You are on page 1of 17

METEOROLOGICAL APPLICATIONS

Meteorol. Appl. 21: 803–819 (2014)


Published online 5 August 2014 in Wiley Online Library
(wileyonlinelibrary.com) DOI: 10.1002/met.1472

Review
Current issues in wind energy meteorology
Stefan Emeis*
Institute of Meteorology and Climate Research, Karlsruhe Institute of Technology, Garmisch-Partenkirchen, Germany

ABSTRACT: This review discusses some of the current issues in wind energy meteorology from the viewpoint of a
meteorologist. The focus is on four major subjects: (1) the wind potential, (2) the influence of major terrain inhomogeneities
on this wind potential, (3) diurnal wind variations and (4) the impact of wind turbines and wind parks on the flow. The wind
potential is addressed by describing vertical profiles of wind and turbulence, trying to give profile laws that are valid throughout
the major part of the atmospheric boundary layer and specifying specific offshore turbulence conditions. Flow over forests and
gently sloping hills are presented as two examples for the very broad spectrum of possible terrain inhomogeneities. Diurnal
variations of wind speed turn out to be height dependent. A major diurnal variation, the formation of low-level jets, is discussed
in detail. Finally, wakes of single turbines and of entire wind parks are addressed and an analytical model for park efficiency
and wake is presented. This review gives only a short overview on the spectrum of issues in wind energy meteorology. A few
analytical approaches are presented to explain first-order effects. Detailed investigations of wind and turbulence conditions
usually require the use of non-linear high-resolution flow models.

KEY WORDS wind energy; vertical wind profiles; complex terrain; low-level jets; wind park efficiency; wind park wakes
Received 19 November 2013; Revised 6 June 2014; Accepted 18 June 2014

1. Introduction distribution of the 10 min averaged wind speeds at hub height


and the overall turbulence intensity was sufficient to supply
Energy meteorology has recently become a new branch in
the necessary background information for the siting of single
applied meteorology after the significant development and
turbines and small onshore wind parks. Logarithmic and power
deployment of wind turbines started about 20 years ago. laws that describe the vertical wind profiles in the surface layer
Mankind’s need for energy will persist or even increase for allowed for reliable vertical interpolations and extrapolations
the foreseeable future. A sustainable supply will only be possi- in flat and homogeneous terrain. Two developments have now
ble from renewable energies in the long run (Breton and Moe, changed this situation.
2009). Renewable energies comprise water power, wave and On the one hand, the focus for wind park installation has moved
tidal energy, geothermal energy, biomass, solar energy and, from land to marine sites. Offshore wind parks will deliver a
last but not least, wind energy. Wind energy could potentially larger part of the wind energy in the future, because of the large
contribute a larger share in future energy supply. Presently, available space and the usually higher wind speeds compared
the global primary energy demand is in the order of 20 TW to onshore sites. Therefore, marine boundary-layer meteorology
(EIA, 2013). According to recent calculations, wind energy has become important in energy meteorology as well. Until a
could supply a few tens of TW (Adams and Keith, 2013). The few years ago, experimental data for the marine atmospheric
availability of wind energy and some other renewable energies boundary-layer were available, if any, for only a shallow layer
depend heavily on the weather. Therefore, this review focuses on explored from buoys, ships and oil riggs. A few masts, such as the
some features in atmospheric conditions which permit or limit three German 100 m high Forschungsplattformen In Nord- und
the generation of electricity from wind energy by wind turbines. Ostsee (FINO) masts, have just been erected in the last 10 years
It is presented from the viewpoint of a meteorologist. (Beeken and Neumann, 2008). They are presently delivering
Systematic electricity generation from the wind has been per- long-term information on a deeper layer of the marine boundary
formed for more than 20 years now. In the early years, onshore for the first time.
wind turbines were small with rotor diameters being much On the other hand, the size of turbines has increased in the last
smaller than the vertical extent of the atmospheric surface layer 15 years. The hub height of multi-megawatt (MW) turbines is
which is roughly 80–100 m deep. Onshore surface layer winds often above the atmospheric surface layer and rotor diameters of
are relatively easy to assess, because in situ measurements from >100 m are frequent now. Offshore turbines with diameters of
masts can be made with reasonable effort in order to calculate >160 m and a power of 7 MW have already been designed and
turbine loads and energy yields. The knowledge of the frequency will be deployed in the near future. This leads to much more
complicated interactions between the turbines and the lower
atmosphere. Meteorological features that had been considered as
* Correspondence: S. Emeis, Institute of Meteorology and Climate irrelevant for a long time are now becoming decisive for planning
Research, Karlsruhe Institute of Technology, Kreuzeckbahnstr. 19, and running single, large turbines and increasingly larger wind
82467 Garmisch-Partenkirchen, Germany. E-mail: stefan.emeis@kit.edu parks. In particular, vertical gradients in mean wind speed and

© 2014 Royal Meteorological Society


804 S. Emeis

wind direction as well as in turbulence intensity above the surface 2.1. Unified vertical wind profile law
layer have to be known.
Both developments moreover mean that the relevant wind Modern large wind turbines nearly always reach into the lower
parameters can no longer be fully obtained by masts. New mea- Ekman layer. Therefore, a unified description of the wind profile
surement techniques are required to collect the necessary wind for the entire lower part of the ABL up to several hundreds of
information. This has led to a boom in surface-based remote metres above ground, i.e. the surface layer and the lower Ekman
sensing techniques (see Emeis, 2010a) such as wind lidars. These layer, is desirable for the planning and operation of modern
lidars provide continuous information on vertical wind and tur- large wind turbines. Conflicting assumptions usually made for
bulence profiles with high vertical resolution. Additionally, the description of wind profiles in the surface and the Ekman
modelling techniques have become more important. They help layer are a handicap that prevent easy vertical extrapolation
to assess, for example, spatially inhomogeneous wind fields in from one layer into the other. As a consequence of a vertically
complex terrain or wakes behinds turbines and entire wind parks. constant exchange co-efficient, K M , in the Ekman layer which
This review tries to summarize some important aspects on is necessary for the derivation of an analytical profile law for
atmospheric boundary layers (ABLs), onshore and offshore, with this layer, wind profiles cannot be extended into the surface
respect to wind power generation. The article will focus on layer. On the other hand, the classic logarithmic wind profile
the vertical profiles of wind and turbulence, surface inhomo- cannot be extended from the surface layer into the Ekman
geneities, diurnal wind variations, park efficiency and wakes. layer, due to the underlying assumption of a linearly growing
The review will address features such as vertical profile laws exchange co-efficient K M = u* l with a constant friction velocity,
and turbulence intensity over homogeneous terrain (onshore and u* , a height-dependent mixing length, l = 𝜅z, and von Kármán’s
offshore) in Section 2, modifications to wind and turbulence pro- constant, 𝜅. Therefore, two approaches were tested recently to
files over inhomogeneous terrain (Section 3), instationary phe- overcome this problem. The first idea is to fit the Prandtl and
nomena such as nocturnal low-level jets (LLJs) (Section 4) and Ekman profiles together in such a way that there is a smooth
the complex wind–wakes interactions in and behind larger wind transition in terms of wind speed and wind shear between both
parks (Section 5). A more detailed description of meteorologi- regimes (Etling, 2002; Emeis et al., 2007). The second idea is
cal aspects relevant for wind energy generation can be found in to modify the Prandtl layer mixing length formulation in order
Emeis (2012). Measurement techniques are described in detail to extrapolate the Prandtl layer wind profile into higher layers
in Emeis (2010a, 2011). A full review of modelling issues espe- (Gryning et al., 2007; Peña et al., 2010a, 2010b).
cially for wind energy does not exist. The available models
range from simple analytical models (Emeis, 2010b) via vari- 2.1.1. Fitting Prandtl and Ekman layer profiles
ous mesoscale models (Fitch et al., 2012) to highly resolving Etling (2002) had proposed the first idea by presenting a
large-eddy simulations (LES) (Steinfeld et al., 2010; Witha et al., wind profile description with a linearly increasing exchange
2012). co-efficient K M below the Prandtl layer height, zp and a constant
K M above this height. Introducing the correction function for
non-neutral stratification, Ψm (z/L* ) (see, e.g. Stull, 1988, for
2. Wind and turbulence profiles over homogeneous terrain details) leads to:

Wind speed information is the most important input for any ⎧


wind energy assessment. Mean wind speed distributions and ⎪ ( ( ) ( ))
⎪ u∗ ∕𝜅 ln z∕z0 − Ψm z∕L∗ for z < zp
turbulence as a function of height have to be known in order to ⎪
predict potential yields from wind turbines and to estimate the ⎪
loads on these turbines. Wind speed profile laws for the horizon- ⎪
⎪ ( )
tally homogeneous ABL form the basis for vertical interpolation u (z) = ⎨ ug − sin 𝛼0 + cos 𝛼0 for z = zp
and/or extrapolation of wind speed data from measurements ⎪
or model layer heights to hub height or other heights within ⎪ √
⎪ ug [1 − 2 2e−𝛾 (z−zp )
the rotor plane of a wind turbine. Profile laws also indicate the ⎪ ( ( ) )
vertical wind shear that has to be expected across the rotor plane. ⎪ sin 𝛼0 cos 𝛾 z − zp + 𝜋∕4 − 𝛼0 for z > zp
The growing hub heights of modern wind turbines require a ⎪ +2e −2𝛾 ( z−z p ) sin 𝛼0 ]
2 1∕2

careful investigation of the vertical structure of the boundary (1)
layer in order to describe the wind profiles correctly. Over land, where u is the mean wind speed, z0 is the roughness length, L* is
hub heights of 80 m and more are usually above the surface layer the Monin–Obukhov length, ug is the geostrophic wind speed, 𝛼 0
(or Prandtl layer) which forms about one tenth of the depth of the is the angle between the surface wind and the geostrophic wind
total boundary layer. Over the sea, the surface layer height can and √
be even lower. Driving pressure gradient forces and retarding f
frictional forces are usually balanced in the surface layer. Simple 𝛾= (2)
2𝜅u∗ z
power law or logarithmic profiles (the first is empirical, the latter
is based on the balance of pressure gradient and frictional forces) is an inverse length scale for the Ekman layer depending on the
are strictly valid in the surface layer only. The remaining 90% of Coriolis parameter f which is only latitude dependent.
the ABL is called the Ekman layer where the Coriolis force due The vertical wind profile given by Equation (1) depends on
to the Earth’s rotation adds to the balance of pressure gradient six parameters: the surface roughness, z0 , the geostrophic wind
and frictional forces. The Coriolis force induces a turning of the speed, ug , the height of the Prandtl layer, zp , the friction velocity,
wind direction with height in the Ekman layer while the vertical u* , the Obukhov length, L* and the angle between the surface
wind shear is much reduced in this layer. Details of the Prandtl wind and the geostrophic wind 𝛼 0 . The three variables z0 , L*
layer and Ekman layer wind profiles are available in classic and ug are external parameters imposed by outer conditions, the
textbooks such as Stull (1988). other three are internal parameters of the boundary layer. If a

