Effects of floating (solar PV) platforms on hydrodynamics and primary production in a coastal sea

primary production in a coastal sea Thodoris Karpouzoglou1, Brigitte Vlaswinkel2, and Johan van der Molen3 1Institute for Marine and Atmospheric Research (IMAU), Utrecht University, P.O. Box 80.005, 3508 TA Utrecht, The Netherlands 2Oceans of Energy, Wassenaarseweg 75, 2223LA Katwijk, The Netherlands 3NIOZ Royal Netherlands Institute for Sea Research, Dept. of Coastal Systems, and Utrecht University, P.O. Box 59, 1790 AB Den Burg, The Netherlands Correspondence: Johan van der Molen (johan.van.der.molen@nioz.nl)


Introduction
With a growing world population and growing global energy demand, new options need to be explored to generate energy.
While traditional fossil fuels emit carbon dioxide and other harmful gases which cause global temperature to rise, renewable forms of energy offer a sustainable alternative that can remediate climate change. Two of the most promising sources of renewable energy are the sun and the wind. Wind farms are built both onshore and offshore, but utility scale photovoltaic (PV) 5 solar farms are until now installed only on land. Growing space constraints, higher land costs, increased public resistance and competition with other functions will ultimately set a limit to the potential of onshore solar development, especially in densely populated areas. Such constraints may be less relevant at sea, and offshore solar energy generation has huge potential.
Large scale floating solar farms already exist inshore and are rapidly being developed all around the world (da Silva and 10 Branco, 2018). The effects of these structures on the ecosystem have been discussed mainly for standing water environments (Santafe et al., 2014;Sahu et al., 2016;da Silva and Branco, 2018). These studies argue that (inshore) floating platforms decrease the evaporation rate and increase water quality by reducing primary production due to the shadowing effect of the platforms. However, these studies did not investigate these effects in detail. The potential of offshore solar energy has recently been highlighted in several policy roadmaps in The Netherlands, and the world's first demonstration of an offshore solar farm Pohlmann, 2011;Pickering et al., 2012). In deeper areas further to the north, tidal currents are weaker and wave effects rarely reach the sea bed, allowing temperature stratification during summer (van Leeuwen et al., 2015). Such stratification limits vertical exchange of nutrients and determines the timing of the spring bloom (Sverdrup, 1953;Ruardij et al., 1997). We hypothesise that offshore floating platforms will modify currents, waves and stratification, and primary production. The platforms will cast shadows under water, reducing heat input and likely affecting temperature stratification. We also expect reductions 5 in under-water light intensity to affect phytoplankton growth. The friction of the rigid platforms with the tidal currents and shielding of the water surface from the wind are expected to result in weaker currents. The platforms can also be expected to have an impact on waves. Changes due to these forcings will affect turbulence and the resulting vertical mixing, suspended sediment and nutrient concentrations, and phytoplankton growth. 10 Here, we assess three contrasting locations in the North Sea for which time-series observations of hydrographic and biological quantities are available: a shallow and a deeper well-mixed site, and a summer-stratified site. We focus on changes in net primary production induced by the effects of floating platforms on the physical environment. In absence of field observations with floating platforms present, we used a water-column model to obtain first estimates of the potential effects of covering part of the sea-surface area on hydrodynamics and net primary production. This model allowed for easy development and testing of 15 the implementation of the effects of the floating structures on light (shading), wind forcing (shielding) and currents (platform friction).
The following research questions are addressed in this paper: 20 1) What is the overall potential effect of floating platforms on the net primary production at different locations in the North Sea as a function of coverage density? 2) What is the relative importance of the individual effects of platform shadow, wind shielding and platform friction?

Study sites and observations
Three study sites were selected for which time-series observations of hydrographical and biogeochemical variables were 30 available, with contrasting hydrographic conditions: Oyster Grounds (54.41 N,4.02 E),Noordwijk (52.301 N,4.303 E) and West Gabbard (51.9895 N,2.08983 E) (figure 1). Oyster Grounds is located at 45 m depth, and stratifies every summer between April and October (Tijssen and Wetsteyn, 1984). It is characterized by relatively low tidal current velocities, low suspended . sediment concentrations and low primary production. The sites West Gabbard and Noordwijk are located at 32 m and 18 m depth respectively. These locations remain well mixed during the entire year and they are characterized by relatively strong tidal currents, high suspended sediment concentration and high primary production.
At the three study sites, time-series observations were collected using SmartBuoys deployed by the Center for Environ-5 mental Fisheries and Aquaculture Science (Cefas) (www.cefas.co.uk/publications-data/smartbuoys). SmartBuoys are moored, automated, multi-parameter recording platforms which are used to collect marine environmental data. They measure, at 1 m below the sea surface, salinity, temperature, turbidity, oxygen saturation, chlorophyll fluorescence and nitrate and silicate concentration. Data were collected in 10-minute bursts; here we have used daily averages. The buoys also collected and preserved water samples which were used to calibrate the sensor data. For this study, we used observations from the following years:

