A review of tidal energy—Resource, feedbacks, and environmental interactions A

The ocean contains a variety of renewable energy resources, little of which has been exploited. Here, we review both tidal range and tidal stream energy, with a focus on the resource, feedbacks, and environmental interactions. The review covers a wide range of timescales of relevance to tidal energy, from fortnightly (spring-neap) and semi-diurnal variability, down to array, and device-scale turbulence. When simulating the regional tidal energy resource, and to assess environmental impacts, it is necessary to account for feedbacks between the tidal array and the resource itself. We critically review various methods for simulating energy extraction, from insights gained through theoretical studies of “tidal fences” in idealized channels, to realistic three-dimensional model studies with complex geometry and arrays of turbines represented by momentum sinks and additional turbulence due to the presence of rotors and support structures. We discuss how variability can be reduced by developing multiple (aggregated) sites with a consideration of the enhanced phase diversity offered by exploiting less energetic tidal currents. This leads to future research questions that have not yet been explored in depth at ﬁrst-generation tidal sites in relatively shel-tered channels (e.g., the interaction of waves with currents). Such enhanced understanding of real sea conditions, including the effects of wind and waves, leads to our other identiﬁed primary future research direction—reduced uncertainties in turbulence predictions, including the development of realistic models that simulate the interaction between ambient turbulence and the turbulence resulting from multiple wakes, and changes to system-wide hydrodynamics, water quality, and sedimentation.


I. INTRODUCTION
Investment in emerging renewable energy technologies is essential if the global energy sector is to transition from fossil-based toward zero-carbon by the second half of this century, limiting the impacts of climate change. 1 Many of these emerging technologies are based on a resource that surrounds us-the ocean. Although there are many forms of ocean energy conversion, including wave energy, 2 Ocean Thermal Energy Conversion (OTEC), 3 and ocean currents, 4 one ocean energy resource has the dual advantage of high predictability and excellent opportunities for grid connectivity-tidal energy-the focus of this review.
Tides are predictable because of their origin in (astronomical) tide generating forces. These manifest as a number of tidal constituents, the dominant of which are the semi-diurnal lunar (M2) and solar (S2) constituents, with periods of 12.42 and 12 h, respectively ( Table I). The combination of M2 and S2 leads to the fortnightly spring-neap cycle, with enhanced tidal range (spring tide) when the Earth-Moon-Sun system is in-line (either New Moon or Full Moon) and reduced tidal range (neap tide) when the Earth-Moon-Sun system is perpendicular (either First Quarter or Third Quarter Moon). It can therefore be seen that there is variability of the tides (and hence the potential of the tides for electricity generation) at both semidiurnal and fortnightly time scales (Fig. 1). In addition to M2 and S2, other important tidal constituents that significantly affect the temporal variability of the tides are the lunar (O1) and lunisolar (K1) diurnal constituents, and the larger lunar elliptic semi-diurnal constituent (N2) ( Table I). At longer timescales, the tidal resource is also influenced by the effect of the 18.6-year lunar cycle, which is mainly driven by the changes of the inclination of the Moon's orbital path relative to the plane of the Earth's equator. 5 Global tidal dissipation is around 2.4 TW, with the majority of this, 1.7 TW, occurring in shelf sea environments. 6 This represents an upper theoretical bound for tidal power, but due to interaction between tidal energy extraction and the resource (e.g., Ref. 7), in addition to technical and practical constraints, the available resource is likely to be considerably less. To put this in perspective, annual mean global electricity consumption is around 3 TW, 8 and so even if 10% of the shelf sea resource was exploited, that is, 170 GW, tidal energy could have a substantial contribution to the global energy mix. Further, in some regions such as the UK, tidal dissipation is around 200 GW, whereas mean demand for electricity is currently around 40 GW; therefore in some areas, the potential contribution of the tides to the electricity mix significantly outweighs the global mean.
There are two main ways of converting the energy of the tides into electricity-tidal range power plants and tidal stream turbines. Tidal range power plants (Sec. II) rely on the potential energy of the tides. Water is impounded behind an artificial embankment and released through turbines when there is sufficient head difference between the artificial water level inside the barrage and the natural tidal level outside of the barrage. In contrast, tidal stream turbines (Sec. III) rely on the kinetic energy of the tides. In regions where the tidal streams are of sufficient magnitude, the flow is intercepted by an in-stream tidal turbine, which spins a generator and produces electricity. Although tidal stream turbines can be installed individually, it is only when they are deployed in arrays (i.e., like a wind array) that they can reach their full potential.
In this review, we consider these two forms of tidal energy conversion-tidal range energy and tidal stream energy-discussing the resource, projects, and research questions. We also discuss turbulence and wakes, and the environmental consequences of extracting tidal energy, particularly at significant scale, from the oceans. Finally, we identify future research areas that need to be addressed before the tidal energy resource can be exploited to its full potential.

