SurfaceLayer Similarity Functions for Dissipation Rate and Structure Parameters of Temperature and Humidity Based on Eleven Field Experiments
 1k Downloads
 5 Citations
Abstract
In the literature, no consensus can be found on the exact form of the universal funtions of MoninObukhov similarity theory (MOST) for the structure parameters of temperature, \({C_T}^2\), and humidity, \({C_q}^2\), and the dissipation rate of turbulent kinetic energy, \(\varepsilon \). By combining 11 datasets and applying data treatment with spectral data filtering and errorweighted curvefitting we first derived robust MOST functions of \({C_T}^2, {C_q}^2\) and \(\varepsilon \) that cover a large stability range for both unstable and stable conditions. Second, as all data were gathered with the same instrumentation and were processed in the same way—in contrast to earlier studies—we were able to investigate the similarity of MOST functions across different datasets by defining MOST functions for all datasets individually. For \({C_T}^2\) and \(\varepsilon \) we found no substantial differences in MOST functions for datasets over different surface types or moisture regimes. MOST functions of \({C_q}^2\) differ from that of \({C_T}^2\), but we could not relate these differences to turbulence parameters often associated with nonlocal effects. Furthermore, we showed that limited stability ranges and a limited number of data points are plausible reasons for variations of MOST functions in the literature. Last, we investigated the sensitivity of fluxes to the uncertainty of MOST functions. We provide an overview of the uncertainty range for MOST functions of \({C_T}^2, {C_q}^2\) and \(\varepsilon \), and suggest their use in determining the uncertainty in surface fluxes.
Keywords
MoninObukhov similarity theory Scintillometry Structure parameters Turbulent kinetic energy dissipation Turbulence1 Introduction
MoninObukhov similarity theory (MOST; Monin and Obukhov 1954) is used to describe turbulence characteristics in the atmospheric surface layer (Stull 1988). MOST applies to statistics (e.g. variances, vertical gradients or structure parameters) of scalars such as temperature (T), specific humidity (q), the three velocity components (u,v,w) and the dissipation rate of turbulent kinetic energy (\(\varepsilon \)). The application of MOST allows measured turbulence statistics to be related to temperature, humidity and wind scales, which are defined in terms of the turbulent fluxes of heat (H), moisture (\(L_vE\)), momentum (\(\tau \)), the Obukhov length scale (L) and the height above the surface (z). Through these relations (MOST functions) it is possible to determine turbulent fluxes from measured turbulence statistics once the MOST functions are known. These MOST functions are assumed to be universal, which means that once they are determined they are equal for all similar circumstances.
Here we focus on MOST functions that are relevant to scintillometry, notably that of the structure parameters of temperature (\({C_T}^2\)) and humidity (\({C_q}^2\)), and \(\varepsilon \). Scintillometers measure the pathaveraged turbulence intensity of the refractive index of air, which is mainly dependent on temperature and humidity fluctuations (e.g. De Bruin et al. 1995). The sensitivity of a scintillometer for these temperature and/or humidity fluctuations is dependent on the wavelength at which it operates; \({C_T}^2\) can be obtained from an optical scintillometer, and \({C_q}^2\) can be acquired from combined optical and microwave scintillometers (e.g. Andreas 1989; Lüdi et al. 2005; Meijninger et al. 2006). Furthermore, \(\varepsilon \) can be obtained with a smallaperture scintillometer (Hartogensis et al. 2002). The main application that we have in mind is to obtain \(H, L_vE\) and \(\tau \) using \({C_T}^2, {C_q}^2\) and \(\varepsilon \) determined from scintillometer data.
Several MOST functions for \({C_T}^2, {C_q}^2\) and \(\varepsilon \) have been reported in the literature, of which an overview is given in Online Appendix.^{1} This overview illustrates that MOST functions of the same parameter vary largely across different studies. The individual functions have been derived from different measurement instruments, environmental and meteorological conditions and data analysis procedures of separate datasets, as listed in Braam et al. (2014). As all of these factors may have contributed to the variations in the MOST functions, it cannot be concluded from the overview in the online Appendix which factors have caused the variations and if MOST functions can actually be applied across different datasets. Hartogensis et al. (2003) made an inventory of the uncertainty of all parameters that are used in the scintillometer heatflux calculations and showed that differences in MOST functions cause up to 20 % difference in H at the free convection limit. This difference is substantial and can have a considerable influence on intercomparison studies of different measurement techniques. This clearly demonstrates the need to determine welldefined MOST relations that do not suffer from the issues listed above and that consensus is needed on the MOST functions to be used in flux calculations.