Model description
For the purpose of this work the coupled physical-biogeochemical model GOTM-ERSEM-BFM was used. The General Ocean Turbulence Model (GOTM; Burchard et al. (2006); www.gotm.net) is a public domain, one-dimensional Finite Differ-15 ences water column model, that includes the most important hydrodynamic and thermodynamic processes related to vertical mixing in natural waters. The model solves the 1DV Reynolds-averaged Navier Stokes equations and the Reynolds-averaged transport equations of temperature and salinity, under the Boussinesq and hydrostatic approximations. In this offshore application of GOTM, salinity was considered constant. The model was forced with meteorological hindcast data obtained from the European Centre for Medium-Range Weather Forecast (ECMWF) ERA-40 (datasets/data/era40-daily/levtype=sfc/). Moreover, it was forced with time series of depth-averaged tidal velocities reconstructed from the harmonic analysis of a 3D model (van der Molen et al., 2017).
Coupled with GOTM, the European Regional Seas Ecosystem Model-Biogeochemical Flux Model (ERSEM-BFM) was As inclusion of phaeocystis without a riverine nutrient source led to spurious interannual variations, it was excluded from the calculations. 20 GOTM-ERSEM-BFM was modified to allow representation of the effects of the floating platforms on the hydrodynamics and ecosystem dynamics of the water column. The model accounted for the platforms through the introduction of three individual effects that can be activated separately or together: the shadow of the platforms, shielding of the water surface from the wind and the friction of the platforms acting on the currents. The implementation allowed for variable platform coverage as a fraction of the model surface. As the model represents averaged conditions over a unit surface area at each depth interval, it 25 can not distinguish between different ways of distributing this coverage over the unit surface area, and for the purpose of this study we assume the coverage to be distributed uniformly in space. The shadow of the platforms and the wind shielding effects were expressed by a linear reduction of surface irradiance and surface wind stress with coverage. The frictional effects of the platforms on the currents was represented, in similarity to the bottom friction, by an additional surface shear stress that was calculated with the logarithmic law of the wall, applied as a linear function of coverage. For mathematical expressions of the 30 implementation of the floating structures, see Appendix A. In absence of design details of operational systems, the roughness of the platforms is as yet not known, and may also vary during deployment due to biofouling. As a first approximation, the roughness height of the floating structures was assumed equal to that of the sea bed (h 0s = 0.05 m). A series of experiments with varying values of h 0s between 0.0125 and 0.4 m was carried out to provide insight into sensitivity of the model results to this parameter. Apart from coverage, this was the only parameter associated with the addition of floating platforms to the model. A sensitivity analysis of other parameters is beyond the scope of this paper, and the reader is referred to section 3.1 for a comparison with observations.

5
For each site, a water-column model was set up with 40 vertical levels with increased resolution near the surface and bottom.
Time steps were 300 s for the hydrodynamics, and 3600 s for the biology. Site-specific values for the porosity of the sea bed, the light-extinction factor for suspended sediment (the contribution to the light-extinction coefficient by suspended sediment is this factor multiplied by the suspended sediment concentration), and salinity were defined based on observations (table 1). As the water-column model is a closed system that conserves nutrient mass, it can only reproduce observations if the total amount 10 for each nutrient integrated across all ecosystem components reflects the average amount present in the vicinity of the site. In absence of direct observations of the amounts of nutrients in all ecosystem compartments, we tuned the initial concentrations of nitrate, silicate, phosphate and detritus in such a way that the model results, after a spin-up period of 26 years, matched the observed biogeochemical data as well as possible for each site. Because for two of the three sites only a few years of observations were available, and differences between years had to be accounted for in the tuning process, we did not have enough   1.5 10 5 6 10 5 1.8 10 5

Model experiments
The chlorophyll-a (figure 2, panels a,d,g), nitrate (panels b,e,h) and silicate (panels c,f,i).
For silicate and nitrate the agreement between model and observations was better for the locations of Oyster Ground and West Gabbard than for Noordwijk (see also Table 3). For chlorophyll-a, the model reproduced the seasonal cycle at the three sites, but underestimated the high concentrations during the spring bloom at West Gabbard and Noordwijk (figure 2). These  3.2 Sensitivity of net primary production to coverage