II. TIDAL RANGE
Converting tidal range into other useful forms of energy has a history that extends back to the tide mills of the 6th Century. 9 However, converting the potential energy of the tides into electricity began with the construction of La Rance power station (France) in 1966. 10 La Rance comprises a 720-m-long barrage and impounds an area of approximately 22 km 2 in a region that has a mean tidal range of around 8 m. The barrage houses 24 Kaplan bulb turbines, which provide a combined rated power output of 240 MW. Four other tidal range power plants have been constructed (Table II), but it is notable that no scheme has been constructed in the last 25 years, setting aside the complication that the embankment for the 254 MW Lake Sihwa barrage was constructed in 1994, and the power plant in 2011. Further, all of the existing tidal range power plants are of the barrage type; that is, they span the full width of an estuary or channel. It is generally considered that the next generation of tidal range power plants will be lagoons, which only partially span an estuary, since these are associated with lower capital cost and reduced environmental impacts. 11 A tidal range power plant is based on a number of components, the most important of which is the embankment. The embankment, which constitutes the majority of the capital cost of the power plant, is a barrier that is used to impound an area of sea, the filling or release of which is controlled by diverting the flow either through sluice gates (e.g., to transfer water into or out of the enclosed basin) or through turbines (i.e., to generate electricity). There are two main modes of operation of a tidal range power plant, and a third operation mode that is a combination of the other two. For ebb generation, the flooding tide enters the enclosed basin through sluice gates and idling turbines. Once the maximum level inside the lagoon is achieved, these gates are closed, until a sufficient head develops on the ebbing tide. Power is subsequently generated by the turbines and generators until a predetermined minimum head difference, when turbines are no longer operating efficiently. For flood generation, the process is reversed to produce electricity during the flood phase of the tidal cycle. Two-way generation is a combination of the two and can be used to reduce variability in the resulting power time series. Two-way generation can also be supplemented by pumping to further reduce variability. 12 Tidal waves propagate at a phase speed c ¼ ffiffiffiffi ffi gh p , where g is gravitational acceleration and h is water depth. Therefore, tidal waves travel much slower in shelf sea environments than they would in the deep ocean. For example, the mean depth of the world's oceans is 4000 m, and a tidal wave would travel at 200 m/s. In a shelf sea with water depth of 100 m, the tidal wave would propagate at 31 m/s. When a wave propagates along a channel that is relatively long compared to the tidal wavelength, at a distance of 1/4 of a wavelength from the head of the bay or estuary the crest of the incoming wave passes at the same time as the trough of the reflected wave (and vice versa). The two waves (incident and reflected) cancel at this point, known as a node, and so there is no tide at this location. Since the incoming and reflected waves combine, the tidal amplitude at the head of the bay is doubled in such a standing wave system.
Tidal resonance occurs when the length of a channel coincides with the quarter wavelength. In such a case, the tidal range at the entrance to the seaway is amplified at the head of the estuary. The most famous case of tidal resonance is the Bay of Fundy on the east coast of Canada, which has the highest tidal range in the world (16-m spring tidal range). With a mean depth of 66 m, the tidal wave propagates at a mean speed of 25 m/s along the channel. For quarter wavelength resonance with the semi-diurnal tide (T ¼ 12.42 h- Table I), this represents a length of 280 km, which is very close to the actual length of the Bay of Fundy (290 km), that is, the channel is in near resonance with the semi-diurnal tide. Since tidal power is related to the basin area and the square of tidal range, these regions of high tidal range are sought for the placement of large tidal power plants; for example, the mean tidal range at La Rance is 8.2 m, and 5.6 m at Lake Sihwa.
Since the tidal range resource is a function of tidal range squared, the resource is discretely distributed across shelf sea regions (Fig. 2), with Australia alone containing 30% of the global resource, and Canada (Fundy) containing 23% due to its near resonance semidiurnal tides. The UK and France each have an equal share of around 13% of the global tidal range resource, 11 with much of this focused in the Severn Estuary (UK) and the Gulf of St. Malo (France). Since we know the speed of propagation of the tidal wave (a function of water depth), we know the timing of the tides at any single location, but also the relationship between the phase of the tides between multiple locations. It is therefore possible to stagger a series of tidal range power plants along a coastline to reduce variability in the aggregated power output. 13 Tidal range power has more potential for this "tidal phasing" compared to tidal stream power (Sec. III), where the majority of energetic sites are in phase with one another. 14 As stated earlier, all of the existing tidal range power plants are barrages (Table II), but a newer concept is that of a tidal lagoon, where only part of an estuary is blocked off to create an artificial basin. A much studied location for a tidal lagoon is Swansea Bay in the Bristol Channel, UK, 15 where the spring tidal range is 10.5 m. This site has also been the subject of a UK Government commissioned review, known as the Hendry Review, which investigated the role of tidal lagoons in the future UK energy mix. 16 The proposed Swansea Bay tidal lagoon would comprise a 9.5 km embankment, 17 enclosing an area of 11.5 km 2 . Using 16 Â 7:2 m-diameter turbines, installed in 60 m draft tubes, the power plant would have a nameplate capacity of 320 MW. Using dual (flood/ebb) mode, the project is estimated to have a capacity factor of 19%. The turbine design is triple regulated, 18 that is, adjustable guide vanes, blade pitch angle, and variable speedthe latter potentially reducing harm to fish. The Hendry Review supported the development of the Swansea Bay tidal lagoon, known as a "pathfinder" project; however, the review further recommended that the pathfinder project should be operational for a reasonable period of time before construction of any further (and potentially larger) tidal lagoon projects. This is to allow the full range of environmental impacts to be monitored over time. One important point about tidal range schemes, in comparison with other energy projects, is the lifetime of the embankment, which is generally estimated to exceed 100 years, and this could be accounted for when estimating the levelized cost of energy (LCoE).
Little research has been conducted into how climate change, particularly changes in mean sea level, will affect the tidal range resource. This is particularly relevant, since the embankment of a tidal range power plant is estimated to have a lifetime that is highly likely to witness significant changes in the climate. For example, an increase in sea level in a tidal basin will lead to an increase in the speed of propagation of the tidal wave, altering the resonance characteristics. A model study of the Bay of Fundy shows that a 1 m rise in mean sea level would lead to an increase in approximately 0.1 m in tidal amplitude due to the basin approaching true resonance. 19 This may seem like a small change, but since the tidal range resource is related to the tidal range squared, it could lead to a significant change in the resource. Global tidal modeling under a more extreme scenario of 2 m sea-level rise leads to a complex picture of regional changes in tidal amplitude, with no consistent pattern emerging of increase or decrease in tidal amplitude over shelf sea regions. 20 Further, such model studies, which The Swansea Bay tidal lagoon is in italics as it is a proposed scheme.
Journal of Renewable and Sustainable Energy REVIEW scitation.org/journal/rse examine future changes in tidal dynamics due to sea-level rise, should do so concurrently with changes in developed energy infrastructure. 21 On non-tidal influences to the tidal range resource, it has been demonstrated that storm surges will not significantly impact the resource compared to that estimated by astronomical constituents alone. 22