Besides the large variability in the MOST functions, earlier studies found that MOST functions of \({C_T}^2\) and \({C_q}^2\) are not always equal (Li et al. 2012; Maronga 2014). This dissimilarity is in contradiction to the assumptions of MOST that heat and moisture are transported by the same mechanism and therefore should have the same MOST function (De Bruin et al. 1993). MOST functions of \({C_T}^2\) and \({C_q}^2\) need to be considered separately to investigate if the dissimilarity between MOST functions of \({C_T}^2\) and \({C_q}^2\) is consistent for different datasets.

The large amount of data over a wide stability range allows us to determine a sound uncertainty range of MOST parameters, which can be used in estimating the uncertainty in the calculated fluxes.

We define MOST functions from all individual datasets. Based on the spread in these functions we then discuss plausible reasons for the large range of functions reported in the literature. As part of this we also investigate if the (dis)similarity between MOST functions of \({{C_T}^2}\) and \({{C_q}^2}\) differs between datasets.

We discuss the validity of the derived MOST relations in two ways. First we show if and to what extent the MOST relations are influenced by crosscorrelation of equal terms in the dependant and independent variables of the MOST relations (also known as selfcorrelation). Second, we determine the impact of the uncertainty of the MOST functions on the fluxes. For any issue that we encounter regarding the violation of MOST, we discuss the extent to which the fluxes are affected.
2 Theory
2.1 MOST Framework
In Sect. 1 we introduced the assumption of MOST that heat and moisture are transported by the same mechanism and therefore in principle should have the same MOST function. However, there are two reasons for dissimilar transport mechanisms between heat and moisture. The first is due to a dissimilar transport efficiency, expressed by different Prandtl and Schmidt numbers, in idealized conditions. The second is due to the inclusion of extra sources of heat or moisture due to horizontal or vertical advection (entrainment) or no collocation of the sources that violate MOST (De Bruin et al. 1999). For example, for horizontal advection, the vegetationair interaction affects moisture, but not the temperature of the air (Katul et al. 2008). Under these conditions the fluctuations of T and q may not be correlated, and thus \({C_q}^2\) may be more vulnerable to nonlocal effects than \({C_T}^2\). Moreover, under very stable conditions, turbulence may cease, and the flow is governed by intermittent turbulence, drainage flows along slopes, and pressure perturbations (Mahrt 1999). In Sect. 2.4 past work is discussed that deals with these nonlocal influences that disturb the heat–moisture similarity. We attempt to filter out the nonlocal effects as much as possible from our data and attribute differences between MOST relations of \({C_T}^2\) and \({C_q}^2\) to differences in the Prandtl and Schmidt numbers.
2.2 Definition of \({C_X}^2\) and \(\varepsilon \)
2.3 MOST Functions \({{C_X}^2}\) and \({\varepsilon }\) Based on FluxProfile Relations
2.4 MOST Functions of \(f_{{C_X}^2}\) and \(f_{\varepsilon }\) in the Literature
Different field experiments are described in the literature, which has resulted in different formulations for \(f_{{C_X}^2}\) and \(f_{\varepsilon }\). In Online Appendix we give an overview of earlier reported MOST functions. Several formulations of \(f_{{C_X}^2}\) follow the base function proposed by Wyngaard et al. (1971), which consists of two coefficients \(c_1\) and \(c_2\). In this formulation, \(c_1\) indicates the neutral limit and \(1/c_2\) the inflection point of the function on a semilogarithmic scale. For \(f_{\varepsilon }\), the most common base function is that proposed by Wyngaard and Coté (1971). In the online Appendix we specify the neutral limit of the functions and for \(f_{{C_X}^2}\) under unstable conditions we give an additional term \(c_1c_2^{2/3}\) when the base function of Wyngaard et al. (1971) was used; \(c_1c_2^{2/3}\) is a measure of the shape of the function under free convective conditions (De Bruin et al. 1995). Note that the values for \(c_1\) and \(c_2\) are interconnected; if the neutral range yields a relatively high \(c_1\), then \(c_2\) should be high as well to conserve the scaling coefficient for free convection. The parameters to compare are therefore \(c_1\) for neutral conditions and \(c_1c_2^{2/3}\) for free convection.