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 805

fixed value is chosen for zp then two further equations are needed Turbulence in the ABL is either generated by shear or by thermal
to determine u* and 𝛼 0 . Extending the idea of Etling (2002), a instability. While for lower wind speeds thermal production of
unified vertical wind profile should be generated from the more turbulence is dominant, it becomes nearly negligible for high
realistic physical requirement such that both the wind speed as wind speeds when compared to shear production. The shear
well as the wind shear are continuous at the height z = zp (Emeis production is proportional to the surface roughness.
et al., 2007). This can be achieved by the system of equations Turbulence intensity is defined as the ratio of the standard devi-
given in the Appendix. ation of the wind speed, 𝜎 u over the mean wind speed, u(z). Of
note is that 𝜎 u increases with the length of the averaging interval
2.1.2. Modifying mixing length description (see Appendix A in Emeis, 2012). For a change in the averaging
interval from 10 to 30 min the increase is negligible a few metres
The second idea is to modify the dependence of the mixing length
above ground, because the turbulence length scale is small close
on height and has been proposed by Gryning et al. (2007). They
to the ground and enough turbulence elements are already con-
reformulated the height-dependent mixing length, l (which is
tained in a 10 min interval. The increase becomes larger further
denoted as kz = LL in the Prandtl layer in the following equations,
away from the surface where the turbulence length scale is larger
where k = 0.4 is the von Kármán constant) in order to limit
and less turbulence elements pass the measurement instrument
its growth with height and thus to extend the validity of the
in a given time period. In Appendix A of Emeis (2012), a sample
logarithmic law to heights above the surface layer. The following
from the FINO1 data 80 m above the sea surface showed that 𝜎 u
has been chosen:
nearly doubled when the averaging period was prolonged from
1 1 1 1 10 to 30 min. Using the logarithmic wind profile law, yields the
= + + (3)
l LL LM LU below equation for the turbulence intensity (Wieringa, 1973):
A modified mixing length is formed in Equation (3) by intro- 𝜎u 𝜎 ( ( ))−1 ( ( ))−1
ducing two additional length scales, one for the middle part of Iu (z) = = u 𝜅 ln z∕z0 = A ln z∕z0 (9)
u (z) u∗
the boundary layer, LM = u* /f (−2 ln(u* /(fz0 )) + 55)−1 and one
for the upper part of the boundary layer, LU = (zi – z). Here, zi is Wieringa (1973) set A = 1 based on the assumption that the
the depth of the entire boundary layer. ratio of the standard deviation of the wind speed over the friction
Peña et al. (2010a) suggest a similar approach for the mixing velocity, u* , is 1/𝜅 = 2.5 (Stull, 1988; Arya, 1995). Principally,
length starting from Blackadar’s (1962) principal approach for A should vary with the length of the averaging interval, and
the mixing length, l: the FINO1 data showed values for the ratio 𝜎 u /u* which were
close to 2.5 for wind speeds above 7 m s−1 (Türk, 2008) using an
1 1 (𝜅z)d−1 averaging interval of 10 min. Thus, A is close to unity in this data
= + (4)
l 𝜅z 𝜂d set as well.
where 𝜂 is a limiting value for the mixing length in the upper part Equation (9) means that turbulence intensity in the neutrally
of the boundary layer in the order of 40 m and d is an empirical stratified surface layer is a function of surface roughness only.
factor close to or slightly above unity (Peña et al., 2010a). The Increasing roughness lengths will lead to higher turbulence
alternative approaches by Gryning et al. (2007) and Peña et al. intensities. For a given roughness length, turbulence intensity
(2010a, 2010b) yield the following unified vertical wind profiles: decreases with height in the surface layer. Because it is assumed
( ( ) ) that the standard deviation, 𝜎 u is vertically constant in the surface
u∗ z z z z z layer also for stable stratification (Arya, 1995), the turbulence
u (z) = ln + T + − (5)
𝜅 z0 L∗ LM zi 2LM intensity for stable stratification is given by:
( ( ) ( ))−1
( ( ) ( )d Iu (z) = ln z∕z0 − Ψm z∕L∗ (10)
u z z 1 𝜅z
u (z) = ∗ ln +T +
𝜅 z0 L∗ d 𝜂 The situation is more complicated for unstable stratification.
( )d ) Here 𝜎 u varies with boundary layer height, zi (Panofsky et al.,
1 z 𝜅z z 1977; Arya, 1995) and the turbulence intensity reads:
− − (6)
1 + d zi 𝜂 zi ( )
z 1∕3 ( ( ) ( ))−1
Iu (z) = 15.625 − 0.5 i 𝜅 ln z∕z0 − Ψm z∕L∗
where zi is the boundary layer height and T(z/L* ) is a stability L∗
correction function that is different for unstable and stable strat- (11)
ifications. In unstable situations it reads (Peña et al., 2010b): As turbulence is mainly shear-induced, there is an interesting
( ) ( ) similarity between the vertical profiles of turbulence intensity
T z∕L∗ = −Ψm z∕L∗ (7) and the exponent of the power law for the vertical wind profile.
Equating the slopes of the logarithmic wind profile and that of
and in stable situations (Peña et al., 2010b):
( ) the power law at a given height, z, delivers a relation between
( ) ( ) z the power law exponent, a, and the surface roughness length
T z∕L∗ = −Ψm z∕L∗ 1− (8)
2zi (Sedefian, 1980; Emeis, 2012):
The work on such unified profiles is ongoing; see, for example, ( ( ))−1
a = a (z) = ln z∕z0 (12)
Sathe et al. (2011).
Comparison with the definition of the turbulence intensity
2.2. Turbulence intensity and power law profiles (Equation (9)) reveals that the vertical profile for the exponent
a is equal to the profile of the turbulence intensity for neutral
Turbulence influences the yields from wind turbines as well as stratification. This means that a logarithmic wind profile and a
the loads on wind turbines. Increasing turbulence leads to higher power law profile have the same slope at a given height if the
yields and to higher loads, where the latter is more important. power law exponent equals the turbulence intensity at this height.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


806 S. Emeis

Figure 1. Vertical profiles of turbulence intensity, Iu (z) from Equations (9) to (11) (thick lines) and of the power law exponent, a, from Equations (12)
to (14) (thin lines). Full line: neutral stratification, dashed lines: stable stratification, dash-dotted lines: unstable stratification. (a) Onshore conditions
(z0 = 0.1 m), (b) offshore conditions (z0 = 0.0001 m), (c) offshore conditions (z0 = 0.0001 m), but Iu profiles only compared to FINO1 data from
October 2004 to January 2005 (lines with markers). (Please see note on limited validity with height in the text).

Therefore, the curves for neutral conditions from Equations (9) while the power law exponent drastically increases with height
and (12) are indistinguishable in Figures 1(a) and (b). for stable stratification in a layer within which the Obukhov
For non-neutrally stratified flow, turbulence intensity does similarity theory is valid (Figure 1). This property makes it
not depend only on shear, so the same equality as between difficult to use the power law with a single value for the exponent
Equations (9) and (12) for neutral flow cannot be expected. The for vertical extrapolation over larger height distances.
exponent a is now given by (Sedefian, 1980; Emeis, 2012): It should be noted that although all curves have been plotted up
( ( ) ( ))−1 to 250 m above ground in Figure 1 these curves are strictly valid
z z 1 z only up to those heights where the Obukhov similarity theory
a = a (z) = ln −Ψ for < 0 (13)
z0 L∗ x L∗ is applicable. This height of applicability can easily be 250 m
( ( ) ( ))−1 ( ) in strongly unstable convective boundary layers while it can be
a = a(z) = ln
z
+ 4.7
z
1 + 4.7
z
for
z
>0 much <100 m in stable situations.
z0 L∗ L∗ L∗
(14) 2.3. Turbulence intensity as function of wind speed
where x = (1 − bz/L* )1/4 and b = 16. Equations (13) and (14) show
that the exponent a is smaller with unstable stratification than According to Equations (9)–(11) turbulence intensity does not
with neutral, but is larger with stable stratification, because x depend directly on wind speed. Onshore, this is not true for
and the expression within the brackets containing z/L* are both all wind speeds. Turbulence intensities increase with decreas-
larger than unity. The change in turbulence intensity is opposite ing wind speeds as can be seen from Figure 2(a). In Equations
with larger turbulence intensity in unstable regimes and smaller (9)–(11) it is assumed that the standard deviation of the wind
turbulence intensity in stable regimes (Figure 1). speed, 𝜎 u is directly proportional to the wind speed, u. This is
The second important message from Equations (9)–(14) is true for fully developed mechanically induced turbulence, but
that turbulence intensity and power law exponent decrease with when approaching zero wind speed, the regime of fully devel-
height in the surface layer with neutral and unstable stratification oped turbulence ceases and 𝜎 u decreases more slowly than the