III. TIDAL STREAM
Tidal streams are regions with concentrated tidal flows, posing an energy resource with significant potential. The kinetic energy may be extracted from these tidal streams using hydrokinetic turbines, analogous to wind energy. Similar to offshore wind, turbines are typically deployed in arrays, increasing system redundancy, and resilience. The vast majority of tidal turbines use rotors to convert the kinetic energy of the flow into mechanical energy, either with the horizontal axis parallel to the flow or with the axis perpendicular to the flow. However, there are examples of turbines using other means such as oscillating hydrofoils and underwater kites. 23 Regardless of the specific type of technology being used, all projects require resource assessments to first determine project feasibility, then to design the turbine array layout, and finally to compute the project annual energy production (AEP). The International Electrotechnical Commission (IEC) has published a technical specification for performing tidal stream energy resource assessments for projects at both feasibility and design stages. 24 Starting with small-scale resource assessments, the theoretical power available for an individual turbine is computed as a cube of the free stream velocity where q is the density of the fluid and V the flow speed. Accounting for the efficiency of a particular turbine, this equation can be modified by multiplying by the power coefficient C p , where P d is the electrical power density. Sometimes C p is referred to as the performance coefficient and represents the shaft power rather than the electrical power. An upper bound (16/27, or 59.3%) for turbines in an unconstrained flow was derived using actuator disk theory and is formally referred to as the Lanchester-Betz-Joukowsky limit, but is most frequently called the Betz limit. [25][26][27] This analysis was performed for wind turbines; however, it is generally accepted that it also applies to tidal energy, providing the turbine swept area is small relative to the channel depth and width. 28 The actual power output depends on the properties of the turbine. In particular, the power for the turbine (P) is found by multiplying the power density by the swept area of the turbine, A s , The power coefficient can be a function of the velocity with variable efficiencies, including a minimum cut in speed below which the turbine does not operate and a rated speed at which the turbine does not generate more power for higher flow velocities. The IEC has a technical specification outlining the methodology for empirically determining the power coefficient as a function of the incoming current speed. 29 Finally, the total energy converted by the turbine is found by integrating the power over the desired period of time, most commonly a year, to get the AEP. Optimal efficiencies of tidal power converters are generally lower than the Betz limit and are reported to be typically between 16-50%. 30,31 However, for many tidal streams, the Betz limit is irrelevant and may be exceeded under conditions with constrained flow 32 or when using turbines with ducted intakes. 33 On a much broader scale, high-level reconnaissance studies are implemented at country or regional scales to identify regions with significant tidal energy resources by evaluating existing data or using numerical simulations of large regions. One example of a reconnaissance study is the geodatabase of tidal constituents developed using the Regional Ocean Modeling System (ROMS), which presents the regional assessment of tidal stream power resource and identifies locations with high kinetic power density in the USA. 34 Feasibility studies are used to compute a preliminary estimate of the upper bound for the potential available power for a project.
These are generally done using simulations with depth-averaged 2D models that reasonably reproduce the general characteristics of the tidal flows. However, estimating the power for larger projects with the current velocities from model simulations without accounting for the effect of energy extraction can be erroneous. Vennell et al. 35 suggested that the any project with a total swept area larger than 2-5% of the crosssectional area may be considered large, requiring consideration of the effect of the turbines on the flow field. As illustrated in Fig. 3, starting with a small array with a few turbines, the inclusion of additional turbines will reduce the residual available kinetic power density, and each individual turbine will produce less overall power. However, the power produced by the array will still be increased. The rate of the increase in total power will be reduced as more turbines are included, reaching a threshold where adding further turbines will actually reduce the total power output from the full array, thereby leading to a maximum possible power that may be extracted for a given channel configuration.
In order to provide an estimate of the maximum power available, Garrett and Cummins, 36 hereafter referred to as G&C, identified that the backflow effect created by the obstructions in the channel would increase the pressure differential driving the flow and enhance the flow velocity. For a constricted channel connecting two large bodies of water in which the tides at both ends are assumed to be unaffected by the currents through the channel with a tidal fence consisting of turbines across the entire channel cross section, a general formula gives the maximum average power between 20-24% of the peak tidal pressure head multiplied by the peak of the undisturbed mass flux through the channel. Maximum average tidal stream power, P max , is given as where g is a parameter, a is the amplitude of the tidal water level constituent, and Q max is the maximum corresponding tidal flow rate. The maximum average power may be estimated with an accuracy of 10% using g ¼ 0.22, without any need to understand the basic dynamical balance. 36 A multiplying factor is used to account for additional constituents (-, -) given as This upper bound on the available power ignores losses associated with turbine operation and assumes that turbines are deployed in uniform fences, with all the water passing through the turbines at each fence. Using numerical simulations of tidal flows, Haas et al. 37 applied Eqs. (4) and (5) to produce the United States national tidal energy reconnaissance resource assessment. Other studies have looked at various applications of G&C. The validity of the value of g in Eq. (4) was demonstrated by Sutherland et al. 38 using numerical simulations of channel-wide energy extraction. However, they did find that for more complex channel geometries such as split channels, Eq. (4) overestimated the maximum power by up to 50%, demonstrating one of the limits to the applicability of this method. Recognizing the unlikely scenario of installing tidal fences across an entire channel, Garrett and Cummins 32 analyzed partial fences across a channel. Analytically, it was found that the resultant maximum power was reduced by a factor of 1/3 to 2/3. This work was extended by Whelan et al. 39 for the case of high flow blockage where the free surface drops immediately downstream of the turbine, and by Nishino and Willden 40 who considered the case of a tidal fence partially blocking a wide channel. Vennell 41,42 also looked at the effects of partially blocked channels, finding that by adjusting the flow reduction (ratio of the wake velocity to the incoming velocity), the turbine array could be better optimized. Vennell argues that changes to individual turbines will affect the overall array drag and therefore found that flow reduction ratios could be tuned in the range from 1/3 to 1, in contrast to Garrett and Cummins 32 where the optimal limit was 1/3. Vennell 42 also discusses the need to tune multiple rows of turbines "in-concert" because even if the rows are widely separated such that the wake sufficiently recovers, they will still have interactions due to the full array's effect on the overall array drag coefficient. An excellent discussion comparing the results from Garrett and Cummins with the work by Vennell is found in Vennell. 43 The previously discussed studies generally focused on relatively simple channel geometry and idealized flow conditions. The comparison of array efficiency estimates obtained from analytical and numerical models suggests that analytical models underestimate the tidal energy yield potential, particularly as they do not realistically reproduce the flow structure in the vicinity of the turbines, especially threedimensional effects and turbulent mixing. 44 Further, tidal estuaries are frequently much more complex and may contain multiple channel branches and intertidal storage. Under these more complex scenarios, kinetic power density is sometimes still used to estimate the resource. 45-47 However, additional work has been completed to extend the analytical approaches above to more complex geometries. Blanchfield et al. 48 modeled a closed bay and open ocean, which was successfully verified by Yang et al. 49 numerically. Polagye and Malte 50 treated tidal networks like electrical circuits and found the most power efficient turbine deployment in networks required equally deploying turbines across sub-channels, or deploying prior to the channel bifurcation to a reduce flow diversion.
The project design process and final AEP calculations require much higher levels of resource assessments, using additional levels of Starting from the top left, as more turbines are added the residual kinetic power decreases, and the generated power increases. There is an optimal point where generated power is maximized, beyond which the generated power is decreased as further turbines are added.