Note that base functions other than those of Wyngaard et al. (1971) and Wyngaard and Coté (1971) have been used in other studies. Besides that, different MOST function coefficients were found when the same base function was used. For example, the base function of Wyngaard et al. (1971) was applied to other datasets (Hill et al. 1992; De Bruin et al. 1993; Li et al. 2012; Maronga 2014; Braam et al. 2014), which resulted in coefficients that deviate from the coefficients used by Wyngaard et al. (1971). For \(f_{{C_T}^2}, c_1\) ranges from 4.4 to 8.1, and \(c_1c_2^{2/3}\) from 0.93 to 1.58.
Overall, the neutral limit of \(f_{{C_T}^2}\) ranges from 4.3 (Kanda et al. 2002) to 8.1 (Hill et al. 1992). In the field of scintillometry, the function of Wyngaard et al. (1971) with a neutral limit of 4.9 is most frequently used to determine H, whereas it differs from the neutral limit of 6 that we derived in the previous section. The neutral limit found by Maronga (2014), which is equal to 6.1, is closest to the neutral limit of 6 that we derived. In the studies from Kohsiek (1982) and Roth et al. (2006) it is shown that the data do not level down to a neutral limit but rather show a further increase towards neutral conditions. We reason that this is because H and \(\theta _{*}\) go towards zero at the transition from unstable to stable conditions, which makes the dimensionless group go towards infinity. Braam et al. (2014) demonstrated the influence of different stability ranges and regression techniques on MOST functions of \(f_{{C_T}^2}\) under unstable conditions, which explains part of the variation in reported MOST functions. Moreover, they concluded with measurements at various heights between 2.3 m and 58.1 m that MOST functions are heightdependent.
For \(f_{{C_q}^2}\), neutral limits of \(3.5\pm 0.1\) and 6.3 were found by Li et al. (2012) and Maronga (2014) respectively. As was discussed in Sect. 2.1, MOST functions of \({{C_T}^2}\) and \({{C_q}^2}\) may differ under the influence of nonlocal effects or due to different Prandtl and Schmidt numbers. A problem with nonlocal terms is that they cannot be quantified from typically available turbulence measurements. These data are available through LES models and turbulence can be simulated under ideal and nonideal circumstances with such a model, which provides a means to study the effect of MOST violation on MOST functions. Although LES modelling results are restricted to nearfree convection conditions, Maronga (2014) demonstrated in his LES study that when entrainment was simulated, \(f_{{C_q}^2}\) diverges from \(f_{{C_T}^2}\). Maronga et al. (2014) quantified the effects of surface heterogeneity on \({{C_T}^2}\) and \({{C_q}^2}\) and found that an heterogeneous surface generates additional temperature and moisture fluctuations leading to higher \({{C_T}^2}\) and \({{C_q}^2}\) values. This effect did modify the MOST functions but the effect was very small. Li et al. (2012) and Van de Boer et al. (2014) made an effort to parametrize nonlocal effects using local turbulence parameters, i.e., the correlation coefficient \(R_{Tq}\) and the skewness of q. Li et al. (2012) showed that from weakly unstable to neutral conditions, \(R_{Tq}\) decreases progressively from near one to near zero and derived a dissimilarity ratio that scales with \(R_{Tq}\). Van de Boer et al. (2014) observed dissimilarity for MOST functions of the variance of humidity that they were able to relate to entrainment using the skewness of q. They suggested use of an extra term in the fluxvariance MOST functions to capture the effect of entrainment, where this term includes the entrainment ratio for humidity and the boundarylayer height. The fact that similarity between \(f_{{C_T}^2}\) and \(f_{{C_q}^2}\) may not always be valid confirms that we should consider functions for \(f_{{C_q}^2}\) and \(f_{{C_T}^2}\) separately. Note, however, that our study does not investigate the dissimilarity between \(f_{{C_T}^2}\) and \(f_{{C_q}^2}\) but rather the differences between MOST functions of different datasets.
