WES Wind Energy Science WESWind Energ. Sci. 2366-7451 Copernicus Publications Göttingen, Germany 10.5194/wes-3-533-2018From standard wind measurements to spectral characterization: turbulence length scale and distributionTurbulence length scale statistics – from standard observations to spectra KellyMark mkel@dtu.dk https://orcid.org/0000-0003-2882-4450 Wind Energy Department, Risø Lab./Campus, Danish Technical University, Roskilde 4000, Denmark Mark Kelly (mkel@dtu.dk)16August2018 3 2 533543 11February2018 14February2018 3July2018 24July2018 This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/4.0/This article is available from https://wes.copernicus.org/articles/3/533/2018/wes-3-533-2018.htmlThe full text article is available as a PDF file from https://wes.copernicus.org/articles/3/533/2018/wes-3-533-2018.pdf

In wind energy, the effect of turbulence upon turbines is typically simulated using wind “input” time series based on turbulence spectra. The velocity components' spectra are characterized by the amplitude of turbulent fluctuations, as well as the length scale corresponding to the dominant eddies. Following the IEC standard, turbine load calculations commonly involve use of the Mann spectral-tensor model to generate time series of the turbulent three-dimensional velocity field. In practice, this spectral-tensor model is employed by adjusting its three parameters: the dominant turbulence length scale LMM (peak length scale of an undistorted isotropic velocity spectrum), the rate of dissipation of turbulent kinetic energy ε, and the turbulent eddy-lifetime (anisotropy) parameter Γ. Deviation from “ideal” neutral sheared turbulence – i.e., for non-zero heat flux and/or heights above the surface layer – is, in effect, captured by setting these parameters according to observations.

Previously, site-specific {LMM,ε,Γ} values were obtainable through fits to measured three-dimensional velocity component spectra recorded with sample rates resolving the inertial range of turbulence (≳1 Hz); however, this is not feasible in most industrial wind energy projects, which lack multi-dimensional sonic anemometers and employ loggers that record measurements averaged over intervals of minutes. Here a form is derived for the shear dependence implied by the eddy-lifetime prescription within the Mann spectral-tensor model, which leads to derivation of useful forms of the turbulence length scale. Subsequently it is shown how LMM can be calculated from commonly measured site-specific atmospheric parameters, namely mean wind shear (dU/dz) and standard deviation of streamwise fluctuations (σu). The derived LMM can be obtained from standard (10 min average) cup anemometer measurements, in contrast with an earlier form based on friction velocity.

The new form is tested across several different conditions and sites, and it is found to be more robust and accurate than estimates relying on friction velocity observations. Assumptions behind the derivations are also tested, giving new insight into rapid-distortion theory and eddy-lifetime modeling – and application – within the atmospheric boundary layer. The work herein further shows that distributions of turbulence length scale, obtained using the new form with typical measurements, compare well with distributions P(LMM) obtained by fitting to spectra from research-grade sonic anemometer measurements for the various flow regimes and sites analyzed. The new form is thus motivated by and amenable to site-specific probabilistic loads characterization.

 Introduction

Of the atmospheric parameters which are generally input into (or required by) wind turbine load calculation codes, several stand out due to their prominence in load contributions: the “mean” wind speed U, the standard deviation of streamwise turbulent velocity σu, the shear dU/dz or shear exponent α, and the characteristic turbulence length scale L corresponding to the most energetic turbulent motions e.g.,. explored the importance of shear (α); found that both fatigue and extreme turbine loads can be sensitive to L in addition to the dominant influences of mean wind speed U and streamwise turbulence “strength” σu

To a lesser extent, some sensitivity to the Mann-model anisotropy parameter Γ has also been found.

. These are also consistent with the earlier finding of that stability could affect fatigue loads through α and σu.

Within the context of obtaining site-dependent statistics of the most crucial load-driving parameters (U,σu,α,L) from conventional industrial wind measurements, this work focuses on the one parameter which has thus far been most difficult to measure: the turbulence length scale L. The turbulence length scale corresponds to the “energy-containing sub-range” of turbulent velocity fluctuations associated with the peak of the streamwise velocity spectrum, which contribute most to turbulent kinetic energy (and σu) – and which can dominate the turbulence contribution to wind turbine loads. Measurements used in wind energy are usually stored as 10 min statistics (average and standard deviation of wind speed and direction), so one cannot obtain turbulence spectra from them, nor can one calculate integral time or length scales from such observations.

Because of its widespread use in the wind industry and its inclusion in the standard on design requirements for wind turbines, here we consider the spectral turbulence model of and L as prescribed for this model. Within the “Mann model”, which uses rapid-distortion theory (RDT) to account for shear-induced distortion of isotropic turbulence e.g.,, there is also a prescription for the scale-dependent time over which turbulent eddies of a given size are distorted. This timescale is key to proper representation of atmospheric turbulence and reproduction of component spectra via RDT. However, the eddy lifetime was not directly derived, but rather cleverly prescribed, by . Concurrent to and independent of the work herein, also derived some relations to create a model for time-varying eddy lifetime. The present article provides direct derivation of the eddy lifetime, which results in a relation between the three (spectral) parameters of the Mann model and measurable quantities. More importantly, the derivations here include connection of the turbulence length scale to routinely available quantities from typical 10 min industrial wind records. The turbulence length scale is in fact that corresponding to the spectral form, and thus the relation here is applicable to other turbulence models used in wind engineering, such as those relying on the spectrum.

After deriving the eddy lifetime and giving subsequent expressions for the turbulence length scale, this article proceeds to validation of the underlying assumptions. Constraints implied by fitting the Mann model to measured spectra in non-neutral conditions, given eddy lifetime and mixing-length relations, are also tested. This includes dependence of predicted velocity variance on model anisotropy parameter (Γ), as well as implications in the surface layer and connection to previous findings in boundary-layer meteorology. Finally, the length scale obtained from conventional 10 min wind measurements via the new expression is compared to the length scale found from fits of Mann-model output to measured component spectra; this is done using data from multiple sites, representing several types of site conditions.