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 807

Figure 2. Turbulence intensity as function of wind speed. (a) Onshore. Functional relationship for Class B turbulence intensity as given in IEC
61400-3. (b) Offshore. Evaluation of data at 90 m height at the FINO1 mast in the German Bight from September 2003 until August 2007 using
1 m s−1 bins. Please note the different scaling of the y-axis in both plots. The right-hand y-axis in plot (b) refers to the thin curve with the crosses
which indicates how many values are in each 1 m s−1 bin.

wind speed, leading to higher turbulence intensities. Figure 2(a) surface temperatures significantly above the air temperature.
shows the assumptions for turbulence intensity made for onshore When the wind speed (and so the roughness length) increases
conditions in the standard IEC 61400-3 (2005). further, the mechanical part of the turbulence intensity begins
Offshore this is not true either. On the one hand, the same to dominate over the thermal effects and turbulence intensity
asymptotic behaviour for low wind speeds is found as over land, increases again (Barthelmie, 1999).
leading to higher turbulence intensities (Figure 2(b)). But on The absolute minimum of turbulence intensity for each 1 m s−1
the other hand there is a second influence which is not present wind speed bin in Figure 2(b) is <1% for wind speeds of up to
onshore: the ocean roughness length increases with increasing 20 m s−1 . Inspection of the synoptic conditions suggests that
wind speed due to the formation of waves (Foreman and Emeis, under very stable atmospheric conditions, situations with very
2010, 2012; Türk and Emeis, 2010). The dependence of median, low turbulence intensity can occur offshore even at relatively
arithmetic mean, minimum, maximum, the 10th , 25th , 75th and high wind speeds. The influence of the increasingly rough sur-
90th percentiles of turbulence intensity on wind speed, 10 min face for wind speeds above 20 m s−1 can break the stable layering
means for the measuring period from September 2003 to August and so the absolute minimum of turbulence intensity begins to
2007 and a measuring height of 90 m at the offshore mast increase. At higher wind speeds, the spread of turbulence inten-
FINO1, are shown in Figure 2(b). For low wind speeds, the sity values within one wind speed class continuously becomes
mean of turbulence intensity rapidly decreases with increasing smaller (Large and Pond, 1981).
wind speed to a minimum value of about 4.5% at 12 m s−1 For the calculation of loads on wind turbines the 90th percentile
wind speed. Above this minimum, turbulence intensity increases of the turbulence intensity for a given wind speed bin is impor-
nearly linearly with increasing wind speed. The high turbulence tant. The international standard IEC 61400-3 (2005) suggests the
intensity values at wind speeds below about 12 m s−1 originated following dependence on wind speed:
from the dominance of thermal induced turbulence at low u ( )
𝜎u90 = ( h ) + 1.28 1.44 m s−1 I15 (15)
wind speeds during unstable atmospheric conditions with water ln zh ∕z0

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


808 S. Emeis

Figure 3. 90th percentile of turbulence intensity as function of wind speed. (a) Data at two heights (30 and 90 m) at FINO1 compared to the functional
dependence given in the IEC 61400-3 (2005). (b) Data at 90 m at FINO1 compared to the modified relation given in Equation (15).

where uh is the wind speed at hub height of the wind turbines, zh is calculated according to IEC 61400-3 (2005). Above wind speeds
the height of the hub above sea level and I 15 is the average turbu- of about 22 m s−1 , the slopes of measured and calculated curves
lence intensity at hub height at 15 m s−1 wind speed. Figure 3(a) of the 90th percentile of the turbulence intensity become nearly
shows the 90th percentiles of measured offshore turbulence inten- identical.
sity at the FINO1 mast at different heights from 10 min means for At 70 and 90 m of the FINO1 mast, the values according to the
the period September 2003–August 2007 (solid lines) compared IEC standard (2005) lie permanently above the measured values,
to the offshore turbulence intensity given by IEC 61400-3 (2005, while at the heights of 50 and 30 m, the measured values lie above
dashed lines). Similar to the turbulence intensity (Figure 2(b)) the the IEC-values for some wind speed bins. The data presented
values of the 90th percentiles of turbulence intensity also decrease here suggest a modification of the relation (15) given in the IEC
with increasing wind speed until a minimum of ∼7–8.5% at standard. A better fit is possible from:
10–12 m s−1 wind speed and then increase again with further-
more increasing wind speed; 90th percentiles of turbulence inten- uh 2uIu,min ( )
𝜎u90 = a ( )+ 1.44 m s−1 I15 + buh (16)
sity also decrease with height (Figure 3(a)). Compared to tur- ln zh ∕z0 uh
bulence intensities given by the IEC standard, three sectors are
detected. Below wind speeds of 8–10 m s−1 , at wind speeds that where a and b are two tunable factors and uIu, min is the wind
are not really load-relevant for wind turbines, measured 90th speed at which the minimum turbulence intensity occurs. The
percentiles of the turbulence intensities are not well covered first term on the right-hand side of Equation (15) and apart
by the IEC-curves. At wind speeds between 10 and 22 m s−1 from the factor a also in Equation (16) gives the mean standard
the 90th percentiles of measured turbulence intensities lie below deviation of the wind speed for thermally neutral stratification,
the values given by the IEC standard except for two values at assuming a logarithmic vertical wind profile (see Equation (9)).
30 m height. In this sector, the slopes of the curves of measured For the second term it is assumed in Equation (15) that the values
turbulence intensity are steeper than the slopes of the curves for the standard deviation of the wind speed follow a Gaussian

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 809

distribution around its mean, so that the 90th percentile of the 3.1.1. Near-equilibrium flow over large forests
wind speed, 𝜎 u90 is 1.28 times the standard deviation of the
Flow equilibrium is usually expected to be found 20–30 tree
standard deviation of the wind speed, the latter represented by
heights away from the edges of a large forest (Träumner et al.,
1.44 m s−1 times I 15 . In Equation (16) the constant factor 1.28 has
2012). Wind and turbulence profiles in such an equilibrium for-
been replaced by twice the ratio between the wind speed where
est boundary layer decisively depend on the spacing of the trees.
minimum turbulence intensity occurs and the hub height by the
If trees grow very close together, the flow is displaced vertically
wind speed. In Figure 3(b) this better fit (Equation (16)) is shown
where a = 0.63 and b = 0.0012 have been used. The variability of above the forest. The crowns of the forest trees form a new rough
the two empirical factors a and b for different offshore sites has surface which has much in common with an impervious rough
not been investigated yet. grass land (Raupach, 1979). The vertical displacement, d of the
flow is considered by a large displacement height on the order
of two thirds of the canopy height. This height substitutes the
Earth’s real surface in all profile laws for flows over forests, i.e.
3. Wind and turbulence profiles over inhomogeneous z is to be replaced by z–d in the Equations (1), (5) and (6). If the
terrain trees grow sparser, the displacement height is lower and the rough
The features described in the previous Section are found when surface has to be considered as pervious. Therefore, the main
the atmospheric flow is in equilibrium with a horizontally homo- difference between a densely vegetated forest and a sparsely veg-
geneous surface. This is an idealized description, because usually etated forest is that in a sparsely vegetated forest larger air parcels
land surfaces are not horizontally homogeneous and increasingly can enter (these movements are sometimes called sweeps) and
more onshore wind turbines are erected in complex terrains. Two leave (also called ejections) the forest canopy sublayer. This per-
major inhomogeneities are different land use characteristics and meability of the rough surface of the forest canopy sublayer leads
orography, which usually occur together. Here, flow over forests to anomalously higher turbulence intensities in the lower part of
in flat terrain and simple orographic features with a homogeneous the forest sublayer than expected from the mean vertical wind
surface are addressed as two simple examples. gradient in this layer (see Högström et al., 1989 for details). This
anomalous enhanced turbulence may extend to about three tree
heights (see big arrow in the central part of Figure 4).
3.1. Wind and turbulence profiles over forests
Therefore, wind turbines at forest sites should have hub heights
In recent years, the deployment of wind turbines in forests of more than three times the tree height in order to avoid unneces-
has become an interesting option as these sites are usually sary fatigue due to enhanced turbulence. Together with the above
away from larger settlements and are no longer visible for those mentioned large displacement height on the order of two thirds
inhabitants who may be disturbed by their presence. In terms of the tree height, this usually means wind turbine hub heights
of boundary-layer meteorology, forest-covered surfaces are a above the mean tree height of considerably >100 m, i.e. total
special form of vegetated surfaces. A distinction must be made hub heights of ∼130–150 m above ground. This means addi-
between equilibrium flow features over larger forest canopies tional constructive efforts to build these turbines. On the other
away from their edges and special flow features at the edges itself. hand, mean wind speeds are typically higher at greater heights

Figure 4. Schematic of flow over a forest with edge effects. Long curves with arrows symbolize streamlines; small curved arrows indicate additional
turbulence. The large black arrow indicates sweeps and ejections between the canopy layer and the forest sublayer in the near-equilibrium flow over
larger forests. H is average tree height.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


810 S. Emeis

above ground so that such turbines can still be economically turbulence is observed up to about 15 tree heights downwind
meaningful. of the rear forest edge (Träumner et al., 2012). A comparison
between the mentioned results from wind lidar measurements,
3.1.2. Flow features and forest edges wind tunnel data and the output from a high-resolution numerical
model study with a large-eddy simulation (LES) is given in
Turbines close to forest edges are subject to special flow features Kanani et al. (2014).
that must be expected at these sites (Figure 4). At the upwind
edge, the vertical displacement of the flow leads to the formation 3.2. Wind and turbulence over orography
of a jet-like feature and additional small-scale turbulence above
the crowns of the trees at three to five tree heights. In a data set The influence of mountains on the flow is usually much more
from wind lidar measurements this jet phenomenon disappears important than spatial changes in surface properties. Principally,
when the flow is in a new equilibrium after about 30 tree mountains influence the upcoming wind for turbines in two dif-
heights behind the forest edge (Shannak et al., 2011; Träumner ferent ways. There is a dynamic influence on flows crossing
et al., 2012). At the lee edge, the flow starts to deviate about mountains (Figure 5(a)) and there is a thermal influence produc-
20 tree heights before the forest edge (Träumner et al., 2012). ing secondary flow features (Figure 5(b)). The thermally induced
Once again, increased wind speeds and additional small-scale flows can be very regular, especially with fine weather and clear