For \(f_{\varepsilon }\), Frenzen and Vogel (2001), Pahlow et al. (2001) and Hartogensis and De Bruin (2005) reported neutral limits below 1, which implies that the production by shear and the dissipation of TKE are not in balance. Hartogensis and De Bruin (2005) further discuss the neutral limits that were reported before. It is clear from the studies that do not keep the neutral limit fixed at 1 that the data do not fulfil the expectation that the neutral limit it equal to 1. Instead, the deviations from the ideal behaviour are the rule rather than the exception.
3 Methodology
Essential to our approach is that we use many datasets that are all processed with the same algorithms. In the following we describe the data processing and we discuss how we fit MOST functions to the data.
3.1 Data Description
Overview of field experiments including the surface cover, type of surrounding area, measurement height, vegetation height, and if available, a reference with more detailed information about the experiment or environmental conditions
Experiment  Surface cover  Surroundings  Measurement height (m)  Vegetation height (m)  References 

BBC Aug–Sep 2001  Grass \(++\)  Grass \(++\)  3.32  0.10  Crewell et al. (2004)) 
Matador May–Jun 2002  Desert \(\) \(\)  Croplands \(+\)/\(\) \(\)  8.61  0.00  Renno et al. (2004) 
SudMed Oct–Nov 2002  Olive grove \(+\)  City \(\) \(\)  9.09 \(^\mathrm{a}\)  6.50  Hoedjes et al. (2007) 
8.87 \(^\mathrm{b}\)  
LITFASS’03 May–Jun 2003  Crop: Colza \(++\)/\(+\), Rye \(+\)/\(\)  Croplands \(+\)/\(\)  3.51 \(^\mathrm{c}\)  1.15 \(^\mathrm{c}\)  
3.53 \(^\mathrm{d}\)  0.65 \(^\mathrm{d}\)  
Transregio32’08 Aug 2008  Sugarbeet \(++\)  Croplands \(++\)  2.20  0.30  Graf et al. (2010) 
Chilbolton Oct–Nov 2008  Grass \(\)  Grass \(\)  4.46  0.10  Kesteren and Hartogensis (2010) 
LITFASS’09 Jun–Jul 2009  Crop: Triticale o/\(\)  Croplands \(+\)/\(\)  3.20  1.00  Beyrich et al. (2012) 
Transregio32’09 Aug 2009  Crop: Wheat \(++\)  Croplands \(++\)  2.40  0.40  Kesteren et al. (2013) 
BLLAST Jun–Jul 2011  Crop: Grass \(+\), Wheat o  Croplands \(+\)/o  2.55 \(^\mathrm{e}\)  0.30 \(^\mathrm{e}\)  Lothon et al. (2014) 
3.00 \(^\mathrm{f}\)  1.00 \(^\mathrm{f}\)  
LITFASS’12 May–Jul 2012  Crop: Colza \(+\)/\(\), Rye \(+\)/\(\)  Croplands \(+\)/\(\)  3.16 \(^\mathrm{c}\) 3.17 \(^\mathrm{d}\)  1.15 \(^\mathrm{c}\) 1.50 \(^\mathrm{d}\)  
Omseson Apr 2013  Irrigated desert; Crop: Chile \(++\)  Croplands \(+\)/\(\)  2.74  0.70  Hoedjes et al. (2002) 
We used the fluxsoftware package EddyPro (v5.1.1) from LICOR Biosciences Inc. (Lincoln, Nebraska, USA) to calculate turbulent fluxes. All standard data treatment and fluxcorrection procedures were included, most notably axis rotation with the planarfit procedure (Wilczak et al. 2001), raw data screening (Vickers and Mahrt 1997) and lowpass filtering effect (Moncrieff et al. 1997). In addition, error estimates of the fluxes were determined within EddyPro based on Finkelstein and Sims (2001), and quality flags were determined based on Mauder and Foken (2004).
