Other important direct measurements of the atmosphere include air temperature, atmospheric pressure, and relative humidity. All three are used to determine air density, which directly impacts turbine performance; as such, they should be measured at turbine hub height. However, if this is not feasible, height adjustments can be made using simple assumptions. Relative humidity and temperature also influence the corrosion potential of materials and coatings. Air temperature is also a factor in determining the probability of a turbine exceeding its safe operating and survival envelope (Stout, 2013). The vertical profile of temperature can be used to estimate the thermal stability of the atmosphere, as can the temperature differential between the sea surface and the overlying air.

For cases where wind and other meteorological variables are not observed at hub height, extrapolation techniques are available to adjust values from other heights. For example, measured wind speeds can be adjusted to hub height using the logarithmic wind profile assumption or the power law using an appropriate shear exponent. Both extrapolation techniques are sensitive to the atmospheric stability and work most reliably under near-neutral stability conditions. To minimize the uncertainties associated with extrapolation, measurements should be taken as close to the desired height as possible.

Additional weather variables that can impact wind farms and their operation include precipitation, solar radiation, lightning, and visibility. Precipitation comes in many forms—rain, freezing rain, hail, snow, etc.—and can impact turbine performance, such as from ice buildup on blades. Solar radiation data can be used to approximate blade deterioration rates and for sizing ancillary power supplies; it can also be an input for some atmospheric stability classification methods. Lightning frequency and characteristics data are useful in estimating damage and downtime risks. Visibility data are relevant to wind farm navigation marking requirements, for assessing visual impacts from shore, and to support vessel operations for construction and operations.

Table 3.1 lists recommended meteorological data parameters to be measured and derived. Most nonwind meteorological parameters are sampled and recorded at the same frequency as wind data. Cloud-to-ground lightning statistics can be obtained from government- or privately run detection networks. Hurricane/typhoon/cyclone statistics are available from government meteorological offices.

Currents within the water column, which can vary with depth, are also referred to as current profiles. A current profile consists of wind-generated near-surface currents; subsurface currents induced by tidal fluctuations, large-scale circulations, or density gradients; and near-shore currents. The water level and its range consist of an astronomical tidal fluctuation and any storm surge. The juxtaposition of the two results in the maximum range of water levels expected at a site.

Other relevant water-related variables include water temperature, density, salinity, conductivity, and ice. The parameters affecting the density of sea water, such as salinity and temperature, also affect the structural loading due to the water flow. In addition, the presence of sea ice and its physical properties can greatly affect structural loading in cold climates.

Corrosion potential may be estimated from observation of water chemistry or pollution, and salinity. Water conductivity measurements are frequently used to estimate water salinity, given known or assumed proportions of dissolved salts. Water temperature also affects corrosion rates, in addition to influencing structural loading characteristics through marine growth.

Estimates of storm surge and sea ice properties are ocean surface observations. For all other parameters listed here, observations throughout the depth of the water column are essential for accurately gauging conditions at development sites.

Table 3.2 summarizes the recommended set of oceanographic data parameters relevant to offshore wind farms.

Table 3.3 lists relevant joint metocean parameters. The desired parameters, their applications, and the available measurement and modeling technologies will likely expand over time, so it is imperative to remain abreast of offshore wind industry developments and new data needs.

It is worth noting that many of the parameters presented in Tables 3.1–3.3 are common to multiple references, recommended practices, and applications; however, the means of measuring, calculating, and/or analyzing them can vary significantly. Design standards and guidelines provide procedures for certain parameters where consensus or industry best-practices are available. Among these references, however, differences exist in procedures and requirements, and many do not address all parameters equally. Some meteorological parameters, for example, may be called out as relevant for design consideration, with no guidance on how to collect, analyze or interpret them. In other cases, characteristics of certain metocean parameters—relevant measurement frequency, return period, extrapolation method, etc.—will be affected by project location and application.

There are three main types of sensors used on satellites that detect ocean winds. Passive microwave radiometers measure different microwave frequencies which are passively emitted from the ocean's surface. The spatial resolution is on the order of 25 km, which limits the ability to resolve winds within this distance of a coast or island due to the corrupting influence of land. The first satellite of this type—SSM/I—was launched in 1987 by NASA, and several others have been launched since. Other satellites of this type include TMI, AMSR-E, and WindSAT.