Comparison between locations
To compare the effect of floating platforms between the three locations (research question 1), the relative change in net primary production was plotted as a function of coverage ( Figure 3). The response was different at each of the three locations, but all sites showed, with increasing coverage, a limited reduction in net primary production followed by an accelerated reduction 5 leading to a complete collapse of net primary productivity. Taking all sites together, three ranges of coverage can be distinguished. From 0% to approximately 20% coverage the difference in response between the three locations was relatively small.
Also, the impact of the floating platforms on net primary production was relatively small (less than 10% reduction), while for West Gabbard even a small increase was simulated. Within this range of coverage, the two well-mixed locations appeared more resilient to the effects of the platforms than the stratified location of Oyster Grounds. From roughly 20% to approximately 10 40% coverage an increased spread in the results occurred between the three sites. Beyond approximately 40% of coverage, the net primary production at the two well-mixed locations sloped down rapidly, indicating collapse of the ecosystem. A similar collapse at the Oyster Grounds occurred later, at 60-80% coverage. These results suggest a different response for the stratified than for the two well-mixed locations. The two well-mixed locations appeared more resilient to small percentages of coverage, while they experienced an earlier ecosystem collapse.
The resilience of the well-mixed locations for small percentages of coverage with floating platforms can be explained by the migration of their spring bloom towards the sunnier summer months (Figure 4) and by the compensating effect of decreased surface suspended sediment on irradiance ( Figure 5 a,b). In contrast, the timing of the spring bloom at the stratified location of 5 Oyster Grounds, which is known to coincide with the onset of stratification (Ruardij et al., 1997), did not change substantially for coverage up to at least 60% ( Figure 4c). Considering irradiance near the surface (Figure 5a), for small percentages of coverage, a weaker reduction of subsurface irradiance occurred at the two well-mixed locations in response to a stronger reduction of suspended sediment at the surface (figure 5b), which allowed more light to penetrate the water column. The change in surface suspended sediment concentration with coverage followed the behavior of subsurface eddy diffusivity (figure 5c) in accordance 10 with theory, as lower values of eddy diffusivity result in less upward mixing of suspended sediment (Burchard et al., 1999).  displacement of the net primary production maximum that is located below the surface mixed layer (figure 6a). Due to its shift towards the surface and hence towards the light, the subsurface maximum of time-averaged net primary production (which happened mainly in summer) increased, while a reduction of time-averaged net primary production occurred within the surface mixed layer (which happened during the spring bloom). Above 60% of coverage, insufficient light reached the thermocline in summer, and the net primary production maximum observed at the stratified location of Oyster Grounds disappeared. The

Contributions to changes in net primary production by separate processes
To compare the importance of the individual effects of the floating platforms (platform shadow, wind shielding, platform friction) (research question 2) the response of net primary production to the different effects is presented in figure 7. Platform shadow was the dominant effect for all three locations. For the two well-mixed locations (figure 7 a and b) platform friction increased primary productivity, resulting in an overall effect that was smaller than the individual effect of platform shadow. In later spring bloom on net primary production that occurred at the well-mixed sites.