Figure 5. (a) Schematic of wind profile distortion in a flow over a hill crest (full line) compared to the shape of a wind profile over flat terrain (dashed
line). The dotted line following the hill contour indicates the depth of the inner layer. The maximum wind speed-up occurs at the top of this inner
layer. (b) Thermally induced secondary flows in mountainous terrain. Thin arrows: slope winds and vertical winds over crests, full arrows: valley
and mountain winds, open arrows: regional winds towards and away from the mountains.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 811

skies, but wind speeds are rarely so high that they can be used for
wind energy generation. Therefore, only the dynamic influence
is addressed in this review. The mountain-induced flow features
can be very complex and non-linear. Non-linearity such as flow
separation cannot adequately be described with analytical mod-
els but must be addressed with non-linear numerical flow models;
see, for example, Zeman and Jensen (1987).
The most prominent feature of flow over hills and mountains
is the speed-up over the crest (Figure 5(a)). This (fractional)
speed-up, Δs is defined as the ratio of the wind speed at a
given height over a crest, u, to the wind speed at the same
height above ground over level terrain, u∞ minus unity. For
attached flow over gentle slopes (H/L > 0.2, where H is the height
and L is the half-width of the hill) with no flow separation, a
simple analytical analysis is sufficient. Some of the analytical
approaches are quite old and date back to work of, for example,
Jackson and Hunt (1975). These analytical approaches have
always been accompanied by numerical efforts; see, for example,
the work of Taylor (1977). Also, the well-known Wind Atlas
Analysis and Application Program (WAsP) model is based on
such simplifying analytical approaches (Troen and Petersen,
1989). For such flow over gentle hills Hoff (1987) found for
neutral thermal stratification:
u (L) H ( x z )
Figure 6. Stability dependence of the fractional speed-up over a gentle
u (x, z)
Δs (x, z) = −1= ∞ 𝜎 , (17) hill from Equation (18) with L = 1000 m, l = 16.44 m and z0 = 0.2 m.
u∞ (z) u∞ (l) L L L Dashed: unstable stratification (L* = −500 m), full: neutral stratification,
short-dashed: stable stratification (L* = 500 m).
where x is the horizontal co-ordinate (origin at the hill top), z
is the vertical co-ordinate above ground, l is the depth of the should be made with considerably reduced power law exponents
so-called inner layer over the hill (usually a few percent of L, compared to level terrain, and (2) shear-produced turbulence
see Figure 5(a)) and 𝜎 describes the shape of a cross-section may be lower. This means that above the height l the vertical
through the hill. At the hill top 𝜎 has the value of unity and increase in wind speed is lower than that in the same height above
approaches zero far away from the hill. Near the hill foot this flat terrain. Therefore, over hill tops, less additional energy is
function is slightly negative. The absolute value of 𝜎 decreases harvested when turbine hub heights are increased. But loads on
with height, above the hill surface. Usually, with a few exceptions the turbines could be lower due to the reduced shear turbulence.
(see, e.g. Emeis, 2012), 𝜎 cannot be given analytically. Also Δs These changes in the vertical wind profiles over the hill have
increases with increasing steepness, H/L of the hill and increases consequences for the vertical profiles of the two parameters of
with height in the inner layer and reaches its maximum value at the Weibull distribution over the hills. The vertical profile of the
the height l over the hill top. Simultaneously, height l denotes the scale parameter of the Weibull distribution more or less follows
depth of this inner layer. the changes in the mean wind speed profile described above.
Bradley (1983) extended this relation for non-neutral stratifi- The change in the vertical profile of the shape parameter, k, is
cation: ( ) more drastic (Figure 7). The frequently observed maximum of
ln zL − Ψ LL ( ) the shape parameter at the top of the surface layer which is typical
H x z
( ) 𝜎 ,
0 ∗
Δs (x, z) = (18)
ln zl − Ψ Ll L L L
0 ∗

where z0 is the surface roughness length and L* is the


Monin–Obukhov length. The stability correction function
(Ψm (z/L* )) is the same as in Equation (1). Δs increases with
increasing stability and is reduced with unstable flow (Figure 6),
because increasing stability opposes the vertical displacement of
the streamlines over the hill. Thus, the streamlines are squeezed
together and the speed-up is increased.
The most important consequence of the speed-up described in
Equations (17) and (18) is that the vertical wind shear is increased
below the height l and that this shear is decreased above height
l. The height of the inner layer (see Figure 5(a)), l, is usually
1–5% of the horizontal scale, L, of the hill. For typical hills
or similar orographic features with a horizontal scale of a few
kilometres this means that the maximum of the speed-up appears
at heights of several tens of metres above the hill top. The rotor
area of modern large wind turbines will be usually above this Figure 7. Sample vertical profiles of the shape parameter, k(z) over flat
height l. Therefore, reduced vertical shear in the rotor area is to terrain (diamonds) and over a hill top (squares) from Equation (19) with
be taken into account. This has two consequences: (1) vertical zm = 75 m and c2 = 0.06 for flat terrain and zm = 50 m and c2 = 0.01 for
extrapolation of wind speeds to greater heights in this layer the hill top.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


812 S. Emeis

for flat homogeneous terrain disappears nearly completely. The daytime while it appears frequently at night time in the Ekman
shape parameter over the hill tops tends to be more or less layer. This has already been observed by Hellmann (1915) and
constant with height. The reasons for this are on the one hand Peppler (1921) and has been shown from numerical studies by
the reduced shear in wind speed above the top of the inner layer Krishna (1968).
and on the other hand the fact that hill tops are often above the In the marine ABL (MABL) the time scales of non-stationarity
nocturnal stable surface layer. This leads to the fact that over the are different from those over land. Over oceans the diurnal
hills in all heights a diurnal variation of wind speed is found cycle is practically absent due to the high heat capacity of the
which is very similar to the diurnal variation above the surface water. Instead, an annual cycle is found. Here, instationarity
layer over flat terrain, i.e. higher wind speeds at night time than at and LLJs appear when the flow transits from warmer land to
daytime (Emeis, 2001). The profiles in Figure 7 have been plotted cooler sea or passes over water bodies with considerably different
using a formula suggested by Wieringa (1989): temperature.
( )
( ) z − zA
k (z) − kA = c2 z − zA exp − (19) 4.2. Low-level jets
zm − zA
where kA is the measured shape parameter at a height zA , zm When the near-surface cooling after sunset occurs rather fast
is the height of the maximum of the k(z)-profile and c2 is a (i.e. more than about 0.5∘ h−1 which leads to inversion strengths
tunable constant. Figure 7 has been plotted with zm = 75 m and of >1 K per 100 m (Kottmeier et al., 1983)) the Ekman layer
c2 = 0.06 for flat terrain and zm = 50 m and c2 = 0.01 for the hill winds can overshoot to super-geostrophic wind speeds. This
top. Together with zA = 30 m and kA = 2.0 the curves in Figure 7 phenomenon is called the nocturnal LLJ. Nocturnal LLJs usually
are close to values observed in Emeis (2001). A slightly modified form at the top of the nocturnal boundary layer (Lettau, 1954;
version of Equation (19) has very recently been proposed by Blackadar, 1957). First climatologies of this phenomenon were
Gryning et al. (2014) who added a further term that allows k to given by Wippermann (1973) and Kottmeier et al. (1983). The
approach a given value of kt ≈ 2 in the upper part of the boundary climatology given by Bonner (1968) deals with a different type
layer. of LLJ over the Great Plains east of the Rocky Mountain chain
The example presented here is valid for gentle hills with homo- in the USA. There, due to orographic forcing by the Rockies
geneous land use only. Steeper hills and mountains lead to and synoptic conditions which lead to strong pressure gradients
non-linear features such as flow separation, channelling in val- across the Great Plains, mountainous LLJs develop much more
leys or flows horizontally around orographic features. Non-linear frequently (63% of all nights) than the simple nocturnal LLJ
flow features can no longer be derived from simple analytical (see Song et al. (2005) for further details.) Here, the purely
relations but require the operation of full Reynolds-averaged stability related nocturnal LLJ first described by Lettau (1954)
numerical simulation models (RANS models). Therefore, wind and Blackadar (1957) is considered.
assessment in rougher terrain where linearity is no longer assured An LLJ can also form when the flow transits from a warm
has to be done by site-specific numerical model simulations. land surface to a cooler water surface. During this transit, the
lower part of the flow stabilizes quite fast and the wind above this
stable layer overshoots as well (Högström and Smedman, 1984;
Smedman et al., 1993). This coastal phenomenon is especially
4. Diurnal wind variations and low level jets (LLJs) pronounced when the water is cooler than the air (Garratt and
4.1. Daily cycles of mean wind speed in the surface layer and Ryan, 1989). Marine LLJs only disappear gradually when the
above newly formed internal boundary layer over the sea surface finally
fills the full depth of the marine boundary layer. Above the LLJ
Generation of energy from the wind is subject to changing wind the wind speed decreases again to the geostrophic level in both
speeds. Wind speeds change due to the synoptic weather con- cases.
ditions and, in addition, they usually exhibit diurnal variations. Typical heights of LLJ cores are often between 150 and 500 m
These diurnal variations can be important because they may above ground. Therefore, they have the ability to influence
lead to a mismatch between electric energy generation from the the energy yield of modern wind turbines with hub heights of
wind (which may peak at night time) and the common electricity >100 m. Recent case studies based on surface-based remote sens-
demand (which peaks at daytime). The usual diurnal variations ing data can be found, for example, in Banta et al. (2002, 2004,
in the thermal stratification of the ABL over land modify the ver- 2006) or in Reitebuch et al. (2000). Figure 8 shows six subse-
tical wind profiles given by stationary wind laws (1), (5) or (6). quent half-hourly mean profiles from SODAR measurements as
Non-stationarity, especially over land in the transition hours at an example. Here, the jet core is at 300–350 m above ground.
dusk and dawn, provokes additional wind features that are not The following statements on typical properties of LLJs are
covered by these stationary wind laws. The most prominent non from 2 years of acoustic remote sensing with a SODAR (sound
stationary feature in the wind field over land is the change in wind detection and ranging) in Northern Germany during the years
speed at dusk. In daytime, when radiative heating of the surface 2001–2003 and serve here as an example. The frequency of
destabilizes the thermal stratification of the ABL, additional tur- occurrence is linked to the appearance of certain weather or
bulence is produced. This enhanced vertical exchange leads to circulation types, because the development of a nocturnal LLJ
increasing wind speed in the surface layer and decreasing wind requires clear skies and a non-vanishing large-scale horizontal
speed in the Ekman layer. After sunset, radiative cooling of the pressure gradient. For Central Europe, the ‘Grosswetterlagen’
surface stabilizes the thermal stratification of the ABL and drasti- (large-scale weather types) have proven to give a good clas-
cally reduces turbulence and vertical momentum exchange. This sification of the weather situation (Gerstengarbe et al., 1999).
leads to low wind speeds close to the ground in the shallow stable Figure 9 shows the frequency of occurrence of LLJs over North-
nocturnal surface layer and to higher wind speed in the Ekman ern Germany as a function of these 29 large-scale weather types.
layer above (LLJs, see next subsection). The result is that in the The three-most relevant types during the 2 years of observations
surface layer the wind speed maximum usually appears during (the three left-most columns in Figure 9) were a high-pressure