REVIEW
scitation.org/journal/rse complexity. Individual effects from turbines that must be considered can be small scale, where turbines interact directly (i.e., wake effects), or large scale, where the energy extraction from the turbines affects the tidal flow in the entire estuary. While it is not possible to resolve all these scales in a single model, it is necessary to resolve a broad range of scales, which is computationally challenging. Resource assessments utilized for siting considerations require a much higher model resolution than feasibility studies. This may be accomplished using models with unstructured grids, such as the resource assessment for New Jersey by Tang  In order to account for array effects, numerical simulations of the project site must quantify the effects of turbines on the flow field. Due to the necessity of resolving a large domain to capture the far field effects, simplified approaches to resolving the effect of turbines are generally utilized. While there are many options for incorporating the impacts of turbines into models, they generally have comparable approaches. 49,[57][58][59][60][61] One approach for simulating energy extraction incorporates an extra retarding force into the momentum equations. This force may be written asF where C ext is an extraction coefficient. To evaluate far-field effects, the grid resolution may be relatively coarse where each grid point may represent multiple turbines, leading to larger overall extraction coefficients. At such a coarse resolution, the turbines within each grid cell cannot be individually tuned, and so the optimal grid layout cannot be determined. In order to optimize the turbine layout, individual turbines must be resolved by one or more grid cells. In these cases, threedimensional (3D) models may also resolve the vertical structure of the flow with multiple layers resolving the swept area. 62,63 Tidal turbines and their support structures will also have a pronounced effect on the turbulent characteristics of the flow field. Therefore, models must include modifications to the turbulence closure scheme. 64 This includes an additional turbulent kinetic energy (TKE) source term, an additional term accounting for the transfer of large-scale turbulence to small-scale turbulence, and a term to model reduction in the spectrum of the turbulent length scales due to the partial generation of turbulence from fluid-structure interactions.
There have been numerous tidal power assessments performed while accounting for the effects of tidal power extraction for different regions around the world. Tidal stream energy resources in northwest Spain were modeled numerically, and the impacts of tidal stream energy were assessed. 45, 53 The maximum tidal power potential of Johnstone Strait, British Columbia, Canada, was studied by Sutherland et al. 38 using a 2D finite element model, and the maximum extractable power in northwestern Johnstone Strait was estimated to be about 1.3 GW. The available tidal power from in-stream turbines placed in the Minas Passage of the Bay of Fundy and the Passamaquoddy-Cobscook Bay located near the entrance to the Bay of Fundy has also been examined. [65][66][67] Polagye et al. 68 studied and characterized the instream tidal energy potential of Puget Sound, Washington, and quantified the far-field, barotropic effects of energy extraction. The Kennebec River of the central Maine coast was found to contain narrow passages where mean tidal energy capacity is sufficient to meet the consumption needs of about 150 homes. 69 The Tanana River at Nenana, Alaska, was studied for its potential for installing and operating hydrokinetic turbines, and suitable locations were recommended. 70 Coles et al. 71 performed 2D model simulations around the Channel Islands and found benefits in developing multiple sites simultaneously. Hakim et al. 72 modeled the Muskeget Channel and found modest impacts on the underlying hydrodynamics. Marsh et al. 73 modeled the Clarence Strait in Australia both with and without the effects of energy extraction and found that modest sized arrays had a limited impact on the underlying hydrodynamics. In contrast, Coles et al. 74 modeled the Alderney Race and documented a reduction in flow due to the impact of the turbines, resulting in a total power estimate 57% lower than the previous assessment without energy extraction.

IV. TURBULENCE AND WAKES
Tidal stream energy sites are characterized by high Reynolds numbers. For instance, the order of magnitude is 2:5 Â 10 7 for the Alderney Race, France. 75 The flows are thus highly turbulent, and this has strong implications on the design of turbines. Whereas the smallest scales of turbulence increase the fatigue of the turbines by inducing vibrations, the larger scales are responsible for fluctuations of power 76,77 and loads. 78 Indeed, instantaneous velocities (and associated loads) result from the superposition of turbulence-induced velocity fluctuations and time-mean velocities. The characterization of turbulence is thus crucial for the design of tidal turbines. Turbulence is also known to have a significant effect on the wakes that form behind the turbines, as demonstrated experimentally 79 and numerically. 80 Basically, the higher the turbulence intensity, the greater the mixing behind the turbines and the faster the flow recovery between two consecutive rows of turbines in an array. This effect should thus be considered when optimally positioning turbines within an array, especially the longitudinal spacing between rows of turbines. Finally, as turbulence acts on the mixing processes and bed shear stress, it influences sediment transport and the transport of solute and suspended substances that pass through the array (e.g., biochemical particles). Wake turbulence should thus be considered when assessing the influence of tidal turbines on the physical conditions.
The turbulence at tidal energy sites results from a combination of processes that interact with each other. In this section, we first review studies dedicated to the characterization of the ambient turbulence (the turbulence that is naturally present in the flow). Then, we review investigations on the turbulence generated by the turbines.