 Theory

Relation of the turbulence length (spectral “peak”) scale to measurable statistics is possible through the eddy-lifetime form of , where the latter is defined in terms of the isotropic von Kármán spectrum that is distorted using RDT.

Eddy lifetime

A number of forms exist to estimate eddy lifetime τe, though these can be generally expressed as the ratio of a length scale (taken as the reciprocal of wavenumber, k-1) to a velocity scale which follows from some integrated form of the (scalar) kinetic energy spectrum E(k): τe∼k-p-1∫k∞κ-2pE(κ)dκ-1/2, where the characteristic velocity scale can be generically described by kp∫k∞κ-2pE(κ)dκ1/2. In contrast to the “coherence-destroying diffusion time” of and the reciprocal of eddy-damping rates from , for use with rapid-distortion theory chose an eddy lifetime that depends on eddy size (wavenumber) according to τM∝k-1∫k∞E(κ)dκ-1/2, i.e., equivalent to p=0 in terms of Eq. (). The choice of Eq. () for eddy lifetime was found to behave more reasonably than both the “diffusion time” (where p=1)

The expression is also equivalent (or at least proportional) to the “convection time” of .

, as well as the timescale [k3E(k)]-1/2 (which in the inertial range is equivalent to p=-1)

The reciprocal of eddy-damping rate, [k3E(k)]-1/2, is equal in the inertial range to Eq. () with p=-1 since E(κ)→κ-5/3 there. This expression is also similar to the “rotation time” or “strain time” given by , but it should be noted that such expressions integrate from 0 to k, i.e., over eddies larger than 1/k.

 implicit in eddy-damped quasi-normal Markovian models ; both of the latter lifetime models do not (reliably) integrate to give finite σu2.

re-writes τM as τM(k)=ΓdU/dz(kLMM)-2/32F113,176;43;-1(kLMM)2, where 2F1 is Gauss' hypergeometric function 

The hypergeometric function 2F113,176;43;-(kLMM)-2 approaches 1 for kLMM≫1 (the inertial range) and simplifies to aHG(kLMM)2/3 for kLMM≪1, where aHG≡(3π/4)fΓ(4/3)/fΓ(17/6)≃0.69 and fΓ(x) is the Euler gamma function.

, and LMM is the turbulence length scale associated with the peak of the turbulent kinetic energy spectrum E(k) as in Eq. (). The eddy lifetime definition () is used in practical implementation of the spectral tensor model e.g.,, and it notably defines a parameter of this model: the eddy lifetime factor Γ, also known as the anisotropy factor. The spectral-tensor model employs RDT, whereby the shear dU/dz distorts turbulence from an isotropic state, based on an initial turbulent kinetic energy spectrum of the von Kármán form EvK(k)=αε2/3LMM5/3(kLMM)4[1+(kLMM)2]17/6, where α=1.7 . This in effect defines the length scale LMM through the peak of the initial spectrum

The peak of the von Kármán isotropic TKE spectrum EvK(k) occurs at kLMM=12/5, i.e., LMM≃1.55/kpeak.

. Using Eq. () in the proportionality expression () produces τM|E→EvK=cτk-2/332αε22/3F113,176;43;-1(kLMM)2, where we have introduced the proportionality constant cτ to write the result of integrating the proportionality relation () as an equation. Now τM can be seen to depend upon k, LMM, and ε. The eddy lifetime can be reduced and clarified via 2F113,176;43,(-kLMM)-2≃[1+3.07(kLMM)-2]-1/3 to give the more transparent von Kármán-like form

Note 2/3≃0.82 and 3.07=aHG-3; cf. footnote 4. In Eq. (), αε2/3 is kept together for comparison with Eq. () and because αε2/3 is commonly used as an input to the spectral-tensor model instead of ε e.g.,.

 τM(k;LMM,ε)≃0.82cταε2/3k-2/31+3.07(kLMM)21/6.

Since Eqs. () and () are equal, we have an expression relating the Mann-model parameters to the shear dU/dz: Γ=cτ3α/2dUdzLMM2/3ε-1/3. The expression () can be made yet more useful to relate the turbulent length scale to measurable parameters, as shown in Sect. .

Eddy lifetime and equilibrium

The parameters {ε,Γ,LMM} are site-dependent and in practice have been obtained from measurements through fits of the model output to observed spectra , relying on (at least three of) F11, F13, F33, and F22 e.g.,. The model starts with an (undistorted) isotropic incompressible turbulence spectral tensor: Φij(k)|0=δijk2-kikj4πk4E(k), where E(k) is taken to be EvK(k) (shown in Eq. ), then the Φij are distorted – i.e., the rapid-distortion equations are solved – per (three-dimensional) wavenumber over a time τM(k) via RDT.

The rapid-distortion equations do not explicitly solve for production of normal stresses (which sum to twice the turbulent kinetic energy) or shear stress, though they do include terms that perturb the stresses

Assuming a constant mean shear dU/dz, the spectral-tensor model solves Fourier-transformed versions of rapid-distortion equations for streamwise normal stress 〈u1u1〉 and shear stress 〈u1u3〉; multiplying these by dU/dz one obtains the corresponding production rates: P11=-2〈u1u3〉dU/dz and P13=-〈u3u3〉dU/dz chap. 11.

to account for the (anisotropic) effect of a constant shear dU/dz. Further, the RDT discussed here does not include dissipation ; instead, in the spectral-tensor model the dissipation rate of turbulent kinetic energy ε is a parameter giving the amplitude of the undistorted (initial) spectrum via Eq. (). In practice ε is obtained via fits of pre-calculated Mann-model output to measured spectra. So ε in effect gives the inertial-range amplitudes of the distorted velocity component spectra, which have been distorted for a time τM(k). From Eq. () one sees that the parameter Γ serves as a factor that determines the amount of distortion and associated anisotropy: increasing Γ corresponds to longer distortion time τM and thus more anisotropy, with Γ=0 corresponding to isotropy (zero distortion of the initial isotropic Φij). The separation between the peaks of the different component spectra increases with Γ; the spectral peak of F33 is at higher wavenumbers (smaller scales) than the F22 peak, which is at higher wavenumbers than the peak of F11 .