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 813

core wind speeds. This means that the mean wind shear under-
neath the jet core varies mainly between 0.04 and 0.08 s−1 .
Typical hub heights of modern onshore wind turbines will be
well above 100 m and upper rotor blade tips will reach heights
in the order of 200 m. Therefore, Figure 10(b) shows the vari-
ation of the wind speed at 160 m above ground as a function
of the 850 hPa wind speed at midnight for nights with LLJs.
The latter is used as an approximation to the geostrophic wind
speed and has been taken from radiosonde data at Bergen, Ger-
many (WMO number 10238, data have been obtained from
the internet from the pages of the University of Wyoming
at http://weather.uwyo.edu/upperair/sounding.html). The wind
speed at 160 m during LLJ events varies between ∼5 and 13 m s−1
without showing a strong correlation with the 850 hPa wind
speed. This indicates that the 160 m wind speed is limited by
some additional factor. This factor could be the maximum shear
Figure 8. Nocturnal low-level jets from SODAR measurements at
which can prevail underneath the LLJ core without producing too
Charles de Gaulle Airport in Paris in June 2005. Each profile shows
much shear-induced turbulence which eventually would lead to a
half-hourly mean data. Grey curves below 200 m give logarithmic wind
profiles for neutral and stable conditions which fit the measured wind breakdown of the LLJ. Wind speeds of about 18 m s−1 seem to be
speed at 50 m height. an upper limit for the formation of LLJs. It turns out that there is
a most favourable 850 hPa wind speed range for LLJ formation
at 9–13 m s−1 . More details can be found in Emeis (2014).
bridge over Central Europe (type ‘BM’), a high-pressure system
over Central Europe (type ‘HM’) and a high-pressure area over
the British Isles (type ‘HB’). Overall, a LLJ appeared in about 5. Wakes
21.3% of all nights. Nearly the same frequency (roughly 20%) 5.1. Single wakes
has been found for the Cabauw tower in the Netherlands (Baas
et al., 2009). Assuming a typical duration of about 6 h for a noc- Wind parks need special treatment, because the flow conditions
turnal LLJ means that a LLJ occurs in about 5% of all hours of a approaching most of the turbines in the park interior are no
year (i.e. in about 440 h of the 8760 h of a year). longer undisturbed. Wakes produced by upwind turbines can
The specific features of a single LLJ event vary quite consid- significantly influence downwind turbines. This means reduced
erably. Figure 10 shows a few more results from the 2 years of wind speeds and enhanced levels of turbulence which will lead
SODAR measurements in Hannover. The maximum wind speed to reduced yields and enhanced loads. For a given land or sea
varies between 7 and 23 m s−1 in the LLJ cases observed dur- area, it is desirable to optimize the number of wind turbines in
ing this period (Figure 10(a)). The height of the jet core can be order to maximize energy production. If wind turbines are too
between 130 and 600 m, the latter being the height up to which closely spaced, wake interference effects could result in a con-
the SODAR data are usually available. There is a visible correla- siderable reduction in the efficiency of the wind park’s energy
tion between the core wind speed and the core height (R2 is 0.31 production up to some tens of percent. Some wind parks with
in this data set). The core height increases from about 200 m for tightly spaced turbines have produced substantially less energy
the lowest LLJ core wind speeds to 400–600 m for the highest than expected based on wind resource assessments. In some

Figure 9. Frequency of occurrence of nocturnal low-level jets (in percent) for all nights of a year from SODAR measurements at Hannover in Northern
Germany as a function of large-scale weather types (Grosswetterlagen).

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


814 S. Emeis

Figure 10. (a) Low-level jet (LLJ) core height in m versus maximum jet wind speed in m s−1 from SODAR measurements at Hannover in northern
Germany. Lines indicate constant mean wind shear underneath the jet core of 0.02, 0.05 and 0.1 s−1 (from left to right). (b) Maximum 160 m wind
speed in m s−1 during a LLJ event versus 850 hPa wind speed in m s−1 from SODAR measurements at Hannover and the Bergen radiosonde in
Northern Germany. A wind speed of 4 m s−1 at 160 m height corresponds to a mean shear underneath this height of 0.025 s−1 ; a wind speed of
16 m s−1 corresponds to a mean shear of 0.1 s−1 .

densely packed parks where turbines have failed prematurely, B varies between 1 and 3 while n takes values between 0.75 and
it has been suspected that these failures might have been 1.25 and principally depends on the ambient turbulence intensity
caused by excessive turbulence associated with wake effects (Vermeer et al., 2003). The WAsP model (Troen and Petersen,
(Elliot, 1991). 1989) uses a similar approach (Barthelmie and Jensen, 2010):
Distinction has to be made between near wakes and far wakes ( )( )2
when considering turbine wakes. The near wake is taken as the uh √ D
= 1 − 1 − Ct (22)
area just behind the rotor, where the special properties of the uh0 D + ks
rotor itself can still be distinguished, approximately up to a few
rotor diameters downstream. Features such as 3D vortices and with the wake decay co-efficient k where k = 0.04 is typical for
tip vortices from single blades in the near wake are found. The offshore conditions (Barthelmie and Jensen, 2010) while 0.075
far wake is the region beyond the near wake, where modelling is the default (onshore) value in WAsP (Barthelmie et al., 2004).
the actual rotor is less important. The distance-dependent relative The added turbulence intensity in the wake, ΔI, decreases more
velocity deficit, Δu, in the far wake reads according to Vermeer slowly than the velocity deficit. Vermeer et al. (2003) give three
et al. (2003): empirical formulae from three different sources which describe
u − uh ( )n the measured data quite well. According to Quarton (1989), the
Δu D
= h0 =B (20) added turbulence intensity decreases as:
uh uh s
where uh0 is the undisturbed wind speed upwind of the turbine, √ ( s )0.57
2 = 4.8C 0.7 I 0.68 N
uh is the wind speed at hub height, D is the rotor diameter, s is the ΔI = I 2 − I∞ T ∞ (23)
s
distance from the turbine and B and n are constants. B depends
on the turbine thrust co-efficient, Ct , and increases with it. The where I ∞ is the undisturbed turbulence intensity and sN is the
thrust co-efficient has a constant value of about 0.85 for lower length of the near wake which is between one and three rotor
wind speeds around the cut-in wind speed and decreases around diameters. The width of the wake, DW , is proportional to the
and above the rated wind speed of the turbines with increasing one third power of the distance from the turbine, s (see Frandsen
wind speed (Jimenez et al., 2007). The exact value depends on et al., 2006 for more details):
the construction of the turbine and its operation. The following
empirical relation for the thrust co-efficient is taken from Figure DW (s) ∝ s1∕3 (24)
9 in Magnusson (1999) and additionally the maximal value at
This spreading of the wake with distance downstream of the
Betz’s limit is considered (see Emeis, 2012 for details):
( ( ( )) ) turbine leads unavoidably to complex wake–wake interactions
CT = min max 0.25; 0.5 + 0.05 14 − uh ; 0.89 (21) in larger wind parks.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 815

Figure 11. Normalized wind speed reduction (a) and wind power reduction (b) at hub height, h within large wind parks from Equation (28) as function
atmospheric stability (h/L* = −1: unstable, = 0: neutral, = +1: stable) and surface roughness (z0 = 0.0001 m: smooth sea surface, z0 = 1.0: very rough
land surface).