We calculate \({C_X}^2\) and \(\varepsilon \) from the power spectra of T, q and u for every 30min interval using Eqs. 4 and 5. The time series of T, q and u used in this procedure were taken from EddyPro after their processing level 6, i.e., after data pass all screening and correction procedures described above that are applicable to the raw data series. In this way we ensured that the data used for flux calculations and for the turbulence parameters are identical.
We use the MATLAB algorithms as described by Hartogensis (2006), firstly to determine \(S_X(f)\) and \(S_u(f)\) from measured time series using the fast Fourier tranform, and secondly to obtain \({C_X}^2\) and \(\varepsilon \). It is essential in this procedure to distinguish the inertial range of the spectra as Eqs. 4 and 5 are only valid in the inertial range.

at least 15 % of the total spectrum has an inertial range behaviour, which is fulfilled when the calculated r.m.s. of \({C_X}^2\) and \(\varepsilon \) is within 20 % of the average value for that part of the spectrum, and when the slope of the spectrum is within 10 % of the theoretical \(\)5/3 slope (Hartogensis and De Bruin 2005).

\({C_T}^2, {C_q}^2\) and \(\varepsilon \), appropriately scaled with z for neutral conditions, i.e., \(z^{2/3}\) for \({C_T}^2\) and \({C_q}^2\), and z for \(\varepsilon \), have a maximum value of \(10^{1.5}, 10^{9}\) and \(10^{4}\) respectively, whereas the dimensionless groups of \({C_T}^2, {C_q}^2\) and \(\varepsilon \), are set to a maximum value of 25, 25 and 10 respectively.

the number of samples used for flux calculations should at least be 99 % of the total number of samples expected over a 30min interval with 20 Hz.

the flux quality control flag based on Mauder and Foken (2004) should equal zero (the highest possible quality class). This flagging system follows the agreement made during the CarboEuropeIP workshop in 2004 (Mauder et al. 2013).

data for unstable conditions can only include daytime data, whereas stable conditions can occur both during the day and night.

heat fluxes H close to zero (\( 2 < H \) [W m\(^{2}\)] \( < 5\)) are discarded.

data points with excessive large or small error weights were removed from the dataset. This was done using an initial regression and fitting a Gumbel distribution to the weights of the data points. Next, a power density function of the weights was determined based on 100 bins, and data points in bins that represented less than 0.25 % of all data points were removed.
3.2 Fitting MOST Functions to Data
In determining the relation between the MOST dimensionless groups we use an errorweighted, semilogarithmic, orthogonal distance regression. We do this procedure many times on subsamples of the dataset that allows us to estimate the uncertainty of the MOST parameters. In the following we explain the regression procedure further.
Braam et al. (2014) discuss various regression approaches for MOST functions and conclude that an errorweighted, logarithmic, orthogonal distance regression is the optimal. We follow their approach but take a semilogarithmic scale as the dimensionless groups do not have a clear lognormal distribution. The conversion of Eqs. 19–22 for application on a semilogarithmic scale is given in Appendix 1.
The overall MOST functions for all datasets combined are based on 1000 subsample regressions consisting of randomly selected subsets that contain 20 % of the data points. The functions of individual datasets are based on 200 subsample regressions consisting of 30 % of the data points. To ensure that for each subsample regression a representative z/L range is included, we divided the z/L range into 10 bins of equal size and sampled an equal amount of samples from each bin. From the distribution of the MOST parameters \(c_1\) and \(c_2\), thus obtained, we determined a centrevalued \(c_1\)\(c_2\) pair that represents the overall MOST function. Additionally, we determined a minimum and maximum \(c_1\) and \(c_2\) estimate that represents the uncertainty range of the MOST parameters. The details of the procedure that is followed to determine \(c_1\) and \(c_2\) and their uncertainty estimates are given in Appendix 2.