A stationary equilibrium result is achieved via the eddy-lifetime prescription together with rapid distortion of the isotropic spectral tensor – with τM and (initial) inertial-range amplitudes depending on ε via Eqs. ()–() and (), whereby shear-production of TKE is in effect balanced by dissipation. That is, the resultant shear stress 〈uw〉 (expressible now in terms of ε) can be multiplied by 2∂U/∂z to give the implied production rate of 〈uu〉, which with vv and ww (through Γ) gives the implied TKE production rate, amounting to P=ε; such an equilibrium, enforced by τM, can also be inferred from .

 Characteristic length scale

Noting that the spectrum of a variable integrates to the variance of said variable, then invoking Eq. () with the isotropic von Kármán form Eq. () for E(k) and exploiting F11(k1)=∫∫Φ11dk2dk3, one obtains the isotropic streamwise turbulence variance σiso2=2∫0∞955αε2/3LMM5/3[1+(k1LMM)2]-5/6dk1=0.69αε2/3LMM2/3, which is the undistorted streamwise variance. The factor 0.69 is the numerical value of 955πfΓ(13)/fΓ(56), and fΓ(x) is the Euler gamma function see also footnote 4. Then using Eq. () in Eq. () we get a relation for the isotropic (undistorted) turbulence length scale implied by the lifetime model (), LMM≃1.5ΓcτσisodU/dz, where the leading term in parentheses is expected to be of the order of 1.

Relation to observations

suggested that the Mann-model length scale is proportional to the classic mixing length ℓ*≡u*/(dU/dz) multiplied by an empirical constant, LMM=cmℓ*=cmu*,obsdU/dz, where they assign cm=1.7. However, we find from observations that on average cm≈2.3 over flat land, i.e., 〈LMM/ℓ*〉=2.3 (see next section). Combining Eqs. ()–() one sees that cτ decreases with the relative magnitude of measured shear stress (as σiso/u*,obs); this is also expressed usefully through the measured ratio of streamwise fluctuation amplitude to friction velocity: cτ≃1.5Γcmσisou*,obs=1.5Γσu,obs/σisoσu,obs/u*,obscm. From the above and Eq. () one subsequently then finds LMM≃σu,obsdU/dzcmσu,obs/u*,obs. For constant (σu,obs/u*,obs), Eq. () implies that the turbulence scale LMM can be expressed independently of Γ, given σu,obs and dU/dz.

reported the mean profile of σu2(z) from the seminal “Kansas experiment”, showing that (σu/u*)02≈5–6 in the homogeneous atmospheric surface layer (their Fig. 5). The corresponding value of (σu/u*)0 is approximately 2.3; thus, if cm≈2.3 as well, then Eq. () reduces to cτ≈1.5Γσu,obs/σiso. Given the definition of cτ through Eq. (), cτ is a constant; since Eq. () shows σiso is independent of Γ, then σu,obs∝Γ. Consistent with this argument, Eq. () reduces to LMM≈σu,obsdU/dz, which is also evident inserting Eq. () into Eq. (). Using Eq. (), LMM can simply be diagnosed from typical measurements, e.g., 10 min average cup-anemometer output at two (or more) heights. The length LMM can also be cast in terms of variables commonly used in wind engineering, notably the turbulence intensity Iu and shear exponent α. Invoking dU/dz=αU/z and defining Iobs≡σu,obs/U, then Eq. () becomes LMM≈zIobsα.

 Modeled spectra: covariances, anisotropy, and <inline-formula><mml:math id="M157" display="inline"><mml:mi mathvariant="normal">Γ</mml:mi></mml:math></inline-formula>

The spectral Mann model (“MM”) distorts the isotropic von Kármán spectral tensor (Φij(k), Eq. ), per wavenumber via rapid-distortion theory over the wavenumber-dependent eddy lifetime τM, such that the component spectra become anisotropic at wavenumbers outside (lower than) the inertial range; the degree of distortion – and thus anisotropy – are consequently represented by the eddy-lifetime parameter Γ. Above we showed via mixing-length arguments that LMM is independent of Γ, resulting in Eq. (). Possible Γ dependences can also be examined by considering the shear stress 〈uw〉MM=-u*,MM2=2∫0∞F13(k1)dk1 obtained from the modeled spectral tensor component F13(k1)=∫∫Φ13dk2dk3, which is expected to be a function of Γ. Indeed Fig. 4 shows this to be the case, with modeled stress 〈uw〉MM/σiso2 varying almost linearly between 0 and -1 for 0<Γ<5; then u*,MM2/σiso2≈Γ/5. Subsequently from Eq. () one has cτ≈1.55Γcmu*,MMu*,obs≈0.64Γu*,obs/σiso for cm≈2.3, in analogy with Eq. (); thus we expect u*,obs∝Γ, similar to the expected behavior of σu,obs∝Γ following Eq. ().

In addition to the approximate expression (), which is based on the simplified relation u*,MM2/σiso2≈Γ/5, it is possible to derive an exact relation based on the Mann-model shear stress (Eq. ) – but this is cumbersome and analytically intractable. Though derived implicit expressions toward relating {Γ, dU/dz, LMM} to the eddy lifetime and integral of the modeled stress spectrum (Eq. ), these must be evaluated numerically or graphically. An explicit expression corresponding to cm-1=ℓ*/LMM (like Eq.  here) was derived by , but it depends on numerically integrating the stress spectrum.

As spectra fitted to Mann-model outputs correspond to distorted anisotropic turbulence, and noting the Γ dependence of u*,MM discussed above, we expect σu,MM to also depend on Γ. From Fig. 4 of we find σu,MM2/σiso2≃(1+0.14Γ2), which for Γ≳2, the range corresponding to atmospheric boundary layer (ABL) observations e.g.,, becomes roughly σu,MM≈σiso(0.61+0.3Γ).