A. Ambient turbulence characterization
Whereas measuring time-mean flows at tidal energy sites is relatively common today, measuring turbulence in fast currents is still  [86][87][88] Such measurements are performed with acoustic Doppler velocimeters (ADV) that provide high-frequency data with low noise at a single location, acoustic Doppler current profilers (ADCP) that provide lower-frequency data along a profile (generally, the sensor is upward looking from a sea bed mooring), or two coupled ADCPs that allow the six components of the Reynolds stress tensor to be evaluated. [86][87][88] The most popular metrics used to characterize turbulence are turbulence intensity (i.e., the turbulent velocity fluctuations normalized by the tidal currents), turbulent kinetic energy (TKE), the production and dissipation of TKE, and the time and length scales of the turbulence. Turbulence intensity (in percent) is and turbulent kinetic energy (TKE) is where, according to the Reynolds decomposition, the current velocity (u, v, w) is the sum of a time averaged velocity ð u; v; wÞ and a fluctuating part ðu 0 ; v 0 ; w 0 Þ.
The investigations mentioned above showed that (1) ambient turbulence intensity is of the order of 10% at hub height, 83,86 and (2) turbulence is highly site-specific and strongly dependent on the (ebbing or flooding) tidal conditions because its characteristics are influenced by the (local and upstream) seabed features. 85,86 They also highlighted that (3) turbulence is anisotropic with greater turbulence intensities and time-length scales in the streamwise direction than in the transverse or vertical directions. 88 In situ measurements are essential for understanding the turbulence characteristics at a given tidal energy site. However, the data are generally only available at one particular location (or along a profile if measurements are performed using an ADCP). Hence, numerical models are required to map the turbulence characteristics over the entire tidal stream energy site. There are two main ways of modeling the turbulence in tidal flows. The RANS (Reynolds-averaged Navier-Stokes) approach consists of resolving the time-mean variables and using closure schemes (such as the k À e model) to simulate the effect of the turbulence on the flow. This is a temporal filtering of the Navier-Stokes equations. The other way is LES (large eddy simulation) and consists in resolving the greatest scales of the turbulence and in mimicking the effect of the smallest (dissipative) scales with subgrid models (e.g., Smagorinsky 89 ). This is a spatial filtering of the Navier-Stokes equations. The RANS approach is generally used in regional models because it is suited to simulating the slowly varying tidal dynamics over large domains. Numerous regional models, such as Telemac, ROMS, or FVCOM (Finite Volume Community Ocean Model), have thus been applied to appraise the time-mean flow hydrodynamic characteristics of tidal energy sites as well as the resource. 90 Several attempts were also undertaken to characterize turbulence from the turbulence closures of RANS models. Togneri et al. 91 used the k À e closure of the ROMS model to predict the turbulence properties at the West Anglesey Demonstration Zone off the coast of Wales (UK). The comparison with ADCP data showed good agreement in terms of TKE, especially in the lower part of the water column, where the influence of wave action was minimal. Dissipation agreed less well. Applying a comparable methodology to the Chacao Channel (Chile), Guerra et al. 92 also showed that the regional model FVCOM can suitably predict the strength of turbulence at tidal energy sites. The comparison of the model predictions to ADV measurements showed good agreement in terms of dissipation rate. Nevertheless, TKE was significantly underestimated because the model was not able to capture the anisotropy of the low-frequency turbulence scales. LES models simulate the transient flow characteristics and thus enable more insights into the processes controlling the turbulence at tidal stream energy sites. LES has been performed in Ramsey Sound 93 and the Alderney Race. 94,95 The simulations of Zangiabadi et al. 93 relied on the Fluent model and covered a 800 Â 1800 m 2 domain. This highlighted the capabilities of LES to simulate flow patterns over complex seabed morphology, especially the dynamics of large eddies that form behind a large rock. Focusing on comparable length and time-scales, Mercier et al. 95 characterized quantitatively the turbulence in a 240 Â 960 m 2 zone of the Alderney Race with a LES model based on the Lattice Boltzmann method (Fig. 4). The model validation relied on measurements performed by two coupled ADCPs and showed a remarkable agreement for all components of the Reynolds stress tensor. Because of their high computational cost, LES focusses on small domains and is used to investigate only a particular snapshot of the tidal cycle (generally peak ebb or peak flood current).

REVIEW scitation.org/journal/rse
To cover larger temporal and spatial scales, Bourgoin et al. 94 modified Telemac3D so that it solves both the RANS and the LES equations. In the outer domain, solving the RANS equations resolved the tidal propagation, whereas in the inner domain, application of the LES equations resolved the transient characteristics of the flow and the largest turbulence scales. This model was applied to simulate a tide cycle in the Alderney Race and was used to map the turbulence characteristics, investigating the differences in turbulence properties between the flood and ebb phases of the tidal cycle. Table III synthesizes the main characteristics of the models mentioned above.

B. Wake turbulence
The turbulence induced by tidal stream turbines is the result of different processes. In the near wake (typically within five turbine diameters downstream of the device), the turbulence is mostly governed by the swirl (rotational motion of the flow generated by the blades) and by the vortices shed by the blades and the support structure. In the far wake (typically greater than five turbine diameters), turbulence results mainly from the shear between the accelerated flows that bypass the turbine and the slowly moving flows in the turbine wake. The interaction of these processes with the ambient turbulence determines the wake properties. To predict wake characteristics, different techniques can be applied. Here, we review measurements in real sea conditions, laboratory experiments of scaled studies, and numerical models.
In situ experiments can either be used to assess the influence of turbulence on turbine performance (thrust and power) or to map the wake characteristics (in terms of velocity deficit and turbulence intensity for instance). Whereas several in situ experiments highlighted the influence of ambient turbulence on the performance of full-scale tidal turbines, 76,77,96,97 measurements of wake turbulence in real sea conditions are sparse in the scientific literature. The main reason is that acquiring data in the wake of a full-scale tidal turbine (in operating conditions) is highly challenging. Indeed, in addition to the technical limitations of current measuring devices, sensors are difficult to deploy close to the turbine, and multiple sensors are required to capture the shape and properties of the entire wake. In addition, data are generally not publicly available. The only existing studies on wakes found by the authors are those of Schmitt et al. 98 and Verbeek et al. 99 Schmitt et al. 98 investigated the spatial distribution of the velocity deficit in the wake of a 4 m diameter turbine with two ADCPs and an ADV.
Verbeek et al. 99 measured, using two ADCPs, the velocity deficit, turbulence intensity, and the integral time-length scales behind a row of 5.3 m diameter turbines.
As wake data acquired in real sea conditions are sparse, the knowledge on the wakes of tidal turbines comes mainly from experiments on scaled turbines in laboratory flumes. 100 The first measurements were carried out by Myers and Bahaj 101 who mapped the near wake behind a 0.8 m diameter turbine with laser and acoustic Doppler velocimeters. Measurements in the far wake were also performed using an experiment in which the turbine was represented by a porous disk of diameter 0.1 m. Numerous experimental studies were then performed on the wake of a single turbine and showed, in particular, that the ambient turbulence has a strong influence on the flow recovery behind the turbine. 79,[102][103][104][105] Investigations were also conducted on the interactions between the wakes of two turbines, 106 three turbines, 107 and up to ten turbines. 108 Finally, large arrays of tidal turbines were investigated by Coles et al. 71 with porous fences representing several rows of devices. This experiment enabled an investigation of flow development through up to ten rows of turbines.
Experimental studies are essential for understanding the processes controlling wake characteristics. However, the results are difficult to transpose to full scale, especially because the Reynolds similarity is not achievable experimentally. Numerical investigations are thus widely used to characterize the wakes of tidal stream turbines. Here, we present different types of wake models starting with the most sophisticated. Blade resolved models (e.g., Afgan et al. 109 ) can be used to simulate the performance of turbines under different flow conditions. In such an approach, the rotating blades, the hub, and the support structure are physically represented in the model. Since they resolve the blade and support structure, they are able to simulate all turbulent processes generating turbulence in the near wake. However, due to their high computational cost, they cannot be applied to simulate an entire wake (or several interacting wakes). Blade element method (BEM) consists of applying on the fluid the forces experienced by different sections (elements) of the rotating blades (lift and drag forces). BEM is particularly suited to investigating turbine and array scales, and numerous investigations have been conducted to simulate isolated or superimposed wakes. 110,111 The vortex method 112 focuses on comparable scales and relies on a velocity-vorticity numerical implementation of the Navier-Stokes equations. In this approach, the vorticity is carried by particles emitted at the trailing edge of the turbine blades and advected in a Lagrangian framework. This method showed good agreement with experiments on a scaled turbine both in terms of performance and wake characteristics. The Actuator Line  113 Actuator Disk (AD), 80,114,115 or Actuator cylinder (for vertical-axis turbines) concepts represent turbines by applying on the fluid the force exerted by the turbines (thrust and eventually drag). This modeling approach has a lower level of fidelity then the methods mentioned above (especially because the rotational motion of the blades is neglected), but can successfully simulate far-field wakes (Fig. 5), provided the turbulence models are adequately tuned 116 (and suitably mimic the effects of the unresolved turbulence processes, especially swirl and the shedding of vortices). The main interest of AD methods is that, due to their simplicity, they can be integrated into regional models (e.g., ROMS, Telemac3D). Hence, they enable wake studies to be performed under realistic sea conditions (e.g., Roc et al., 64 Michelet et al., 62 Thi ebot et al. 63 ), which are required to assess the energy conversion potential of tidal arrays.