5.2. Wind park models flow above the wind park:

In principle, two different approaches for modelling the effects 𝜏 u − uh


= Km 0 (26)
of large wind parks are possible: a bottom-up approach and a 𝜌 Δz
top-down approach. The bottom-up approach is based on a super- The turbulent exchange co-efficient K m describes the ability
position of the different wakes of the turbines in a wind park. It of the atmosphere to transfer momentum vertically by turbu-
requires a good representation of each single wake (see Section lent motion. Following Frandsen (2007), the wind park drag
5.1) in a 3D flow model (Lissaman, 1979; Jensen, 1983) and a co-efficient ct is defined as a function of the park area A, the rotor
wake combination model. Reviews are available in Crespo et al. area 0.25𝜋D2 , the number of turbines N and the turbine thrust
(1999) and Vermeer et al. (2003). Numerically, this approach is co-efficient CT :
supported by LES (Jimenez et al., 2007; Wussow et al., 2007; 1 N𝜋D2
Steinfeld et al., 2010; Troldborg et al., 2010). The top-down ct = CT (27)
8 A
approach considers the wind park as a whole as an additional
surface roughness, as an additional momentum sink or as a grav- Finally, the ratio Rt between the wind speed at hub height inside
ity wave generator in association with a temperature inversion the wind park to the undisturbed wind speed upstream reads as
aloft at the top of the boundary layer (for the latter idea see below (the entire derivation is documented in Emeis, 2012):
Smith, 2010), which modifies the mean flow above it (New- ( )
𝜙
man, 1977; Bossanyi et al., 1980; Frandsen, 1992). An analyt- fh,Δz Iu + 𝜅m2 cs,h
ical model which follows the top-down approach and which is Rt = ( ) (28)
𝜙
described in detail in Emeis (2012) is shortly introduced. fh,Δz Iu + 𝜅m2 cteff
The starting point for the analytical wind park model is the
overall mass-specific momentum consumption m of the turbines Relation (28) permits easily to add the turbulence intensity pro-
which is proportional to the drag of the turbines, ct and the wind duced by the turbines during operation to the upstream turbulence
2
speed, uh at hub height h: intensity (Iu,eff = Iu2 + Iu,t
2 ). Following Barthelmie et al. (2003),
0
the additional turbulence I u,t can be parameterized as a function
m = ct u2h (25) of the thrust co-efficient (Equation (21)) and a mean turbine dis-
tance 𝛿: √
In an indefinitely large wind park, this momentum loss can only 1.2CT
be accomplished by a turbulent momentum flux 𝜏 from above. Iu,t = (29)
𝛿 2∕D2
Here, u0 is the undisturbed wind speed above the wind park,
K m is the momentum exchange co-efficient and Δz is the height Figure 11(a), in displaying Rt from Equation (28), shows how
difference between hub height of the turbines and the undisturbed much the wind speed at hub height will be reduced as a function

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


816 S. Emeis

of the atmospheric instability and the surface roughness. The time domain. It can be converted into the space domain by assum-
presented results have been found for turbines with a hub height ing an average wind speed over the wake.
of 92 m, a rotor diameter of 90 m and a mean distance between Figure 12(a) shows wake lengths as a function of surface
two turbines of 10 rotor diameters. The figure demonstrates that roughness for neutral stability (h/L* = 0) by plotting the third
the reduction is smallest (a few percent) for unstable thermal power of Rn from Equation (28). If the distance necessary for
stratification of the ABL and high surface roughness. That is, a recovery of the available power to 95% of its undisturbed value
the reduction is smallest over a rough land surface with trees upstream of the park is defined as wake length, then a wake
and other obstacles for cold air flowing over a warm surface length of 4 km for rough land surfaces and a wake length of
(usually during daytime with strong solar insolation). The largest about 18 km for smooth sea surfaces are found. Figures 12(a)
reduction (up to 45%) occurs for very smooth sea surfaces and (b) have been produced for the same park parameters as
when warm air flows over cold waters. This may happen most Figures 11(a) and (b). Actually, the results from Figure 11
commonly in springtime. Figure 11(b) translates this wind speed serve as left boundary conditions for Figure 12. Figure 12(b)
reduction into a reduction of the available wind power by plotting demonstrates the strong influence of atmospheric stability on
the third power of Rt from Equation (28). The strong stability the wake length for an offshore wind park over a smooth sea
dependence of the reduction of the available power can be surface (z0 = 0.0001 m). Taking once again the 95% criterion,
confirmed from measurements taken at the Nysted wind park in the wake length for very unstable atmospheric conditions is
Denmark (Barthelmie et al., 2007). about 10 km. For very stable conditions, the wake length is
The dependence of wind and available power reduction as >30 km. Such long wakes have been confirmed, for example,
a function of surface roughness has consequences for offshore from satellite observations (Christiansen and Hasager, 2005) and
wind parks which are likely to become the major facilities for from numerical simulations (Fitch et al., 2012).
wind power generation in the near future (Breton and Moe,
2009). The lower turbulence production due to the relative
smoothness of the sea surface compared to land surfaces ham-
6. Conclusions
pers the momentum re-supply from the undisturbed flow above.
In order to limit the wind speed reduction at hub height in the This review has shown several relevant meteorological features
interior of the wind park to values known from onshore parks, which must be considered when planning and building wind
the turbines within an offshore wind park must have a larger turbines and farms. The most important are:
spacing than within an onshore park. The above presented analyt-
ical equation (Equation (28)) shows that, roughly speaking, the 1. For hub heights above ∼80 m pure logarithmic or power
number of turbines per unit area in an offshore park with rough- law wind speed profiles, valid for the surface layer, are not
ness z0 = 0.001 m must be approximately 40% lower than in an sufficient to assess the true wind conditions. Suggestions for
onshore park with z0 = 0.1 m in order to have the same power unified wind profile laws which are simultaneously valid in
yield for a given hub height wind speed and atmospheric stabil- the surface layer and in the Ekman layer are presented in
ity. For a given geostrophic wind speed aloft, the lower efficiency Section 2.1.
of offshore parks is overcompensated by the higher hub height 2. Turbulence intensity and the power law exponent are
wind speed under offshore conditions. Inversely, Equation (28) closely related to each other and both of them are height
may be used to determine the optimal areal density of turbines in dependent. Therefore, vertical wind speed extrapolations
a large wind park for given surface roughness and atmospheric over larger vertical ranges (>10–20 m) using the power law
stability conditions. should be made only with the height-dependent exponents
given in Section 2.2.
3. Turbulence intensity at a given height is wind speed
5.3. Wind park wakes dependent. Onshore, turbulence intensity decreases and
The estimation of the length of the wakes of large wind parks approaches asymptotically a constant value with increas-
is essential for the planning of the necessary distance between ing wind speed. Offshore, turbulence intensity decreases
adjacent wind parks. This estimation can be made using the towards a minimum (around 10 m s−1 wind speed at
same principal idea as in the previous subsection: the missing FINO1) and then increases again with increasing wind
momentum in the wake of an indefinitely broad wind park can speeds (see Section 2.3).
only be replenished from above. To move with an air parcel, 4. Flow over forests show distinct features. Especially, turbu-
the acceleration of the speed of this parcel uhn from uhn0 at lence intensity is modified compared to flat terrain (Section
the rear end of the park to the original undisturbed value, uh0 , 3.1).
which had prevailed upstream of the park (neglecting the Coriolis 5. Vertical wind profiles and thus also the vertical profiles
force) is found. This leads to a differential equation (for details of the parameters of the Weibull distribution over com-
of the derivation see Emeis, 2012) which can be solved by an plex terrain are different from those over flat terrain (see
exponential approach giving the ratio Rn between the wind speed Section 3.2). Wind assessment in real complex terrain
at hub height in the wake and the undisturbed wind speed at the requires the application of non-linear numerical flow mod-
same height uh0 : els (Reynolds-averaged numerical simulation (RANS) or
large-eddy simulations (LES) models, also called CFD
( )
u (t) uhn0 models).
Rn = hn =1+ − 1 exp (−𝛼t) (30) 6. The daily wind speed maximum above the surface layer is
uh0 uh0
usually during night time whilst the surface layer maximum
The factor 𝛼 is given by 𝜅u* z/Δz2 (where Δz is the height dif- is during daytime (Section 4.1).
ference between the hub height and the height of the undisturbed 7. Nocturnal low-level jets (LLJs) are a frequent phenomenon
wind above) and depends on the surface roughness and the ther- and appear in about 21% of all nights over the lowlands
mal stratification of the boundary layer. This solution is in the of Central Europe. During these events typical vertical

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 817

Figure 12. Normalized reduction of available wind power at hub height, h behind an indefinitely large wind park as a function of the distance from
the rear edge of the park in km from Equation (30). (a) As function of surface roughness (z0 = 0.0001 m: very smooth sea surface, 0.001 m: rough
sea surface, 0.1 m: smooth land surface, 1.0 m: rough land surface) with neutral stability. (b) As function of atmospheric instability (h/L* = 1: strong
instability, 0: neutral stability, +1: stable stratification) for a smooth sea surface. Note the different scales of the y-axis.

wind shear across the rotor plane of large wind turbines park wakes increases with increasing atmospheric stability
is between 3 and 10 m s−1 per 100 m height interval (see and decreasing surface roughness (Section 5.3).
Section 4.2).
8. LLJs in the marine boundary layer frequently appear when
the flow transits from a warmer land surface over a cooler 7. Outlook
water surface (Section 4.2).
9. Flow in wind parks is characterized by wake–wake interac- Growing turbine sizes and the erection of turbines and entire
tion. Different approaches have been designed to describe wind parks offshore and in complex terrain have changed the
this interaction. The top-down approach presented here importance of meteorological influences on turbine siting and
leads to an analytical park model. The efficiency of a wind resource assessment in recent years. In order to cope
wind park increase with increasing surface roughness and with this increasing complexity in future, remote sensing tech-
decreasing atmospheric stability (Section 5.2). niques (such as wind lidars) and non-linear high-resolution
10. The basic message from the analytical wind park model numerical models (CFD models) must become the standard
is that the generation of energy from the wind over larger tools in assessing wind resources. Meteorological expertise,
areas is limited by the re-supply of momentum from above. which partly is already available, has to be revisited and eval-
This re-supply purely depends on the turbulent state of the uated for this purpose. This expertise is definitely needed for
atmosphere. Thus, atmospheric physics limits the amount of the operation of these complex observation and modelling
energy which can be extracted from the wind. It is exactly tools.
this limitation which has been neglected in earlier assess- As a consequence, a new field within the discipline of mete-
ment of the global potential for wind energy (e.g. Grubb and orology has formed in recent years: energy meteorology. This
Meyer, 1993; Archer and Jacobson, 2005; Lu et al., 2009). sub-discipline of meteorology deals with all meteorological
The study by Adams and Keith (2013) mentioned in the aspects relevant to generation of renewable energies and the
introduction is one of the first which takes this limit into operation of electrical grids. It is expected that energy meteo-
account. rology will become more visible in the near future. Wind energy
11. The wakes of entire wind parks can be modelled in an meteorology as reviewed in this paper is only part of this new
extension of the wake–wake formulation. The length of sub-discipline.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