 Ideal neutral surface-layer implications

Within the atmospheric surface layer (ASL), in the homogeneous stationary limit under neutral conditions, dU/dz→u*/(κz) so that Eq. () reduces to LMM→cmκz≈0.92z. Similarly, in this “log-law regime” εASL,N=u*3/(κz) so that Eq. () becomes ΓASL,N=cτ(3α/2)-1/2(LMM/κz)2/3, or equivalently LMM|ASL,N=(3α/2)3/4κz[Γ/cτ]3/2, which via Eq. () can be written LMM|ASL,N=3α23/4κzcmu*,obs1.5σiso3/2≃1.1κzcmσu,obs/u*,obsσu,obsσiso3/2. Thus for cm=σu,obs/u*,obs, we see that the eddy-lifetime formulation () implies LMM→1.1κz(σu,obs/σiso)3/2 in the neutral ASL. Meanwhile, as noted just above, the mixing-length form () implies LMM→cmκz; this is consistent with Eq. () under the condition that (σu,obs/σiso)≃(cm/1.1)2/3 or roughly σu,obs≈1.6σiso for cm≃2.3.

 Observations and results

Since the choice of eddy lifetime form () leads to a shear-dependent relation () between the spectral-tensor model parameters, one obtains Eq. () for the undistorted (isotropic) length scale, with LMM∝σiso(dU/dz)-1; further invoking a mixing-length argument then leads to a relation () for LMM in terms of quantities that are directly measurable via standard wind-industry (one-dimensional cup) anemometers. Here we test Eq. () as well the assumptions leading to it, through measured wind speed, shear, and turbulent velocity component spectra. We also find a form for the distribution of LMM over all conditions – as would be needed in practice to represent the turbulence length scales of flows experienced by wind turbines at a given site.

For the assumption testing in this section, the spectra used are measured via three-dimensional sonic anemometers on the primary meteorological mast located at the Danish National Test Centre for Large Wind Turbines (Høvsøre), 1.75 km from the western coast of Denmark . The anemometers give 20 Hz samples of all three velocity components and temperature

The sonic anemometers actually give a temperature very close to the virtual temperature, i.e., the temperature including buoyant effects of water vapor.

at heights of 10, 20, 40, 60, 80, and 100 m. This allows calculation of mean speeds, directions, and vertical shear of mean speed over individual 10 min records; in particular we focus on heights of z=20 m and z=80 m, as we are able to calculate shear at (across) these heights using the measurements at 10, 40, 60, and 100 m while also using the measured wind speed components and subsequent spectra at z={20,80} m. The parameters {LMM,Γ,ϵ} are obtained via fits of precalculated Mann-model spectra to the measured velocity-component and stress spectra F11(f), F22(f), F33(f), and F13(f); this is done via Taylor's hypothesis (k1=2πf/U) along with combined least-squares fits .

Testing of assumptions and predicted constraints

The implications of Eqs. ()–() included the independence of LMM and cτ from Γ, as well as (for example) the expected dependence σu,obs∝Γσiso. Indeed we find that LMM is independent of Γ, with no significant statistical correlation: 〈LMMΓ〉/〈LMM2〉〈Γ2〉<0.15 for land or sea sectors at any given height. We also confirm that σu,obs∝Γσiso, which is demonstrated by Figs. –. The first figure displays the joint probability density P(σu,obs,σiso), where σu,obs is the streamwise turbulent variance measured in 10 min intervals, and σiso is calculated using Eq. () with LMM and ϵ from spectral fits corresponding to the same intervals. One can see from Fig.  that σu,obs generally follows σiso, and we find σu,obs/σiso≈5/3. Such evidence corresponds closely to the predicted constraint following Eq. () that σu,obs/σiso should have a value of roughly 1.6 in the neutral surface layer; this is reasonable in the mean, since conditions on average are essentially neutral due to the shape of the stability distribution at Høvsøre . Figure  further shows that σu,obs/σiso∝Γ, consistent with cτ being a constant independent of Γ following Eq. (). The slope of the line in Fig.  also corresponds to the approximate Mann-model behavior σu,MM≈σiso(0.61+0.3Γ) for Γ≳2, outlined at the end of Sect.  above.

Joint distribution of isotropic (un-distorted) variance σiso2(ε,LMM) obtained from fits to measured spectra and observed streamwise variance σu,obs, from height z=80 m over homogeneous land sectors at Høvsøre; dashed line corresponds to σu,obs/σiso=5/3.

Considering wind speeds in the typical turbine operating range of 4–25 m s-1, the Høvsøre data also confirm that 〈σu/u*〉obs≈2.3, consistent with the findings of . Further, the data also show that 〈cm〉≈2.3 so that Eq. () reduces approximately to Eq. (). It is also found that the same approximate trends are seen when considering only U>7 m s-1 (not shown), but with slightly less scatter (narrower joint distributions) away from the predicted σu,obs/σiso behaviors shown in Figs. – (dashed/dotted lines) and discussed above.

The data also show that σu/u* is not correlated with LMM, whether we include all speeds or limit the wind speed range to 7–25 or 4–25 m s-1. Thus this ratio can be treated as a constant in Eq. () for a given height (or throughout the surface layer), using Eq. () over a range of wind speeds.

Ratio of observed streamwise to isotropic fluctuation magnitude versus Γ obtained from spectral fits, plotted as joint probability density function P(Γ,σu,obs/σiso). Dashed (horizontal) line shows σu,obs/σiso=5/3 corresponding to slope of dashed line in Fig. ; dotted line shows the mean linear Γ dependence of σu,obs/σiso.

Turbulence length-scale distributions <inline-formula><mml:math id="M251" display="inline"><mml:mrow><mml:mi>P</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mtext>MM</mml:mtext></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula>

The efficacy of using Eq. () to estimate the spectral length scale LMM can be seen by considering Fig. . The figure displays the joint distribution of turbulence length scale at a height of z=80 m, i.e., PLMM,obs,σu,obs|dU/dz|-1; this is obtained through Eq. () from 10 min measurements and via fitting observed spectra. Figure  is usefully interpreted as the probability-weighted performance of Eq. () for predicting LMM (from σu,obs measured at z=80 m and the shear dU/dz observed over z=60–100 m) versus the LMM obtained from fits of the spectral-tensor model to corresponding 10 min spectra. One sees a 1 : 1 relationship, particularly for the most commonly found values of the length scale; these LMM values range ∼15–50 m

The spectral fits were done using spectral-tensor model output over the parameter ranges of 5<LMM<500 m and 0≤Γ≤5. Some spectra were poorly fitted; since these occurred when Γ=5, cases with Γ>4.95 were excluded from the analysis here. As justification, I note that only a small fraction of the cases (<10 %) had such Γ, and we only consider well-fit spectra for reliable comparison of parameters.

