levels are modest, farming success could be sensitive to the timing of the autumn onset and spring drawdown of nutrient levels. This result should be further tested and investigated using more detailed field and laboratory observations.
The results of farm yield (Table 6), in relation to farm size, can be used as a first indication of the magnitude of potential carbon removal from the marine environment if farming is scaled up in areas with good farming potential. If the produce is used for biofuel production, the farming activity would result in a related reduction in fossil fuel consumption, when allowing for conversion losses and fuel consumption as part of the production cycle. As nutrient-to-carbon ratios were very low, and the model did not detect significant changes in pelagic nutrient concentrations, macroalgae farming would need to happen on a substantially larger scale and/or intensity than simulated here to have an effect on eutrophication.
The current shelf-wide model allows for first assessments of macroalgae farm performance at a wide range of locations on the North-west European Shelf. The relatively coarse model resolution, however, clearly limits the accuracy of the farm production results, in particular in areas with large gradients in topography at scales finer than the model resolution. If more accurate simulations are desired for such locations, or if within-farm gradients in productivity need to be studied, local high-resolution models could be developed that take boundary conditions from the current model. The current model assumes horizontal lines, whereas some experimental farm configurations include vertical or diagonally undulating lines. The Rhine plume farm uses undulating lines, and hence the performance could be different from the simulations presented here. To better represent these different farm configurations and/or investigate potential differences in performance between such configurations, the model could be adapted by introducing the ability to distribute macroalgae biomass in the vertical. Similar adaptations would also be required for the simulation of natural populations, in which plants are anchored to hard substrate and grow vertically towards the surface. The approximation for frond erosion suggested by Broch and Slagstad (2012) does not explicitly include effects of the environment and might be refined with a suitable set of laboratory experiments and field observations. However, the currently simulated values are small, and experience from the experimental farms indicates that higher values only occur later in the season. Crops can be harvested before such mortality occurs, so accurate predictions of mortality are not of high relevance for the current application.
This modelling exercise is a proof of concept and did not aim for a detailed representation of the farm localities, nor did it involve extensive tuning to reproduce detail of farm performance. We used an existing 3-D model setup of the North-west European Shelf (Sect. 2.1), which allowed all farms to be included in one model, albeit with a very coarse representation of coastal geometries. Farm implementation included a level of sub-grid parameterisation. The model was run with forcings for years predating farm deployments, so comparisons with observations collected during the actual deployments can only be qualitative. Despite these limitations, we obtained reasonable confidence in the model, as well as valuable results in terms of farm functioning and performance, macroalgae quality and farming-induced changes in environmental conditions. These predictions are necessary to progress the future development of this fledgling industry.
This model study did not detect large-scale changes in environmental conditions in the vicinity of the simulated farms. Although this is encouraging, we do not consider this finding to be a generic result, and further and specific investigations should be carried out for specific proposed farm implementations. Such work could include application of and contrasting with other models, and further upscaling of farm size and intensity to explore safe limits. Moreover, the current model (as any model) only captured a subset of environmental processes. Also, simulations with a high-resolution model are recommended to confirm and refine the results obtained here, if further farm implementation plans are developed.
The results for the hypothetical Norfolk farm site (Fig. 1, Sites E–G) suggest favourable conditions for commercial macroalgae farming. However, suspended particulate matter concentrations may remain an issue, and accurate regulation of a very shallow depth of line below the surface is probably required. A small-scale field experiment is recommended to test this result in reality.
For the Sound of Kerrera site (Fig. 1, Site B), with lower nutrient concentrations and variable farm yield, the model suggested a relationship between farm yield and autumn and spring nutrient concentrations coinciding with light at sufficient levels. This suggested relationship should be investigated further and confirmed with more detailed observations than available for this study, as further understanding of these processes can help to determine minimum required conditions for successful farming.
The model results suggest high rates of macroalgae growth in early summer, accompanied by an increase in carbohydrate content but also by an increase in mortality. This suggests that there is an optimum window for harvesting, in line with experience from the experimental farms; however, the simulated mortality was not enough to start to reduce biomass. The model suggested differences in this balance between the farm sites, but without further field evidence it is difficult to draw detailed conclusions. It is recommended to continue the field experiments and to gather more detailed information on environmental conditions, carbohydrate content and mortality. This could be accompanied by suitable series of shore-based microcosm experiments. Associated modelling work can help to explain and extrapolate such results.
Concerning the model, improvements could be made in the simulation of nutrients and particulate suspended matter. Also, representations of different farm configurations could be considered (e.g. undulating or vertical lines). Other macroalgae species could be included, as well as a capability to model natural, sea-bed attached macroalgae populations. Finally, inclusion of macroalgae grazers in the model could be investigated, as grazing can be a problem for farm operation.
Satellite observations were obtained from the Ifremer ftp
server (
The authors declare that they have no conflict of interest.
This work was carried out under the SeaGas project, led by the Centre for
Process Innovation (CPI). The SeaGas project is funded by both Innovate UK
(for industrial partners) and BBSRC (for academic partners). The funding
scheme is the Industrial Biotechnology Catalyst, grant number 102298. Queens
University was supported by BBSRC under grant number BB/M028690/1. The work
was carried out under Cefas contract code C6627. The macroalgae modules were
developed at NIOZ. Recent elements of the model development were funded by
Cefas Seedcorn projects DP261 and DP315. Development of postprocessing was
supported by Defra under Cefas contract FC002. Claire Coughlan, while at JRC
(Ispra), created the open-boundary forcing for temperature, salinity and
nutrients for the model. ECMWF and BADC are acknowledged for making the
atmospheric forcing available. Data from Strangford Lough are from the EnAlgae
project, which was funded by the European Regional Development Fund via the
INTERREG IVB NWE programme. Data from the Sound of Kerrera farm were funded
through the European Commission Seventh Framework Programme (FP7) project –
Advanced Textiles for Open Sea Biomass Cultivation (AT