4 Results and Discussion
4.1 Overall MOST Functions for \({C_T}^2\)
Coefficients of MOST functions \(f_{{C_T}^2}, f_{{C_q}^2}\) and \(f_{\varepsilon }\) including the uncertainty based on the subsample regressions as described in Appendix 2. Note that, following Eq. 22, the neutral limit of \(f_{\varepsilon }\) for stable conditions is equal to \({c_1}^{3/2}\)
\(f_{{C_T}^2}\)  \(f_{{C_q}^2}\)  \(f_{\varepsilon }\)  

Unstable  
\(c_1\)  5.6  4.5  0.88 
Uncertainty  5.1–6.3  4.3–4.7  0.83–0.94 
\(c_2\)  6.5  7.3  2.06 
Uncertainty  5.5–7.6  7.0–7.7  1.93–1.56 
\({c_1c_2}^{2/3}\)  1.60  1.20  
Uncertainty  1.64–1.62  1.18–1.21  
Stable  
\(c_1\)  5.5  4.5  0.88 
Uncertainty  5.1–5.9  4.1–4.8  0.84–0.95 
\(c_2\)  1.1  1.1  1.42 
Uncertainty  1.2–1.1  1.2–1.1  1.43–1.35 
 1.
True neutral conditions where, due to strong wind, mechanicallygenerated turbulence is dominant over buoyancygenerated turbulence such that z/L is small but with \(\theta _{*}\) sufficiently large. These data points tend to have the highest weight in the regression in this range.
 2.
Day–night transitions where, independent of wind speed, z/L is small since the heat flux changes from positive to negative and \(\theta _{*}\) has a value near zero. With \(\theta _{*}\) close to zero, the dimensionless group of \({C_T}^2\) goes towards infinity. We applied a data filter to avoid the dominance of nearneutral singularity in the dimensionless group of \({C_T}^2\). The remaining data that show this singularity tend to have relatively large errors and bare little weight in the overall regression result.
The neutral limit that we find for \(f_{{C_T}^2}\) under unstable conditions (5.6) is within the range of neutral limits that were given in the literature. Neutral limits of \(f_{{C_T}^2}\) from Wyngaard et al. (1971) and De Bruin et al. (1993) are well below the neutral limit that we find. Furthermore, the neutral limit of \(f_{{C_T}^2}\) and the free convection parameter \(c_1c_2^{2/3}\) (1.60) are both mostly consistent with that of the function proposed by Maronga (2014), who determined \(f_{{C_T}^2}\) based on simulated turbulence from a LES model. Note that the function from Maronga (2014) is partially hidden by the fluxprofilederived function. The fact that the MOST functions are very similar to those found by Maronga et al. (2014) based on idealized turbulence in a LES model strengthens our case that the selected data indeed follow MOST.
The neutral limit of \(f_{{C_T}^2}\) for stable conditions (5.5) is consistent with that from unstable conditions (5.6). The function that we find for stable conditions is closest to that of Hartogensis and De Bruin (2005) and Li et al. (2012). Several other functions from the literature lie higher than our function. A potential reason for this is that these functions were determined without an errorweighted regression. Larger scatter is found for very stable conditions and the variation between MOST functions from earlier studies is larger as well. It is well known that MOST breaks down under very stable conditions, where fluxes are near zero and mechanisms other than turbulence dominate the flow (Mahrt 1999; BouZeid et al. 2010). The likely violation of MOST under very stable conditions makes that the correct shape of the MOST function for a particular situation is uncertain. For practical applications, when the MOST functions are used to estimate fluxes, the impact of using a nonexact MOST function for very stable conditions is relatively small in absolute terms as the fluxes are close to zero. Towards neutral conditions turbulence is well developed and fluxes are larger, and this is also where the different functions from the literature converge, giving the MOST function a lower uncertainty. In Sect. 4.6 we discuss how the uncertainty of MOST functions affect the fluxes.
From comparison of \(f_{{C_T}^2}\) that we found with the one that follows from the fluxprofile relations, we see that the neutral limit of the fluxprofilederived function (6.0) is within the uncertainty range that we determined for the neutral limit. Moreover, the overall shape of the two functions is also strikingly similar towards more unstable conditions. This indicates that \(f_{{C_T}^2}, \phi _X\) and \(\phi _U\) are consistent, which is a strong experimental evidence for the validity of the MOST framework.