. Compared to the scales calculated from observed spectra, there is some mis-prediction of LMM calculated by Eq. (), but it is relatively rare; this is shown by the low probabilities in Fig.  away from the well-predicted, most commonly occurring LMM.

Joint probability density function of predicted and diagnosed (observed) turbulent length scale, from measurements at Høvsøre over the homogeneous eastern land sectors. x axis: Mann-model scale LMM from spectral fits; y axis: LMM estimated from direct measurements of dU/dz and σu, via Eq. ().

To demonstrate the statistical character of Eq. (), as well as its potential for probabilistic use (e.g., as input to probabilistic load calculations), Fig.  shows the probability density P(LMM). As in Fig. , LMM is again calculated from fits to 10 min spectra and also estimated by σu,obs/(dU/dz), i.e., Eq. (). Additionally Fig.  displays P(LMM) for LMM calculated through Eq. (), i.e., cmu*/(dU/dz); this is done both using the value of cm=1.7 reported by as well as using the approximate mean of 2.3 found to be consistent with measurements and theory in Sects.  and  above. From Fig.  one sees that, for values of turbulent peak scale greater than the mode (∼20 m) up to roughly 150 m, there is a match between the distribution of the diagnosed LMM and distributions of length scale estimated from the forms () based on σu,obs and () based on u* with cm=2.3; these are roughly equivalent for this case over relatively simple homogeneous terrain. It is found that the value of cm=1.7 leads to overprediction of LMM by a factor of 2 or more at scales smaller than 10 m and underprediction by 50 % or more at scales larger than 50 m. The u*-based form () using cm=2.3 matches the spectrally fit diagnosed distribution P(LMM) slightly better than the σu-based form (), with predicted peak (mode) values of LMM being about 3–4 m smaller than the diagnosed peak LMM.

For the homogeneous land case in Fig.  the probability density function of 2.3u*,obs/(dU/dz) matches P(LMM) observed from the spectral fits to within 10 %, over the range 10 m ≲ LMM≲75 m, and the probability density function of σu,obs/(dU/dz) also matches within 10 % over the range 15 m ≲ LMM≲50 m. This is consistent with the darkly colored 1 : 1 patch evident in Fig.  and also shows that Eq. () (and also Eq.  with cm=2.3) is sufficient for probabilistic wind load simulations, for two reasons. First, the well-matched range of scales corresponds to the most commonly found LMM. Secondly, although scales smaller than ∼15 m are not rare (with an occurrence of roughly 1 in 6), they will have a diminishing effect on turbine loads. More specifically, LMM is more than 70 % likely to fall in the 15–75 m range, i.e., P(15 m <LMM<75 m)>0.7, and LMM has more than 86 % likelihood of occurrence between 0 and 75 m, for this homogeneous land case at z=80 m. The relatively common shorter scales correspond to weaker turbulent fluctuations (thus loads), because on average σu,obs∝LMM2/3, as implied by Fig.  and Eqs. ()–(). Further, turbine loads are less influenced by fluctuations characterized by spatial scales significantly smaller than the blade lengths; thus the error in predicted probability for these shorter scales, and the slight underprediction of the most common LMM, should not significantly influence probabilistic load calculations relying on site-specific LMM obtained via measurements and Eq. ().

Probability density function of turbulent length scale from observations at Høvsøre from the homogeneous eastern land sectors. Black: Mann-model scale from fits to spectra; dotted/blue: “mixing-length” formulation (ℓm∝u*/|dU/dz|) with revised constant; dashed/gold:  form for ℓm; red/long-dashed: σu/|dU/dz| form ().

While Eq. () is useful to estimate LMM and P(LMM) as shown above, one expects Eq. () to perform better, as it does not rely on the approximation cm=σu/u*. Indeed 〈cmu*/σu〉 is actually 1.11 (or 1.13 if considering winds only down to 7 m s-1) due to σu/u* being slightly smaller and cm slightly larger than 2.3; using these values in Eq. () gives estimates of LMM closer to the spectrally diagnosed LMM, and within 10 % of P(LMM) over a range of LMM from below 10 m to beyond 100 m. It should also be noted that ignoring speeds below 7 m s-1 can lead to slightly smaller LMM, since these low wind speeds are more influenced by unstable conditions. Indeed for LMM≳50 m, including the lower wind speeds causes both diagnosed and predicted LMM to increase roughly 10 %; this is consistent with larger turbulent eddies being created under unstable conditions.

Estimating <inline-formula><mml:math id="M336" display="inline"><mml:mrow><mml:mi>P</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mtext>MM</mml:mtext></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula> in coastal/offshore conditions

To demonstrate the (probabilistic) use of Eq. () or Eq. () for LMM in somewhat different conditions, we now consider flow from offshore, using data from the same mast and height as above (Høvsøre, z=80 m) but for wind directions between 240 and 300∘. The mast is roughly 1.75 km east of the coastline and subsequently 1.65 km east of a 16–17 m high sand dune that lies 100 m inland, where both are locally oriented in the N–S direction (i.e., for the range of wind directions considered). The dune causes enhanced/accelerated transition of the flow from an offshore (water roughness) to an over-land flow regime ; this results in winds which reflect on-shore and coastal conditions at low heights (below ∼40–80 m depending on stability) and offshore conditions at higher z.