Scatterometers are another sensor type which emit microwave pulses at the earth's surface and receive the return signal using a common antenna. Their resolution is also approximately 25 km, which similarly limits their application in the vicinity of coastlines and islands. Satellites using scatterometers include Quick SCAT/Sea Winds and ASCAT/METOP-1.

The third sensor type, which also actively emits and receives pulses, is synthetic aperture radar (SAR). In addition to surface wind definition, this technology is also used for wave measurements, oil spill detection, and other applications. SAR has the advantage of much finer resolution (generally 25 m) than the other sensor types and can therefore observe offshore winds close to shore. A disadvantage is that SAR has a much narrower field of view and provides coverage of the same given area less frequently. Satellites using SAR include: RADARSAT-1, ERS-2/SAR, ALOS, and RADARSAT-2.

In all cases, the accuracy of satellite-derived wind speed estimates at 10 m above the ocean surface is on the order of 1–2 m/s. Extrapolation to heights approaching 100 m introduces significant uncertainty when estimating hub height wind speeds. It should also be pointed out that the frequency of satellite imagery for any one particular area is usually limited to twice a day or less. This frequency does not provide detail about the diurnal nature of the wind. In general practice for offshore wind energy purposes, satellite information is best used as a first-order siting tool and an indicator of regional wind speeds.

Purpose-built meteorological towers are the primary method for measuring site-specific wind conditions at a proposed offshore project site. The most common design is a self-supporting lattice structure, as shown in Fig. 3.7(a); Fig. 3.7(b) depicts an alternative tapered tubular design. The objective of these structures is to provide a safe and stable platform from which wind, ocean, and meteorological data can be taken for one or more years. Below the tower portion is a foundation section that is attached securely to the seabed. In addition to a large suite of sensors and mounting booms, the towers must also accommodate other features: a power supply (including battery storage), data logging and communications equipment, aviation obstruction lighting, aids to navigation, docking capabilities, biological monitoring systems, and personnel safety features. The entire tower system must be designed to withstand hurricane force winds, extreme waves and currents, and ice loading. Corrosion of structural steel must be addressed, as must the potential for scour in the design of the foundation.

International standards do not yet exist to specify how offshore wind measurements must be taken, but industry best-practices encourage high levels of redundancy and sensor robustness to ensure reliable data capture. For example, three sets of anemometry are recommended for each measurement level, as opposed to two sets for land-based measurements, to achieve the desired target of 90% or greater data recovery. This is because offshore environments are harsher and more difficult to access when maintenance is required. Offshore towers are also more massive and, consequently, more likely to experience flow distortion issues. Sensors extending off of multiple faces of a tower can facilitate the detection and subsequent correction of flow distortions. IEC specifications (IEC 61400-12-1) recommend that wind sensors be placed on booms extending at least three to six tower widths away from the sides of towers, depending on tower solidity, to minimize distortion effects.

While wind measurement programs for land-based wind projects employ multiple towers, offshore projects will most likely invest in at most one. For the most part, this is because of the high cost of an offshore tower ($10 million ± 50%, depending on locale and other factors), which is roughly two orders of magnitude greater than a typical land-based tower. Another consideration is that the need for multiple towers is much less justified offshore because of terrain and surface roughness uniformity within proposed project areas. In practice, a singular tower is usually positioned just outside, and upwind of, the perimeter of the proposed turbine array. This placement allows the tower to continue supplying useful observations even after the array is installed and commissioned. Numerical modeling is used to extrapolate the tower's observed conditions throughout the project area.

The high cost of offshore towers and the practical limits of their installation in waters deeper than approximately 40 m have opened the door to alternative measurement approaches. The leading candidate is profiling lidar, which can be mounted on fixed or floating platforms; sodar has been used on occasion but is suitable only on fixed platforms. On fixed platforms (such as short towers or jack-up barges), measurements taken by either remote sensing technology have shown a high degree of comparability and correlation with concurrent wind observations taken from adjacent tall towers (Cox, 2014; Hung et al., 2014; Barthelmie et al., 2003). These technologies can complement tower-based measurements by sampling winds above the top of the tower, such as across the upper portion of the turbine rotor span.

Unless an existing fixed platform is available, it may be cost-prohibitive to invest in a new, purposely built one to support lidar or sodar. Buoy-mounted floating lidars (see Fig. 3.8) are now commercially available that have demonstrated the ability to measure winds aloft with essentially the same accuracy as tall-tower measurements. However, more extensive validations on long-term measurement reliability are needed before the offshore wind industry fully accepts floating lidar as a replacement for tall towers. A drawback to reliance on lidar data alone is that, unless more than one colocated lidar is used, there is no backup measurement system should there be an operational problem. Furthermore, lidar is unable to measure temperature profiles, which can readily be observed from a tall tower.