V. ENVIRONMENTAL IMPACTS
Tidal stream energy projects are generally located in coastal bays, estuaries, and tidal channels that are characterized by strong tidal currents. These coastal waterbodies present not only a great potential for electricity generation but also provide important nearshore habitats for many species of marine wildlife, including seabirds, fish, and marine mammals. 117 The deployment of tidal turbine arrays in the coastal oceans will cause disturbances to the ambient ocean environments. The level of disturbance, or impact, depends not only on the amount of energy extraction and size of the tidal turbine arrays, but also the geometric features and volume of the tidal system where tidal turbine arrays are deployed. 118 Therefore, to accelerate the development of tidal energy, it is important to assess, understand, and minimize the impacts of the deployment of tidal turbine arrays on coastal ecosystems, and to protect the marine wildlife while maximizing its contribution to the renewable energy portfolio. Environmental effects of tidal turbine array installations on tidal systems typically include the following areas: 117,118 • Physical processes -Effects on the far-field flows and water levels -Transport timescales such as residence time • Biogeochemical processes -Water quality -Sediment erosion and transport near the tidal turbine arrays Common approaches for assessing the environmental impacts of tidal energy extraction include numerical modeling, laboratory experiments, field monitoring, and measurements. However, field monitoring and measurements are limited because very few tidal turbine arrays have been deployed in the world, and the associated costs are extremely high. Therefore, numerical simulations using advanced state-of-the-art models are often employed to evaluate the level of environmental impacts associated with tidal energy extraction in marine systems. 35,63,119 The following two subsections (V A and V B) provide an overall review of the current state of research for the environmental impacts in each subject area.

A. Physical processes
Effects of tidal stream energy extraction on physical processes in marine systems, such as ambient flow fields, are the most evident phenomena in tidal energy projects. Assessment of the effects on physical processes generally can be divided into two groups-system-wide farfield effects and localized near-field effects. In a series of pioneering work on understanding and estimating the maximum power potential in tidal channels, Garrett and Cummins 32,36,120 established a theoretical formula that relates the maximum extractable energy vs the

REVIEW
scitation.org/journal/rse maximum tidal flux across the channel. They discovered that the maximum extractable power is significantly less than the average kinetic energy flux in the undisturbed tidal channel, and the volume flux across the channel will be reduced by 42%. Subsequently, a number of modeling studies confirmed the conclusions by Garrett and Cummins, 36 using numerical simulations in real-world cases and idealized test domains, and further extended the formula to a tidal channel connecting to a coastal bay case. 38,48,49,66,121 It is important to note the previously mentioned studies were based on one-dimensional (1D) or two-dimensional (2D) governing equations. However, in reality, tidal energy extraction is a three-dimensional (3D) process. For example, Yang et al. 49 found that the volume flux is reduced by 32.7% when maximum tidal energy is extracted in a 3D case, which is smaller than the 2D case and the results of Garrett and Cummins. 36 Physicsbased 2D and 3D models are often used to investigate the far-field effects of tidal energy extraction for real-world study sites. Most studies showed that the effect of tidal energy extraction on water level was small. 59,[122][123][124] For example, Zhang et al. 124 applied a 3D model to assess the impact of proposed tidal array with 115 turbines in the Zhoushan Archipelago of China and they found the impact of the tidal turbine array resulted in a water-level change of less than 3 cm. Similarly, Yang et al. 123 also found that a hypothetical tidal array with 100 turbines in Tacoma Narrows, USA, resulted in a change in water level of less than 1 cm. Pacheco and Ferreira 125 investigated the hydrodynamic impacts of a small array of thirty E35 turbines and found the change in a water level due to tidal turbine array operation was less than 2.5 cm. Hakim et al. 72 studied the impacts of tidal stream turbines on hydrodynamics and sediment transport in the Muskeget Channel, Massachusetts, using an unstructured-grid model FVCOM. Model results suggested that water level was only affected by 0.8% when 9% of the natural tidal energy in the channel is extracted. However, some studies found that the presence of large arrays of turbines can potentially result in noticeable changes in water level, especially where tidal flats are present in the system or when tidal phase is affected. For example, Karsten et al. 66 discovered a maximum of 7 GW of power can be extracted by turbines in the Minas Passage based on numerical simulations. However, such a large level of energy extraction could result in the system moving closer to true tidal resonance, increasing tidal amplitudes throughout the Bay of Fundy and Gulf of Maine. In a modeling study on the effect of tidal turbine arrays on the tidal regime, intertidal zones, and flushing time, Nash et al. 126 pointed out the tidal range upstream of the tidal arrays could be affected significantly, varying from approximately 5-42% depending on the size of the tidal array. Similarly, intertidal mudflats could be affected by 14-32% and thus result in loss of nearshore habitat.
In addition to currents and water level, another important parameter used to characterize the far-field, or system-wide effect of tidal stream energy extraction is the transport timescale, such as flushing time, residence time, and age, of a system where the tidal turbine arrays are installed. Information on the effects of tidal energy extraction on residence time and water renewal can be used to guide the optimal siting and layout of tidal turbine arrays during the permitting process. A detailed review on the methodologies of assessing the impact of tidal stream energy extraction on transport timescales was provided by Yang and Wang. 127 For example, based on simulated tracer concentrations with and without tidal turbine arrays (Fig. 6), Yang and Wang 127 showed that if a 10% change in flushing time is the acceptable upper limit for the environmental concern due to tidal energy extraction in Tacoma Narrows, USA, a smaller percentage reduction in volume flux must be considered because the change in flushing time is several times greater than the change in volume flux. Guillou et al. 128 used model simulations to evaluate the effects of tidal turbine arrays on water renewal in a tidal stream energy site of the Fromveur Strait in northwestern coastal waters of Brittany, France. Their study demonstrated that the residence time in the tidal stream energy site was modified by less than 5.3% with an array of 207 tidal turbines. Nelson et al. 129 developed a modeling framework for optimizing tidal turbine deployment in Cobscook Bay in Maine that maximizes the energy extraction with consideration of environmental constraints, such as criteria for the changes to flow fields and bed shear stress. James et al. 130 applied a similar modeling approach to design turbine arrays in the Tanana River, Alaska, and evaluated the effects of different array layouts on power production and environmental impacts. Journal of Renewable and Sustainable Energy REVIEW scitation.org/journal/rse