Figure  displays the distribution P(LMM) of spectral-peak (Mann model) length scales for coastal/offshore winds (from west ±30∘), again using Eq. () to estimate LMM along with LMM diagnosed through spectral fits. For comparison the corresponding P(LMM) for easterly winds from Fig.  is also included. Just as for the homogeneous land case shown in Fig. , one sees in Fig.  that, for inhomogeneous coastal conditions, again Eq. () gives P(LMM) basically matching the spectrally fit observations for scales beyond ∼15 m; in this coastal regime the range of well-predicted LMM extends further, to ∼150 m. While one sees that the distribution of LMM is a bit different for the (western) inhomogeneous coastal case than for the (eastern) homogeneous land case, the simple expression () functions similarly for both flow regimes, with the arguments presented in Sect.  again applying here.

Probability density of turbulence length scale LMM from observations at Høvsøre over both the homogeneous land (eastern) sectors and inhomogeneous coastal (western) sectors. Black: LMM from fits to spectra over land; red/long-dashed: new simplified form () over land; purple: LMM from fits to spectra from offshore; cyan/dot-dashed: new simplified form () from offshore.

The u*-based Eq. () also behaves similarly (not shown) as in the homogeneous land case of Fig. , i.e., with gross overpredictions at small scales and underpredictions at large scales. One difference between the coastal and land cases is that, for small LMM, Eq. () overestimates the distribution P(LMM) a bit more for the coastal regime than for the homogeneous land regime (LMM<20 m); as explained above for the land case, an overprediction at the smallest scales is not expected to significantly impact load calculations, due to the relatively small length scales involved.

 Estimation of <inline-formula><mml:math id="M360" display="inline"><mml:mrow><mml:mi>P</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi>L</mml:mi><mml:mtext>MM</mml:mtext></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula> in more complex conditions

To further show the behavior of LMM and the utility of Eq. () at a site with more complex conditions, we examine data from the inhomogeneous forested Danish National Test Centre for Large Wind Turbines site near Østerild in Denmark e.g.,for details. Here sonic anemometer data are available at heights of 10 and 44 m, with concurrent data from three lidars at z={45,80,140,200,300} m. In this study we consider data from the site's “western lidar”

The “western lidar” at Østerild is located ∼1 km west of the northernmost turbines but less than 100 m east of a forest patch and 5–20 km from the North Sea coastline in the prevailing (W–NW) wind directions .

, to measure winds that flow over the forest more than 70 % of the time, where the canopy height is 10–20 m . The analysis here uses one year (May 2010–May 2011) of wind speeds U≥5 m s-1 from the lidar at 45 and 80 m heights along with the “fast” (20 Hz) data from the sonic anemometer at 44 m. The shear dU/dz is measured across 45–80 m; the spectra and subsequent turbulence/Mann-model parameters {LMM,Γ,ε}, as well as and measured quantities {σu,obs,u*,obs}, are obtained from the sonic anemometer. The measurements are significantly higher than twice the forest canopy height, and thus above the roughness sublayer  and amenable to similarity and mixing-length theory e.g., as well as Mann-model use .

Just as Fig.  showed for flow over homogeneous land at Høvsøre in Sect. , here Fig.  displays the probability density of turbulence (Mann-model) length scale LMM observed via spectral fits at z=44 m for Østerild, along with predictions based on both Eq. () via u*,obs and Eq. () via σu,obs.

Probability density function of turbulent length scale from observations at Østerild from the western mast/lidar. Black: Mann-model scale from fits to spectra; dotted-blue: “mixing-length” formulation (ℓm∝u*/|dU/dz|) with revised constant; red: new form (), σu/|dU/dz|.

As in the cases above (homogeneous land and inhomogeneous coastal), the new form () predicts the distribution rather well, particularly for scales ∼10–100 m – despite the shape of P(LMM) being different due to the trees. For the forest case of Fig.  the σu-based form captures both the peak (most likely LMM) and magnitude of P(LMM), while the u*-based form grossly underpredicts LMM, more so than for the previous cases. The latter is likely due to u*,obs being predominantly affected by the canopy (via larger effective roughness) more so than σu,obs, which tends to be more characteristic of the entire ABL . There is, however, a curious minor peak (with a probability ∼1 % as large as the main peak) around scales of ∼300±50 m in the length-scale distribution obtained from spectral fits shown in Fig. ; this is captured by neither the u*-based form () nor σu-based formulations () and (). Although this peak falls spectrally at small wavenumbers that are more difficult to capture when spectrally fitting the Mann model, it actually corresponds to the distance to the next upwind edge of the forest (orchard segment) in the predominant wind directions.

 Discussion

Towards concluding, we first revisit the motivation for (and thus context of) this work: (1) to “close” the eddy-lifetime (τM) formulation as implemented in rapid-distortion theory – allowing relation between Mann-model parameters (LMM,ε,Γ) and the shear (dU/dz) taken to distort the modeled turbulence; (2) to connect the parameters of the spectral turbulence and eddy-lifetime models with atmospheric statistics, both in theory and in practice; (3) to provide a formulation for the turbulence length scale LMM in terms of quantities commonly measured in wind energy; and (4) to demonstrate that the “measurable” form developed for LMM is robust and amenable to use in (probabilistic) wind turbine load calculations. These four motivating goals have basically been realized, as shown in the previous sections, and this work has a number of implications.

Implications and application

A previously suggested form () for LMM, based on friction velocity u* and (10 min) mean wind shear dU/dz , was confirmed here to be sensitive to its proportionality constant cm. But this constant can vary from site to site (and possibly with height), and the published value of cm=1.7  leads to significant error in prediction of LMM for the different conditions (land and sea directions) at Høvsøre and at the forested site of Østerild. Finding cm from sonic anemometer observations via LMM from fits to spectra and friction velocity measurements, Eq. () may perform slightly better over uniform flat terrain compared to the σu-based form () – but this can be considered a site-dependent fit in itself, as was the case when using a diagnosed value of cm=2.3 for the homogeneous flat land sectors at Høvsøre. However, obtaining cm is generally not possible in industrial practice; where it can be obtained, it relies on LMM – which is the quantity desired – thus negating the purpose of Eq. (). While u* can also in principle be estimated from wind speeds taken at multiple heights by cup anemometers, this too is difficult in practice: one must account for stability, not to mention the need for measurements at multiple heights in the surface layer (or worse, the limited validity of similarity theory above the ASL). Furthermore, it is expected that cm is a function of the (local) surface roughness, as demonstrated by the different results found over the forested Østerild site. Thus the form () is preferable, since it requires only the commonly measured quantities σu and dU/dz. This simple form also gave good estimates of P(LMM) in the forested case – without the need for tuning, whereas the u*-based form () requires a re-calculation of its coefficient cm for such cases.