A measurement period of at least one full year, and preferably two or more, is required to understand the magnitude and variability of winds and other conditions across all seasons. Compared to the 20+ year life of a wind project, this duration is relatively short and will not adequately represent long-term conditions. The measure-correlate-predict (MCP) method is commonly used to derive a long-term climatology of wind conditions from the onsite observations. The MCP process works by correlating the site's wind observations with concurrent data from a high-quality long-term reference located within the same region, such as a coastal weather station or model-generated reanalysis grid data. The established relationship (such as a regression equation) between the two is then applied to the reference's historical record to predict the long-term average wind characteristics for the project site. This approach assumes, of course, that the past is a reliable indicator of the future. This has been found to be generally true within an uncertainty margin of approximately 2% for well-correlated sites (r² > 80%) where the duration of the reference station's wind records is at least 10 years (Brower, 2012).

The lack of available wind speed measurements at or near the hub height of modern offshore wind turbines contributes significant uncertainty to wind speed estimation. The majority of publicly available offshore wind data are collected by buoys at anemometer heights of 5 m or less. Satellite-derived estimates of ocean winds are available at a 10 m height. Fig. 3.9 shows a broad range in hub height speed estimates that would result from using a range of power law shear exponents to extrapolate a known wind speed value from 5 m above the surface up to 120 m. The 5-m wind speed value of 6.7 m/s was the measured annual average observed by a north Atlantic buoy in 2013, while estimated 80-m wind speeds varied from 7.5 to 11 m/s. The average shear exponents represent a range of values that are representative of an offshore environment:

Because the shear exponent that should be used cannot be precisely determined without directly measuring the shear, and the shear may also change with height, buoys or satellite-based estimates alone are insufficient for deriving reliable information about hub height wind conditions.

The typical model resolution for most mesoscale simulations is on the order of a few kilometers, ie, near the interface between the microscale and mesoscale. Since this scale does not provide a very detailed picture of wind conditions within a large wind farm, coupling with a microscale model is often done to obtain the desired detail. It has been demonstrated that a coupled mesoscale NWP and microscale model shows improvement over a mesoscale model alone. Examples of coupled mesoscale and microscale models include AWS Truepower's MesoMap and SiteWind systems (Brower, 1999), Risø National Laboratory's KAMM-WAsP system (Frank et al., 2001), and Environment Canada's AnemoScope system (Yu et al., 2006). Coupled model approaches have been used to create relatively high-resolution wind maps and atlases of the globe. Fig. 3.11 is a representation of the annual average wind speed of the United States, including a 90-km wide zone of offshore winds, at 100 m above the surface at a spatial resolution of 2 km (Elliott et al., 2010; Schwartz et al., 2010). In northern Europe, the NORSEWIND (NORthern Seas Wind Index Database) project, which began in 2008, has develop wind atlases for the Irish Sea, the North Sea, and the Baltic Sea using offshore wind measurements, satellite data, and numerical model data (Hasager et al., 2010).

Mesoscale models take into account subgrid scale effects and physics parameterizations for solar radiation, land surface–atmosphere interaction, the planetary boundary layer (PBL), turbulence, cloud convection, and cloud microphysics. Since they incorporate the dimensions of both energy and time, NWP models are capable of simulating such phenomena as thermally driven mesoscale circulations (eg, sea breezes, thunderstorms) and atmospheric stability, or buoyancy. In the world of mesoscale modeling—as in the real world—the wind is never in equilibrium with the surface because of the constant exchange of energy. This exchange occurs through solar radiation, radiative cooling, evaporation and precipitation, and the cascade of turbulent kinetic energy down to the smallest scales.