Roughness of the platforms
To assess the uncertainty introduced by the assumed value of the roughness height of the platforms (h 0s =0.05 m), and to evaluate the potential importance of the platform design and maintainance, model runs were conducted for different values of 5 h 0s . For coverage up to 20%, the difference was small for all sites (Figure 8). At the well-mixed sites (panels a,b), for higher levels of coverage (>40%), the range of values of platform roughness showed a spread in the impact of the floating platforms on the net primary production equivalent to a difference of approximately 10% in coverage, by modifying the eddy diffusivity, and thus the suspended sediment concentration near the surface. For the Oyster Grounds location (panel c) and coverage levels higher than 60%, the increase in roughness height compensated the impact on net primary production to some extent. This The water-column model assumes a 'unit' horizontal extent and spatial homogeneity, not only in terms of the oceanographic and biogeochemical properties but also in terms of coverage with floating platforms. As the spatial homogeneity assumption implies having the same conditions into infinity, it is not immediately clear how the water-column model results can be related 5 to solar PV farms of a finite extent. We can, however, provide a rough estimate of a minimum spatial scale needed to start to approximate spatial homogeneity. To obtain equivalent (changes in) primary production conditions as simulated by the watercolumn model, phytoplankton, which are transported by the tides, would need to spend a significant amount of time underneath a farm of a certain size (longer than they can chemically buffer solar energy photosynthesized before they were advected into or out of the farm area). Hence, we could take the tidal excursion length as a measure of minimum horizontal size corresponding  (Table 4). For solar PV farms smaller than this 15 length scale, the modelled reductions in net primary production presented here may be over-estimates, and simulations with spatially resolved models are needed to obtain more accurate results. A similar argument holds if substantial residual currents are present in addition to tides. We also note that the results presented here are based on the assumption that platforms are distributed homogeneously in space. Estimates of potential modulations of the current results that may be induced by inhomogeneous distributions of platforms in space can only be made with spatially resolved models.
20 These first model simulations have ignored a number of physical and biological processes that should be considered in further work. The implementation of PV-coverage with a 1DV model does not allow for a realistic representation of the spatial configuration of a solar power plant, the characteristics of which (e.g., the distance between platforms, service lanes) could result in a different response of the ecosystem. Moreover, wave-platform interactions and their effects on the mixing of the 25 water column and the resuspension of sediment have been ignored in this study. To account for these processes in further work, simulations with three-dimensional models are needed. Also, additional ecosystem components could be considered in a three-dimensional model, such as phaeocystis in areas with high nutrient loads, and growth of hard-substrate flora and fauna on the platforms. It may also be possible that there are effects on atmospheric properties and air-sea exchange. 5 We used three contrasting and relatively data-rich locations in the North Sea for this first study to illustrate the effects of floating platforms on net primary production. The differences in the response between the sites indicate that studying new locations will add valuable information. The study focused on the response of the marine (eco)system to floating platforms in terms of water-column structure and net primary production, but other quantities with indicator qualities should also be considered in further work, such as changes in sediment transport, disturbance of the balance of organisms, and the integrity of the sea bed 10 in terms of biomass, species composition and biogeochemical functioning. A good next step would be an examination of the effects of floating platforms with a local fine-resolution 3D model. The water-column model as presented here can, despite its limitations, be used as a test bed to support further work.
This first study was carried out as an exploratory investigation of potential effects and mechanisms, and has elucidated the 15 principle response of the ecosystem. Extreme care should however be taken to use the results for specific planning purposes, and in principle further investigations should be carried out for specific cases. However, as a rough rule of thumb, in absence of better data/models/knowledge, adopting the precautionary principle, and disregarding other effects and criteria that were not considered here (e.g., ecosystem variables other than net primary production, impact on waves, impact of biofouling on the biogeochemistry, specific spatial distribution of floating structures within a farm, acceptable levels of impact, political and 20 planning considerations, etc.), we recommend that real-world field implementations of floating infrastructure in the marine environment should not enter regimes 2 (too uncertain) and 3 (significant disturbance). This implies that, according to our results, coverage density should not exceed approximately 20% for farms of a size in the order of magnitude of the local tidal excursion length or larger. We also advise that for general and individual cases 'acceptable' levels of impact are defined and motivated, and further work is carried out to improve understanding of environmental effects of floating (solar PV) platforms, 25 or any other large floating infrastructure in the marine environment, in general and for specific cases.

Appendix A. Mathematical implementation of the floating structures
The incident radiation with floating strucures is given by with I 0 the incident radiation without platforms, and C the coverage fraction a number between 0 and 1. The surface wind stress with floating structures is given by with τ w the surface wind stress vector without platforms. The surface shear stress by floating structures, according to the logarithmic law of wall, is given by Here, u is the velocity vector in the surface cell, and r s the surface drag coefficient of the floating structures given by r s = κ ln( z0s+h/2 z0s ) 2 (4) with κ the Von Kármán constant, h the height of the surface cell, and z 0s the surface roughness length of the floating structures, defined by 10 z 0s = 0.1 ν u * s + 0.03h 0s .
Here, ν = 1.3E-6 is the molecular viscosity, h 0s the mean height of the roughness elements at the bottom of the platform, and u * s the magnitude of the friction velocity at the underside of the floating structures.
Code availability. The code versions used for the coupled model are available from the authors on request. Stand-alone code for GOTM can 15 be downloaded following instructions on gotm.net.
Author contributions. Thodoris Karpouzoglou did this work as MSc student at Utrecht University. Brigitte Vlaswinkel initiated the study.
Johan van der Molen formulated and planned the project.
Competing interests. Brigitte Vlaswinkel is Research Director of Oceans of Energy, a commercial company that develops offshore solar