Since Eq. () gave yet better performance than both its simplified form () and the u*-based relation (), one might suggest its use. But Eq. () requires cm/(σu/u*), where cm is difficult to obtain, as discussed in the previous paragraph. However, although cm might vary from site to site (or perhaps with height), it was found that the ratio cm/(σu/u*) did not vary appreciably – consistent with the good performance of the simplified form (), which assumes cm≈σu/u*, across sites and regimes.

One interesting implication of the testing of assumptions then follows from the finding that 〈σu/u*〉obs≈2.3, consistent in the surface layer with . Examining the joint behavior of σu/u* and the stability parameter (inverse Obukhov length) L-1, the sonic anemometer data available at multiple heights in this study show no correlation between these two quantities. The dimensionless profiles σu2(z)/u*02 and u*2(z)/u*02 shown by also imply σu2(z)u*2(z)≈(2.3)2, with the ratio converging to a constant above the surface layer (z≳0.1h, where the atmospheric boundary-layer depth h typically ranges from ∼200 m in stable conditions to 1 km or more in convective conditions). The flat-terrain Høvsøre data in fact show the mean value 〈σu/u*〉obs to be independent of z. If one knew how cm varied with height (and stability), then one could also use Eqs. () and () from measurements at one height range, to estimate LMM at higher z (for a given stability range). Over flat terrain, on average the peak spectral scale for streamwise fluctuations (λu) grows with z

The peak length scale also grows with boundary-layer depth h in convective conditions and thus with increasingly negative inverse Obukhov length L-1 e.g.,. But over all stability conditions, which are dominated by neutral conditions , and over an expected distribution of h at a given site, the basic growth of λu with z is consistent with reporting λu∝z for neutral conditions.

. Therefore, if we take LMM∝λu, then with Eq. () one expects the ratio cm/(σu/u*) to increase with z as well. Thus from () the Mann-model length scale LMM will increase with height relative to the mixing length ℓ*≡u*/(dU/dz), so at higher z one would expect the general form () to be yet more accurate than its approximate form (); however, this is not likely for wind turbine rotor heights, except in very stable conditions . Unfortunately the sonic-anemometer measurements available for this study did not include heights well beyond the surface layer, so such variation was difficult to detect.

It is also notable that Fig.  appears to imply the relative error (e.g., in %) in estimating LMM with Eq. () grows for less common values of LMM, particularly very large scales (and also at very small scales if including U<7 m s-1). Thus Eq. () is recommended first for estimation of P(LMM). However, the error at large scales is in part dependent on the limited (10 min) sample lengths and the fitting routine, as there are very few points to fit at the lowest frequencies. Use of 30 min samples can reduce such scatter, and modification of the fitting algorithms may also improve estimations of the larger scales.

Ongoing work includes wind-speed-dependent prediction of LMM, particularly the conditional statistics P(LMM|U). Further concurrent work also entails systematic accounting for the rotor size (shear distance) relative to height (i.e., Δz/z) within the distribution of length scales; following  and a semi-empirical derivation of P(LMM) including Δz/z has been obtained but demands more data for validation and publication. Understanding of the latter facilitates “vertical extrapolation” of LMM and measured turbulence and shear statistics, as well as accounting for the effect of rotor size or shear measurement span.

 Conclusions

The eddy lifetime of , which is part of commonly used turbulence modeling for wind turbine design load cases e.g.,, leads to a relation for turbulence (spectral-peak) length scale LMM of LMM≃cm(σu,obs/u*,obs)σudU/dz, where cm and σu,obs/u*,obs are essentially constants for a given height z, and cm/(σu,obs/u*,obs) is found to fall between 1 and 1.11 for the three flow regimes analyzed.

Theory and measurements support the assumption that cm/(σu,obs/u*,obs)≈1, roughly constant for different atmospheric flow regimes; the turbulence length scale can consequently be approximated by LMM≃σudU/dz. Thus typical 10 min mean cup anemometer measurements can be used to estimate LMM.

LMM is affected by atmospheric stability; this effect is contained within σu and dU/dz.

In terms of the classic mixing-length form u*/|dU/dz|, the turbulence length scale LMM in the spectral-tensor model is observed to be larger (by ca. 30–40 %) than previously reported by .

The data are within an SQL database at DTU and are not publicly available.

The author declares that he has no conflict of interest.

Acknowledgements

The author thanks the reviewers for their time and effort towards constructive criticism of the present article, and thanks are owed to Nikolay Dimitrov for discussions around probabilistic loads. This work was partly supported by the DTU Wind Energy internally funded cross-sectional project “Wind to Loads”. Edited by: Horia Hangan Reviewed by: two anonymous referees