B. Biogeochemical processes
Marine biogeochemical processes, such as water quality, sediment transport, nutrients, and marine habitats, are driven by coastal hydrodynamics. When the physical processes, including flow field, turbulence, vertical mixing, residence time, and bed shear stress, are modified due to the operation of tidal turbine arrays biogeochemical processes will also be affected. 119,123,131,132 Changes to water quality such as dissolved oxygen, nitrogen, turbidity, and primary production in a marine system will ultimately affect the marine ecosystem services. Modification of turbulence characteristics and bottom shear stress within the vicinity of the tidal turbine array will also cause sediment erosion, re-suspension, and deposition. While there have been many studies conducted to investigate the effects of tidal energy extraction on the ambient hydrodynamics, studies on the effect of tidal energy extraction on biogeochemical processes are extremely limited. Therefore, there is a strong urgency to promote the research to assess the effects of tidal energy development on biogeochemical processes and minimize the negative impacts.
Wang et al. 132 discovered that tidal stream energy extraction in a tidal channel can result in either positive or negative impacts to the water quality of a system, based on an idealized case of a tidal channel connecting with a coastal bay. Energy extraction by the tidal turbine array decreased flushing rates in the bay, which causes a negative impact on the water quality in the coastal bay. However, tidal energy extraction also enhanced vertical mixing, which could lead to higher bottom dissolved oxygen. Ashall et al. 133 simulated tidal energy extraction in the Mina Passage of the Bay of Fundy, Canada, and found that a high power extraction case (5.6 GW) will cause 15% change in tidal amplitude, 10% change in velocity, and 37% change in suspended sediment concentration (Fig. 7). Nash et al. 126 showed that the flushing time would be significantly altered if tidal turbine arrays were installed in the Shannon Estuary in the west of Ireland. Ahmadian et al. 134 applied a hydro-environmental model to investigate the impacts of tidal energy extraction on water levels, tidal currents, and sediment and fecal bacteria levels in the Severn Estuary and Bristol Channel, UK. Their study found that 200 tidal turbines with 20 m diameter would cause noticeable changes of current speed and suspended sediment concentrations within 15 and 10 km from the turbine array, respectively. Shapiro 135 applied a 3D coastal ocean model to simulate the effect of tidal energy extraction on far-field current distribution and transport process in the neighboring Celtic Sea. By simulating the presence of a hypothetical tidal array, current speeds in the vicinity of the array were significantly reduced due to tidal energy extraction and the backwater effect. In particular, residual tidal currents and pollutant transport were affected as far as a hundred kilometers away from the tidal turbine array.
A detailed review was provided by Neill et al. 136

REVIEW
scitation.org/journal/rse numerical modeling. Their study showed that the deployment of a tidal energy array resulted in reduced velocity and increased sediment deposition within the vicinity of the tidal array, and increased velocity accompanied by sediment erosion along either side of the tidal array. It is also important to consider the impacts on sediments in relation to natural variability. For example, during summer months, when natural variability of waves and hence sediments is lower, the impact of tidal energy extraction will be relatively greater than during the energetic winter months. 141 Prior to the construction of large arrays of tidal energy devices, impacts on sediments can only be assessed via numerical models. There are various techniques to incorporate energy extraction into numerical models 139,141,142 and no standardized methodology, and therefore, in situ validation of such models and benchmarking against laboratory experiments is an important step. Finally, tidal range schemes (Sec. II) are likely to have significant impacts on the sediment dynamics of a region, due to the presence of the artificial embankment and the operation of the turbines. The region inside the lagoon will be characterized by reduced energy, especially during periods of holding, 136 leading to reduced sediment concentrations in the water column and increased sedimentation. 143 Closely spaced turbines in one section of the lagoon wall will lead to strong wake effects, further disturbing the sediment regime-increasing the spacing of these components of the lagoon along the embankment as much is practically feasible will reduce such impacts. 144

VI. DISCUSSION-FUTURE RESEARCH DIRECTIONS A. Semi-diurnal and fortnightly variability
One of the major advantages of tidal energy is predictability. However, this predictability is associated with variability at two main timescales: semi-diurnal and fortnightly (spring/neap cycle).
It could be possible to phase tidal power plants along a coastline to minimize the semi-diurnal variability of the aggregated power signal, 14 but it has been demonstrated that for shelf sea regions, the currents at the most energetic "hot spot" sites are either in phase with one another or 180 out-of-phase, 145 and hence so would be the aggregated power signal. However, by diversifying tidal stream site selection to incorporate less energetic tidal current amplitudes in the energy mix, the resulting increased phase diversity has considerable advantages in minimizing variability at semi-diurnal timescales. 13 At fortnightly timescales, since the tides everywhere on Earth are simultaneously affected, it is not possible to significantly reduce variability. However, by selecting sites that exhibit larger ratios of M2:S2 amplitudes, it is possible to optimize multiple site selection, reducing fortnightly variability. 90 Such a reduction could also be achieved by optimizing the joint development of a range of semi-diurnal, diurnal, and mixed sites, considering the K1 and O1 constituents alongside the principal semi-diurnal constituents M2 and S2. 146 However, to counteract the spring/neap cycle, other forms of low carbon (renewable) energy conversion, or significant grid storage, will likely be required.
The development of less energetic tidal sites (to increase phase diversity) comes with its own set of challenges. 55 Such locations tend to be in deeper water, further from the coast (and hence grid connections), are potentially more exposed to waves, and the currents are less rectilinear. Therefore, the development of such sites requires a new type of device (e.g., a floating technology with yaw capabilities) and a new set of research questions (e.g., the interaction of waves with currents at such sites). 147 However, in addition to the advantage of increased phase diversity, there is also considerably more sea space available at such sites (Fig. 8). Taking an M2 current amplitude of 2.5 m/s as the current technology standard, 55 lowering the M2 current amplitude to 2 m/s approximately doubles the available sea space. Further reduction of the threshold to 1.5 m/s increases sea space by a factor of 8 compared to sites suitable for current technology. Note also that less energetic tidal sites, far offshore and so not suitable for grid connection, could be used to power the "blue economy," meeting the power demands of, for example, weather buoys, tsunami warning devices, communication, defense, and aquaculture applications. 148