In addition to forecasting weather conditions, NWP models are useful in predicting atmospheric conditions for historical periods, ie, looking backwards in time. This practice is sometimes referred to as “hindcasting” and has the capability of assessing regional wind conditions from existing long-term datasets before launching new measurements at an offshore site of interest. A number of mesoscale gridded datasets are now available on a global basis from various sources spanning several decades. Referred to as “reanalysis,” these datasets have been compiled using a fixed data assimilation approach and NWP model with the primary goal of removing potential biases or artificial trends resulting from the gradual changes in modeling approaches, observation types and regional data collection concentrations over the decades (Kistler et al., 2001). For example, from the 1940s to the 1970s, weather observations were primarily derived from fixed surface weather stations, buoys, weather balloons, ships, and aircraft. Beginning in the 1970s, satellite-based observations of cloud-tracked winds and other parameters began, and since then significant increases in the number of satellites and types of onboard sensors have made satellites the dominant environmental data gatherer across the globe. Even among the nonsatellite types of measurement systems, over time there have been large changes in the density and number of surface and upper air observations, improvements in the quality of the data collected, and the introduction of new data-recording and -sensing technologies and the retirement of old ones. Reanalysis datasets, therefore, provide the most consistent records of atmospheric conditions over long periods of time.

The original reanalysis dataset is known as the National Center for Environmental Predictions/National Center for Atmospheric Research (NCEP/NCAR) Reanalysis 1 (Kalnay et al., 1996). It covers the period from 1948 to present at a spatial resolution of 1.87 degree (approximately 205 km). Since then, several national meteorological agencies and national research laboratories including the European Center for Medium-Range Weather Forecasts (ECMWF), NCEP, National Aeronautics and Space Administration (NASA), and the Japanese Meteorological Agency (JMA), have issued their own reanalysis products. These products include the ERA-Interim, Climate Forecast System Reanalysis (CFSR), Modern-Era Retrospective analysis for Research and Applications (MERRA), and Japanese 55-year Reanalysis (JRA-55). They are based on advanced data assimilation schemes and NWP models and have been generating data at a finer spatial resolution of 0.5–0.75 degree (55 or 83 km) than the NCEP/NCAR Reanalysis 1. Reanalysis data are typically available on a 6-h interval, however there are two exceptions: the MERRA and CFSR, which are available hourly for some surface fields and limited pressure levels.

The relatively coarse grid resolution of reanalysis data (50 km or so) can capture offshore wind flows well if the site of interest is located far enough from the coast (or islands) such that the model grid cell does not include any land portion. However, nested, higher-resolution grids can be modeled to simulate near-shore wind circulations. Reanalysis datasets are also valuable for correlating short-term time series measurements collected on offshore platforms with long-term climatological records. Even though the mean bias between reanalysis and meteorological mast data can be substantial, the value of reanalysis data relies mostly in their correlation to onsite measurements, which are not impacted by a bias. Several studies (Brower et al., 2013; Lileo and Petrik, 2011; Decker et al., 2012; Stoelinga et al., 2012) show that the latest-generation reanalysis datasets—for instance, ERA-Interim, CFSR, and MERRA—have superior accuracy in terms of their correlation to mast data.

A version of CFD models known as Reynolds-averaged Navier–Stokes (RANS) models has emerged as an alternative to Jackson–Hunt models for wind energy applications as personal computers have grown more powerful. They were designed originally to model turbulent fluid flows for airplane bodies, jet engines, and the like. Several RANS models are being used in the wind energy industry: Fluent, CFX, Star-CCM+, OpenFOAM, Meteodyn WT, WindSim, Ventos, etc. The critical difference between RANS and Jackson–Hunt models is that the former solve the nonlinear Navier–Stokes momentum equations, but none of them includes the full conservation of energy equation. The RANS models assume steady-state flows. LES are a promising alternative to RANS models in that they explicitly resolve the energy-containing eddies—those larger than the grid spacing—while simulating the effects of smaller turbulent eddies through a subgrid scale parameterization scheme. LES models are mainly used as a research tool since the necessary computing power is huge. LES can include the full suite of physics parameterization schemes: radiation, microphysics, land surface–atmosphere interaction, turbulence, etc. LES models are based on the raw equations of motion, ie, unsteady, nonlinear Navier–Stokes equations. A fundamental difference between LES and RANS models is that LES solve the conservation of energy equation, which allows LES to fully capture the wind circulations forced by thermal gradients, which is an important driver of offshore wind flows. LES are designed to run at very fine spatial resolutions using a grid spacing in the 1–100-m range. To date, LES have been rarely used to determine the offshore wind resource due to their computational expense and the fact that NWP and microscale models tend to perform relatively well when dealing with the average wind speeds across a large area. However, there is a growing interest for LES to capture the unsteady and nonlinear turbine-induced wakes as well as their two-way interactions with the boundary layer (Jimenez et al., 2007; Calaf et al., 2010; Churchfield et al., 2012).