 References Abramowitz, M. and Stegun, I. A.: Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables; 9th printing, Dover, New York, 1972. Andre, J. C. and Lesieur, M.: Influence of helicity on the evolution of isotropic turbulence at high Reynolds number, J. Fluid Mech., 81, 187–207, 10.1017/S0022112077001979, 1977. Berg, J., Vasiljevic, N., Kelly, M. C., Lea, G., and Courtney, M.: Addressing Spatial Variability of Surface-Layer Wind with Long-Range WindScanners, J. Atmos. Ocean. Tech., 32, 518–527, 10.1175/JTECH-D-14-00123.1, 2015. Caughey, S., Wyngaard, J. C., and Kaimal, J.: Turbulence in the evolving stable boundary-layer, J. Atmos. Sci., 36, 1041–1052, 1979. Chougule, A., Mann, J., Segalini, A., and Dellwik, E.: Spectral tensor parameters for wind turbine load modeling from forested and agricultural landscapes, Wind Energy, 18, 469–481, 2015. Chougule, A. S., Mann, J., Kelly, M., and Larsen, G. C.: Modeling Atmospheric Turbulence via Rapid Distortion Theory: Spectral Tensor of Velocity and Buoyancy, J. Atmos. Sci., 74, 949–974, 10.1175/JAS-D-16-0215.1, 2017. Comte-Bellot, G. and Corrsin, S.: Simple Eulerian time correlation of full- and narrow-band velocity signals in grid-generated, “isotropic” turbulence, J. Fluid Mech., 48, 273–337, 1971. de Mare, M. T. and Mann, J.: On the Space-Time Structure of Sheared Turbulence, Bound.-Lay. Meteorol., 160, 453–474, 10.1007/s10546-016-0143-z, 2016. Dimitrov, N. K., Natarajan, A., and Kelly, M.: Model of Wind Shear Conditional on Turbulence and its impact on Wind Turbine Loads, Wind Energy, 18, 1917–1931, 10.1002/we.1797, 2015. Dimitrov, N. K., Natarajan, A., and Mann, J.: Effects of normal and extreme turbulence spectral parameters on wind turbine loads, Renew. Energ., 101, 1180–1193, 2017. Garratt, J. R.: Surface Influence Upon Vertical Profiles in the Atmospheric Near-Surface Layer, Q. J. Roy. Meteor. Soc., 106, 803–819, 10.1002/qj.49710645011, 1980. Hansen, B. O., Courtney, M., and Mortensen, N. G.: Wind Resource Assessment – Østerild National Test Centre for Large Wind Turbines, Tech. Rep. DTU Wind Energy E-0052(EN), Risø Lab/Campus, Danish Tech. Univ. (DTU), Roskilde, Denmark, 2014. IEC 61400–1, Edition 3: Wind turbine generator systems – Part 1: Safety requirements, International Electrotechnical Comission, Geneva, Switzerland, 2005. Kaimal, J. C., Wyngaard, J. C., Izumi, Y., and Coté, O. R.: Spectral characteristics of surface-layer turbulence, Q. J. Roy. Meteor. Soc., 98, 563–589, 1972. Kelly, M. and Gryning, S.-E.: Long-Term Mean Wind Profiles Based on Similarity Theory, Bound.-Lay. Meteorol., 136, 377–390, 2010. Kelly, M., Larsen, G., Dimitrov, N. K., and Natarajan, A.: Probabilistic Meteorological Characterization for Turbine Loads, J. Phys. Conf. Ser., 524, 012076, 10.1088/1742-6596/524/1/012076, 2014a. Kelly, M., Troen, I., and Jørgensen, H. E.: Weibull-k revisited: “tall” profiles and height variation of wind statistics, Bound.-Lay. Meteorol., 152, 107–124, 2014b. Lesieur, M.: Turbulence in fluids: stochastic and numerical modelling, Springer, Dordrecht, NL, 10.1007/978-94-009-0533-7, 1990. Liu, S. and Liang, X.-Z.: Observed Diurnal Cycle Climatology of Planetary Boundary Layer Height, J. Climate, 23, 5790–5809, 2010. Mann, J.: The spatial structure of neutral atmospheric surface-layer turbulence, J. Fluid Mech., 273, 141–168, 1994. Mann, J.: The spectral velocity tensor in moderately complex terrain, J. Wind Eng., 88, 153–169, 2000. Mann, J., Ott, S., Jørgensen, B. H., and Frank, H. P.: WAsP Engineering 2000, Tech. Rep. Risø-R-1356(EN), Risø National Laboratory, Roskilde, Denmark, 2002. Mann, J., Astrup, P., Jensen, N., Landberg, L., and Jørgensen, H.: The meteorology of “the very large wind turbines”, in: Proc. of the 2004 European Wind Energy Conference, European Wind Energy Association (EWEA), London, 2005. Peltier, L. J., Wyngaard, J. C., Khanna, S., and Brasseur, J. G.: Spectra in the Unstable Surface Layer, J. Atmos. Sci., 53, 49–61, 1996. Peña, A., Gryning, S.-E., and Mann, J.: On the length-scale of the wind profile, Q. J. Roy. Meteor. Soc., 136, 2119–2131, 2010. Peña, A., Floors, R. R., Sathe, A., Gryning, S.-E., Wagner, R., Courtney, M., Larsén, X. G., Hahmann, A. N., and Hasager, C. B.: Ten Years of Boundary-Layer and Wind-Power Meteorology at Høvsøre, Denmark, Bound.-Lay. Meteorol., 158, 1–26, 10.1007/s10546-015-0079-8, 2016. Pope, S. B.: Turbulent Flows, Cambridge University Press, Cambridge, UK, 2000. Raupach, M., Thom, A., and Edwards, I.: A Wind-Tunnel Study of Turbulent-Flow Close to Regularly Arrayed Rough Surfaces, Bound.-Lay. Meteorol., 18, 373–397, 10.1007/BF00119495, 1980. Sathe, A., Mann, J., Barlas, T. K., Bierbooms, W., and van Bussel, G.: Influence of atmospheric stability on wind turbine loads, Wind Energy, 16, 1013–1032, 2013. Savill, A. M.: Recent developments in rapid-distortion theory, Annu. Rev. Fluid Mech., 19, 531–575, 1987. Sogachev, A. and Kelly, M.: On Displacement Height, from Classical to Practical Formulation: Stress, Turbulent Transport and Vorticity Considerations, Bound.-Lay. Meteorol., 158, 361–381, 10.1007/s10546-015-0093-x, 2016. Sogachev, A., Cavar, D., Kelly, M., and Bechmann, A.: Effective roughness and displacement height over forested areas, via reduced-dimension CFD, Tech. Rep. DTU Wind Energy E-0161(EN), Risø Lab/Campus, Danish Technical University (DTU), Roskilde, Denmark, 2017. von Kármán, T.: Progress in the Statistical Theory of Turbulence, P. Natl. Acad. Sci. USA, 34, 530–539, 10.1073/pnas.34.11.530, 1948. Wyngaard, J. C.: Turbulence in the Atmosphere, Cambridge University Press, Cambridge, UK, 2010.