818 S. Emeis

Appendix Barthelmie R, Frandsen ST, Rethore PE, Jensen L. 2007. Analysis of


atmospheric impacts on the development of wind turbine wakes at
Equating the first two equations of the wind profile equation the Nysted wind farm. Proceedings of the European Offshore Wind
(Equation (1)) for z = zp gives the first equation for the friction Conference, 4–6 December 2007, Berlin.
velocity: Barthelmie RJ, Jensen LE. 2010. Evaluation of wind farm efficiency and
( ) wind turbine wakes at the Nysted offshore wind farm. Wind Energy
𝜅ug − sin 𝛼0 + cos 𝛼0 13: 573–586.
u∗ = ( ) ( ) (A.1)
ln zp ∕z0 − Ψm zp ∕L∗ Barthelmie RJ, Pryor S, Frandsen S, Larsen S. 2004. Analytical
modelling of large wind farm clusters. Poster session presented at
Equating the respective equations for the vertical wind shear at Proceedings of the EAWE Conference 2004, Delft. http://www.
the same height z = zp leads to a second equation for u* : risoe.dk/vea/storpark/Papers%20and%20posters/delft_013.pdf
(accessed 7 July 2014).
| |
2 |ug | 𝛾𝜅zp sin 𝛼0 Beeken A, Neumann T. 2008. Five years of offshore measurements at
u∗ = | |( ) (A.2) the Fino1 platform in the German Bight. DEWI Mag. 33: 6–11 http://
𝜑 zp ∕L∗ www.dewi.de/dewi/fileadmin/pdf/publications/Magazin_33/02.pdf
(accessed 7 July 2014).
where 𝜑m (z/L* ) is the vertical derivative of Ψm (z/L* ). For unsta- Blackadar A. 1957. Boundary layer wind maxima and their significance
ble situations, the differential form 𝜑 of the correction function for the growth of nocturnal inversions. Bull. Am. Meteorol. Soc. 38:
283–290.
Ψm for unstable thermal stratification reads: Blackadar A. 1962. The vertical distribution of wind and turbulent
( ) ( )−1∕4 exchange in a neutral atmosphere. J. Geophys. Res. 67: 3095–3102.
𝜑 z∕L∗ = 1 + b z∕L∗ (A.3) Bonner WD. 1968. Climatology of the low level Jet. Mon. Weather Rev.
96: 833–850.
In stable situations, the differential form 𝜑 of the correction Bossanyi EA, Maclean C, Whittle GE, Dunn PD, Lipman NH, Musgrove
function Ψm for stable stratification reads: PJ. 1980. The efficiency of wind turbine clusters. Proceedings of
( ) the Third International Symposium on Wind Energy Systems, 26–29
𝜑 z∕L∗ = 1 + a z∕L∗ (A.4) August 1980, Lyngby; 401–416.
Bradley EF. 1983. The influence of thermal stability and angle of
where b is chosen between 15 and 16 and a is chosen to be about incidence on the acceleration of wind up a slope. J. Wind Eng. Ind.
4.7–5. Aerodyn. 15: 231–242.
The Equations (A.1) and (A.2) must be valid simultaneously. Breton S-P, Moe G. 2009. Status, plans and technologies for offshore
Equating the right hand sides of these two equations yields the wind turbines in Europe and North America. Renew. Energy 34:
desired relation for the turning angle, 𝛼 0 : 646–654.
Christiansen MB, Hasager CB. 2005. Wake effects of large offshore
1 wind farms identified from satellite SAR. Remote Sens. Environ. 98:
𝛼0 = arctg 2𝛾zp ( ( ) ( )) (A.5) 251–268.
1+ 𝜑(zp ∕L∗ )
ln zp ∕z0 − Ψm zp ∕L∗ Crespo A, Hernandez J, Frandsen S. 1999. Survey of modelling methods
for wind turbine wakes and wind farms. Wind Energy 2: 1–24.
Unfortunately, Equation (A.5) still depends on the friction EIA. 2013. International Energy Outlook 2013. U.S. Energy Informa-
velocity u* via the definition of 𝛾 in Equation (2). Thus, the fric- tion Administration. http://www.eia.gov/forecasts/ieo/?src=home-b2
(accessed 18 March 2014).
tion velocity u* has to be determined iteratively starting with Elliot DL. 1991. Status of wake and array loss research. Report
a first guess for u* in Equation (A.1), subsequently comput- PNL-SA--19978, Pacific Northwest Laboratory: Richland, WA,
ing 𝛼 0 from Equation (A.5), and then re-computing u* from September 1991; 17. http://www.osti.gov/scitech/servlets/purl/6211
Equation (A.1) or (A.2). 976 (accessed 7 July 2014).
Emeis S. 2001. Vertical variation of frequency distributions of wind
Inversely the system of Equations (A.1), (A.2) and (A.5) can be
speed in and above the surface layer observed by sodar. Meteorol. Z.
used to determine the height of the Prandtl layer, zp if the friction 10: 141–149.
velocity, u* is known from other sources. Emeis S. 2010a. Measurement Methods in Atmospheric Sciences. In
Situ and Remote, Quantifying the Environment, Vol. 1. Borntraeger:
Stuttgart; 257.
References Emeis S. 2010b. A simple analytical wind park model considering
atmospheric stability. Wind Energy 13: 459–469.
Adams AS, Keith DW. 2013. Are global wind power resource estimates Emeis S. 2011. Surface-Based Remote Sensing of the Atmospheric
overstated? Environ. Res. Lett. 8: 015021. Boundary Layer, Atmospheric and Oceanographic Sciences Library,
Archer CL, Jacobson MZ. 2005. Evaluation of global wind power. J. Vol. 40. Springer: Heidelberg; 174.
Geophys. Res. 110: D12110. Emeis S. 2012. Wind Energy Meteorology – Atmospheric Physics for
Arya SP. 1995. Atmospheric boundary layer and its parameterization. In Wind Power Generation, Green Energy and Technology. Springer:
Wind Climate in Cities, Cermak JE, Davenport AG, Plate EJ, Viegas Heidelberg; 196.
DX (eds). Kluwer: Dordrecht; 41–66. Emeis S. 2014. Wind speed and shear associated with low-level jets
Baas P, Bosveld FC, Klein Baltink H. 2009. A climatology of nocturnal over Northern Germany. Meteorol. Z. 23, pre-published online,
low-level jets at Cabauw. J. Appl. Meteorol. Climatol. 48: 1627–1642. DOI: 10.1127/0941-2948/2014/0551.
Banta RM, Darby LS, Fast JD, Pinto JO, Whiteman CD, Shaw WJ, et al. Emeis S, Baumann-Stanzer K, Piringer M, Kallistratova M, Kouznetsov
2004. Nocturnal low-level jet in a mountain basin complex. Part I: R, Yushkov V. 2007. Wind and turbulence in the urban boundary
Evolution and implications to other flow features. J. Appl. Meteorol. layer – analysis from acoustic remote sensing data and fit to analytical
43: 1348–1365. relations. Meteorol. Z. 16: 393–406.
Banta RM, Newsom RK, Lundquist JK, Pichugina YL, Coulter RL, Etling D. 2002. Theoretische Meteorologie. Eine Einführung. Springer:
Mahrt L. 2002. Nocturnal low-level jet characteristics over Kansas Heidelberg; 354.
during CASES-99. Bound.-Lay. Meteorol. 105: 221–252. Fitch AC, Olson JB, Lundquist JK, Dudhia J, Gupta AK, Michalakes J,
Banta RM, Pichugina YL, Brewer WA. 2006. Turbulent et al. 2012. Local and mesoscale impacts of wind farms as parameter-
velocity-variance profiles in the stable boundary layer generated ized in a mesoscale NWP model. Mon. Weather Rev. 140: 3017–3038.
by a nocturnal low-level jet. J. Atmos. Sci. 63: 2700–2719. Foreman R, Emeis S. 2010. Revisiting the definition of the drag coef-
Barthelmie RJ. 1999. Monitoring offshore wind and turbulence charac- ficient in the marine atmospheric boundary layer. J. Phys. Oceanogr.
teristics in Denmark. Proceedings of the BWEA Wind Energy Confer- 40: 2325–2332.
ence, Cambridge. Foreman R, Emeis S. 2012. Correlation equation for the marine drag
Barthelmie RJ, Folkerts L, Ormel FT, Sanderhoff P, Eecen PJ, Stobbe O, coefficient and wave steepness. Ocean Dyn. 62: 1323–1333.
et al. 2003. Offshore wind turbine wakes measured by sodar. J. Atmos. Frandsen S. 1992. On the wind speed reduction in the center of large
Oceanic Technol. 20: 466–477. cluster of wind turbines. J. Wind Eng. Ind. Aerodyn. 39: 251–265.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)