Historically, wake-effect predictions have been based on a handful of engineering computer models, most importantly Park (Katic et al., 1986; Jensen, 1983) and eddy viscosity (EV) (Ainslie, 1988). The Park model implements a simple formula for the size of the wake deficit and its expansion downstream with a single adjustable parameter, the wake decay constant. The EV model solves an axisymmetric form of the Navier–Stokes equations; it therefore qualifies as a simple RANS model. With the development of large-array wind projects, it has been found that the standard Park and EV models tend to underestimate wake losses in offshore wind farms (Brower and Robinson, 2009; Schlez and Neubert, 2009; Barthelmie et al., 2010). This may be in part because the models assume that wind turbines have no effect on the planetary boundary layer (PBL) other than the wakes they directly generate. As a consequence, new codes have been developed to account for two-way PBL–wake interactions such as the Deep-Array Wake Model (DAWM) in Openwind (Brower and Robinson, 2009) and Large Array Wind Farm (LAWF) model in WindFarmer (Schlez and Neubert, 2009). The DAWM is based on the surface-drag-induced internal boundary layer approach which modifies the wind speed profile within the PBL with increasing distance downstream of the front of a turbine array. The EV or Park model is retained for estimating direct wake effects, thus the term hybrid model. The LAWF model (Johnson et al., 2009) is an extension of standard wake models whereby each turbine is treated as a disturbance analogous to a roughness element that influences the free stream flow, resulting in growth of the internal boundary layer. Both approaches are commonly used today and have demonstrated significant improvement over the original models (Beaucage et al., 2012; Brower and Robinson, 2009).

Another relatively new type of engineering model capable of predicting the turbine-induced wakes is the dynamic wake meander (DWM) model, which is a more detailed model of the flow field behind the upstream turbine (Larsen et al., 2012). This method applies a meandering process within an aeroelastic code in order to simulate the incoming flow field at downstream turbines and calculate both energy production and loading. The Fuga model (Ott et al., 2011) is a linearized RANS model that inserts an actuator disk (an idealized model of a wind turbine rotor's effect on the airflow) to simulate the wakes. It bears some similarities to WAsP, including a mixed spectral solver using precalculated look-up tables. Another relatively new commercial RANS model is WindModeller based on Ansys CFX (Montavon et al., 2011). It incorporates a turbulence closure and an actuator disk. LES models have recently been used to study single and multiple turbine-induced wakes (Wu and Porté-Agel, 2011; Calaf et al., 2010; Churchfield et al., 2012; Troldborg et al., 2014; Mirocha et al., 2014). For single turbine-induced wakes, LES models with a wind turbine parameterization using an actuator disk and/or actuator line model can compare favorably to wind tunnel measurements (Wu and Porté-Agel, 2011). LES codes are a promising approach to simulating wakes if a high-performance computing system is available. Both the open-source Simulator for Offshore/Onshore Wind Farm Applications (SOWFA) from the National Renewable Energy Laboratory (NREL) (Churchfield et al., 2012) and the LES implemented in the WRF model (Mirocha et al., 2014), offer opportunities for the academic and industry sectors to collaborate and develop the next generation of turbine-induced wake models.

A benefit to advancing turbine wake models is to enable turbine and wind farm control systems to optimize operations using wake-related intelligence. For example, under some weather conditions it is possible to enhance overall wind farm output by manipulating (yawing) the orientation of upwind turbines so that their wakes are steered to mitigate negative impacts on downstream turbines. This capability is especially feasible with real-time monitoring of boundary-layer conditions (winds, stability, turbulence) at a wind farm.

In summary, there is a growing number of commercial and research wind turbine wake models but validation datasets from operating offshore wind farms remain limited. Further, those datasets collected by private entities are not typically shared. Because turbine wakes can be a significant cause of power production losses, especially in large arrays, ongoing model advances are desired to optimize wind farm performance through improved layouts and turbine control strategies, and through reduced fatigue loads on turbine structures.

More research focus is needed on processes that couple the turbulent atmospheric and oceanic boundary layers across the interface through the exchange of momentum, mass, and heat. Separate atmospheric and wave models are closed systems that rely on simplified boundary conditions, which can lead to inaccurate solutions. Advancements in the coupling of ocean-atmosphere models will result in increasingly realistic representations of the interplay between the different components of the metocean regime.