B. Turbulence
As seen in this review, numerous investigations have been performed that rely on numerical models and/or measurements performed in real sea conditions or in the laboratory. Despite these efforts, the understanding of processes controlling turbulence of tidalstream energy sites is still incomplete for various reasons: (i) the difficulty with transposing experimental results to realistic sea conditions (which is partially due to similitude problem), (ii) the limitations of actual CFD (Computational Fluid Dynamics) models that cannot resolve the entire range of scales, and (iii) the scarcity of measurements in real sea conditions (for both ambient turbulence and wake-added turbulence). Future research is thus necessary to reduce uncertainties in turbulence predictions. The challenge will be to reliably model the interactions between ambient turbulence and the turbulence resulting from multiple (superimposed) wakes, and to predict the characteristics of wakes in real sea conditions, that is, with irregular sea bed, complex flow structure, and under the combined influence of waves and wind. Furthermore, existing studies show that the turbulence at tidal-stream energy sites is highly variable in space. To explain this heterogeneity, it is necessary to investigate the links between the local characteristics of turbulence and the (local and upstream) seabed morphology. In particular, it is not clear yet how different sea bed features (e.g., micro or macro seabed roughness, the presence of large boulders, abrupt changes of bathymetry) influence the turbulence. Finally, turbulence measurements are only available at specific locations, and so spatial coverage is incomplete. Although numerical models allow us to investigate the spatial distribution of turbulence, computational constraints mean that these models only cover short periods of time (ranging FIG. 8. Availability of UK shelf sea area (for water depths less than 100 m) for a range of M2 current amplitude thresholds.

REVIEW
scitation.org/journal/rse from minutes to hours), thus preventing investigation of the wide range of realistic tidal flow conditions (e.g., spring/neap tidal cycles). It is thus important to combine different methodologies (numerical, laboratory, and in situ measurements) in order to achieve a better understanding of processes controlling the generation and the transport of turbulence.

C. Improved resource assessments
An ongoing challenge for the industry is how to best perform resource assessments and characterizations in a cost-effective manner, particularly for smaller scale projects. One option for addressing this challenge is the application of a measurement-based approach for resource assessments. The IEC technical specification for tidal energy resource assessment 24 specifies a methodology in which a fixed current profiler can be used to collect 90 days of continuous data. However, additional approaches using short-term high-resolution measurements need to be developed to eliminate the requirement for long-term measurements at the location of each turbine. These approaches can adopt recent developments in remote sensing applications for measuring tidal currents. For example, initial analysis of different measurement approaches for reconstructing volumetric currents including vessel mounted profilers, X-band radar, and surface floats has recently been completed. 149 Ongoing work needs to further develop these methodologies, combining short-term and long-term measurements to create the velocity probability distributions necessary for performing accurate AEP calculations.

D. Environmental impacts
Over the last decade, many studies have investigated the potential environmental impacts of tidal energy extraction on physical and biogeochemical processes in coastal bays and estuaries. Most of these studies are based on numerical simulations. However, model validation of the biogeochemical processes was rarely performed because of lack of field measured environmental data, especially pre-and posttidal array installations. Therefore, conclusions drawn from unvalidated model simulations may be associated with high uncertainties. To increase model confidence in predicting the environmental impacts due to tidal energy extraction, it is important to collect field data relevant to environmental impacts to support model validation. Environmental impact modeling and related field measurements should be considered as an integrated element throughout the course of tidal array development. It is important to understand that tidal energy extraction is a 3D problem since tidal turbines are typically located at certain depths (hub height) in the water column, in the form of either free-floating or bottom-moored systems. Therefore, the environmental impacts associated with tidal energy extraction are also three-dimensional. To simulate the far-field and localized environmental impacts using numerical models, selection of models and specification of model resolutions are important. 150 It is not sufficient to use depth-averaged 2D numerical models in the context of tidal energy extraction, since 2D models are not able to simulate vertical variability in the flow field and the associated water quality and suspended sediment dynamics in the water column. To accurately assess the environmental impacts associated with tidal energy extraction, 3D models should be used. More importantly, water temperature, salinity, vertical mixing, stratification, and estuarine circulation induced by density gradients should be considered. 151 The biogeochemical processes must be dynamically coupled with the coastal circulation models with the capability of simulating tidal energy extraction. It is worth noting that the effects of tidal energy extraction on biogeochemical processes may not always be negative, because of the complex interplay of multiple factors. For example, increased mixing in the water column as a result of tidal energy extraction could potentially reduce the hypoxia problem in an estuary. 132 Furthermore, because estuarine biogeochemistry and sediment transport are long-term processes, model simulation periods should be of sufficient length to cover the seasonal variabilities that are governed by other factors such as river discharge, heat flux, and longer-term sea-level rise.

VII. CONCLUSION
Tidal energy, consisting of tidal range and tidal stream, has huge global potential, but there are several research challenges that must be addressed before realizing even a fraction of this potential. In this article, we have identified several evolving research areas that will help enable this goal. These are: • Exploitation of the phasing between multiple tidal range or tidal stream locations to minimize variability in the aggregated power output. Since there is more phase diversity offered by less energetic sites, the unique site characteristics of these locations lead to additional related research topics, including the interaction of waves and currents at such exposed sites. • Improvements in simulating the interactions between the ambient turbulence and the turbulence resulting from multiple (superimposed) wakes, and in predicting the characteristics of wakes in real sea conditions, that is, irregular sea bed, complex flow structure, and under the combined influence of wind and waves. • Improved resource assessment and characterization through novel approaches using short-term high-resolution measurements (e.g., using X-band radar and vessel mounted profiles) to eliminate the requirement for long-term measurements at the location of every turbine within an array. • Increased confidence in predicting the environmental impacts of tidal energy extraction, by collecting appropriate field data relevant for validating numerical models that are used to assess impacts to a range of physical and biogeochemical processes. Environmental impact modeling and supporting field measurements should be considered as an integrated element throughout the entire course of tidal power plant development.