Current issues in wind energy meteorology 819

Frandsen S. 2007. Turbulence and turbulence generated structural load- Peña A, Gryning S-E, Hasager C. 2010a. Comparing mixing-length
ing in wind turbine clusters. Risø-R-1188(EN), Risø National Labo- models of the diabatic wind profile over homogeneous terrain. Theor.
ratory: Roskilde; 130. Appl. Climatol. 100: 325–335.
Frandsen ST, Barthelmie RJ, Pryor SC, Rathmann O, Larsen S, Højstrup Peña A, Gryning S-E, Mann J, Hasager CB. 2010b. Length scales of
J, et al. 2006. Analytical modelling of wind speed deficit in large the neutral wind profile over homogeneous terrain. J. Appl. Meteorol.
offshore wind farms. Wind Energy 9: 39–53. Climatol. 49: 792–806.
Garratt JR, Ryan BF. 1989. The structure of stably stratified internal Peppler A. 1921. Windmessungen auf dem Eilveser Funkenturm. Beitr.
boundary layer in offshore flow over the sea. Bound.-Lay. Meteorol. Phys. Atmos. 9: 114–129.
47: 17–40. Quarton DC. 1989. Characterization of wind turbine wake turbulence
Gerstengarbe F-W, Werner PC, Rüge U (eds). 1999 http://www. and its implications on wind farm spacing. Final Report ETSU WN
deutscher-wetterdienst.de/lexikon/download.php?file=Gross 5096, Department of Energy of the UK, Garrad-Hassan Contract.
wetterlage.pdf or http://www.pik-potsdam.de/∼uwerner/gwl/gwl.pdf. Raupach MR. 1979. Anomalies in flux-gradient relationships over forest.
Katalog der Großwetterlagen Europas (1881 - 1998). Nach Paul Hess Bound.-Lay. Meteorol. 16: 467–486.
und Helmuth Brezowsky, 5th edn. German Meteorological Service: Reitebuch O, Straßburger A, Emeis S, Kuttler W. 2000. Noctur-
Potsdam/Offenbach a. M (accessed 7 July 2014). nal secondary ozone concentration maxima analysed by SODAR
Grubb MJ, Meyer NI. 1993. Wind Energy: Resources, Systems, and observations and surface measurements. Atmos. Environ. 34:
Regional Strategies. In Renewable Energy: Sources for Fuels and 4315–4329.
Electricity, Johansson TB, Burnham L (eds). Island: Washington, DC; Sathe A, Gryning S-E, Peña A. 2011. Comparison of the atmospheric
157–212. stability and wind profiles at two wind farm sites over a long marine
Gryning S-E, Batchvarova E, Brümmer B, Jørgensen H, Larsen S. 2007. fetch in the North Sea. Wind Energy 14: 767–780.
On the extension of the wind profile over homogeneous terrain beyond Sedefian L. 1980. On the vertical extrapolation of mean wind power
the surface layer. Bound.-Lay. Meteorol. 124: 251–268. density. J. Appl. Meteorol. 19: 488–493.
Gryning S-E, Batchvarova E, Floors R, Peña A, Brümmer B, Hahmann Shannak B, Träumner K, Wieser A, Corsmeier U, Kottmeier C. 2011.
AN, et al. 2014. Long-term profiles of wind and Weibull distribution Flow characteristics above a forest using light detection and rang-
parameters up to 600 m in a rural coastal and an inland suburban area. ing measurement data. J. Mech. Eng. Sci. 226: 921–939, DOI:
Bound.-Lay. Meteorol. 150: 167–184. 10.1177/0954406211417944.
Hellmann G. 1915. Über die Bewegung der Luft in den untersten Smedman A, Tjernström M, Högström U. 1993. Analysis of the tur-
Schichten der Atmosphäre. Meteorol. Z. 32: 1–16. bulent structure of a marine low-level jet. Bound.-Lay. Meteorol. 66:
Hoff AM. 1987. Ein analytisches Verfahren zur Bestimmung der mit- 105–126.
tleren horizontalen Windgeschwindigkeiten über zweidimensionalen Smith RB. 2010. Gravity wave effects on wind farm efficiency. Wind
Hügeln, Vol. 28. Ber. Inst. Meteorol. Klimatol. Univ. Hannover: Han- Energy, 13, 449–458.
nover; 68. Smith SD. 1980. Wind stress and heat flux over the ocean in gale force
Högström U, Bergström H, Smedman A-S, Halldin S, Lindroth A. winds. J. Phys. Oceanogr. 10: 709–726.
1989. Turbulent exchange above a pine forest, I: Fluxes and gradients. Song J, Liao K, Coulter RL, Lesht BM. 2005. Climatology of the
Bound.-Lay. Meteorol. 49: 197–217. low-level jet at the southern great plains atmospheric boundary layer
Högström U, Smedman A. 1984. The wind regime in coastal areas with experiments site. J. Appl. Meteorol. 44: 1593–1606.
special reference to results obtained from the Swedish Wind Energy Steinfeld G, Tambke J, Peinke J, Heinemann D. 2010. Application of
Program. Bound.-Lay. Meteorol. 33: 351–373. a large-eddy simulation model to the analysis of flow conditions in
International Electrotechnical Commission. 2005. IEC 61400-3, Ed. 1 offshore wind farms. Geophys. Res. Abstr. 12: EGU2010-8320 http://
(Draft): Wind Turbines – Part 3: Design Requirements for Offshore meetingorganizer.copernicus.org/EGU2010/EGU2010-8320.pdf
Wind Turbines. International Electrotechnical Commission: Geneva. (accessed 7 July 2014).
Jackson PS, Hunt JCR. 1975. Turbulent wind flow over a low hill. Q. J. Stull R. 1988. An Introduction to Boundary-Layer Meteorology. Kluwer
R. Meteorol. Soc. 101: 929–955. Academic Publishers: Dordrecht; 666.
Jensen NO. 1983. A note on wind generator interaction. Risø-M-2411, Taylor PA. 1977. Numerical studies of neutrally stratified planetary
Risø National Laboratory: Roskilde; 16. http://www.risoe.dk/rispubl/ boundary layer flows over gentle topography. I: Two-dimensional
VEA/veapdf/ris-m-2411.pdf (accessed 7 July 2014). cases. Bound.-Lay. Meteorol. 12: 37–60.
Jimenez A, Crespo A, Migoya E, Garcia J. 2007. Advances in large-eddy Träumner K, Wieser A, Ruck B, Frank C, Röhner L, Kottmeier C. 2012.
simulation of a wind turbine wake. J. Phys. Conf. Ser. 75: 012041, The suitability of Doppler lidar for characterizing the wind field above
DOI: 10.1088/1742-6596/75/1/012041. forest edges. Forestry 85: 399–412, DOI: 10.1093/forestry/cps038.
Kanani F, Träumner K, Ruck B, Raasch S. 2014. What determines the Troen I, Petersen EL. 1989. European Wind Atlas. Risø National Labo-
differences found in forest edge flow between physical models and ratory: Roskilde; 656.
atmospheric measurements? An LES study. Meteorol. Z. 23: 33–49. Troldborg N, Sørensen JN, Mikkelsen R. 2010. Numerical simulations of
DOI: 10.1127/0941-2948/2014/0542. wake characteristics of a wind turbine in uniform inflow. Wind Energy
Kottmeier C, Lege D, Roth R. 1983. Ein Beitrag zur Klimatologie 13: 86–99.
und Synoptik der Grenzschicht-Strahlströme über der norddeutschen Türk M. 2008. Ermittlung designrelevanter Belastungsparameter für
Tiefebene. Ann. Meteorol. N.F. 20: 18–19. Offshore-Windkraftanlagen. PhD thesis, University of Cologne; 124.
Krishna K. 1968. A numerical study of the diurnal variation of meteoro- http://kups.ub.uni-koeln.de/2799/ (accessed 7 July 2014).
logical parameters in the planetary boundary layer. I. Diurnal variation Türk M, Emeis S. 2010. The dependence of offshore turbulence intensity
of winds. Mon. Weather Rev. 96: 269–276. on wind speed. J. Wind Eng. Ind. Aerodyn. 98: 466–471.
Large WG, Pond S. 1981. Open ocean momentum flux measurements in Vermeer LJ, Sørensen JN, Crespo A. 2003. Wind turbine wake aerody-
moderate to strong winds. J. Phys. Oceanogr. 11: 324–336. namics. Prog. Aerospace Sci. 39: 467–510.
Lettau H. 1954. Graphs and illustrations of diverse atmospheric states Wieringa J. 1973. Gust factors over open water and built-up country.
and processes observed during the seventh test period of the great Bound.-Lay. Meteorol. 3: 424–441.
plains turbulence field program. Occasional Report 1, Atmospheric Wieringa J. 1989. Shapes of annual frequency distributions of wind speed
Analysis Laboratory, Air Force Cambridge Research Center: Bedford, observed on high meteorological masts. Bound.-Lay. Meteorol. 47:
MA. 85–110.
Lissaman PBS. 1979. Energy effectiveness of arbitrary arrays of wind Wippermann F. 1973. Numerical study on the effects controlling the
turbines. AIAA Paper 79-0114. low-level Jet. Beitr. Phys. Atmos. 46: 137–154.
Lu X, McElroy MB, Kiviluoma J. 2009. Global potential for Witha B, Steinfeld G, Heinemann D, Stütz E. 2012. High-resolution
wind-generated electricity. Proc. Natl. Acad. Sci. U. S. A. 106: offshore wake simulations with the LES model PALM. Geophys.
10933–10938. Res. Abstr. 14: EGU2012-10387 http://meetingorganizer.copernicus.
Magnusson M. 1999. Near-wake behaviour of wind turbines. J. Wind org/EGU2012/EGU2012-10387.pdf (accessed 7 July 2014).
Eng. Ind. Aerodyn. 80: 147–167. Wussow S, Sitzki L, Hahm T. 2007. 3D-simulations of the turbulent
Newman BG. 1977. The spacing of wind turbines in large arrays. J. wake behind a wind turbine. J. Phys. Conf. Ser. 75: 012033, DOI:
Energy Convers. 16: 169–171. 10.1088/1742-6596/75/1/012033.
Panofsky HA, Tennekes H, Lenschow DH, Wyngaard JC. 1977. The Zeman O, Jensen NO. 1987. Modification of turbulence characteristics
characteristics of turbulent velocity components in the surface layer in flow over hills. Q. J. R. Meteorol. Soc. 113: 55–80.
under convective conditions. Bound.-Lay. Meteorol. 11: 355–361.

© 2014 Royal Meteorological Society Meteorol. Appl. 21: 803–819 (2014)

You might also like