4.2 Overall MOST Functions for \({C_q}^2\)
Figure 2 has the same setup as Fig. 1 and shows data and MOST functions of \({C_q}^2; f_{{C_q}^2}\) includes more nearneutral values than \(f_{{C_T}^2}\) because of the stricter nearneutral filtering we adopted for \({C_T}^2\) to avoid the dominance of the day–night transition singularity of the \({C_T}^2\) dimensionless group. In Sect. 2.1 we discussed why \({C_q}^2\) is more vulnerable to nonlocal effects than \({C_T}^2\) and that filtering out these effects is more critical for \({C_q}^2\). To ensure that nonlocal effects are brought to a minimum for \({C_q}^2\) we excluded two datasets from the overall \(f_{{C_q}^2}\) fit, namely Omseson and SudMed. These datasets were gathered over irrigated fields in a desert environment, and are therefore vulnerable to advection induced by thermal and moisture heterogeneities. The Bowen ratio classifications in Table 1 indicate that these two datasets indeed stand out by their heterogeneous surface.
In Fig. 2 we observe that, in contrast to \({C_T}^2\), for unstable (nearneutral) conditions the dimensionless group of \({C_q}^2\) does not go to infinity. We attribute this difference to the fact that \(L_vE\) mostly remains positive under stable conditions and does not go through zero as does H. Where the term \(\theta _{*}\) goes to zero during transitions from stable to unstable, the term \(q_*\) remains large and the dimensionless group does not go to infinity. For this reason \(f_{{C_q}^2}\) has a welldefined neutral limit, which is lower than that of \(f_{{C_T}^2}\).
We emphasize that by removing the Omseson and SudMed datasets, the amount of scatter has been reduced substantially, thereby decreasing the uncertainty of \(f_{{C_q}^2}\). To the contrary, removing these datasets for the overall \({{C_T}^2}\) function did not change the amount of scatter in the data points of \(f_{{C_T}^2}\) (not shown here). This indicates that \({{C_q}^2}\) is indeed more vulnerable to nonlocal effects than \({{C_T}^2}\). Our result are in line with other reported results; the fact that most of the scatter originated from the datasets with the largest likelihood of being influenced by nonlocal effects, indicates that nonlocal influences can indeed affect the shape of \(f_{{C_q}^2}\). Li et al. (2012) showed that from weakly unstable to neutral conditions, \(R_{Tq}\) decreases progressively from near one to near zero, reasoning that this was related to nonlocal effects and derived a dissimilarity ratio that scales with \(R_{Tq}\). Maronga (2014) showed with his model that when entrainment was sufficiently small, the MOST functions of \(f_{{C_T}^2}\) and \(f_{{C_q}^2}\) were identical. However, when entrainment was included he found dissimilarity between \(f_{{C_T}^2}\) and \(f_{{C_q}^2}\). We tried to parametrize nonlocal effects with \(R_{Tq}\), the skewness of q and also the Bowen ratio, but we could not relate any of these parameters to the dissimilarity between \({{C_T}^2}\) and \({{C_q}^2}\). We reason that we did not find a relation because we already minimized the influence of nonlocal effects with our data treatment. Following this argument we cannot ascribe the dissimilarity between \(f_{{C_q}^2}\) and \(f_{{C_T}^2}\) to nonlocal effects per se. What remains is true dissimilarity, i.e., nonequal turbulent Prandtl and Schmidt numbers.
4.3 Overall MOST Functions for \(\varepsilon \)
4.4 Variability of MOST Functions Between Datasets
MOST functions for \({C_T}^2, {C_q}^2\) and \(\varepsilon \) that were fitted using data from individual datasets are presented in Fig. 4 and the corresponding coefficients are given in Online Appendix. Some datasets only consisted of a limited number of data, and those with less than 90 data points were therefore removed.
The top left plot of Fig. 4 shows that the neutral limit of \(f_{{C_T}^2}\) under unstable conditions ranges between 4.5 (SudMed—less dense) and 7.6 (LITFASS’03—colza). Several datasets have a neutral limit close to 4.9, as found by Wyngaard et al. (1971) and De Bruin et al. (1993). Besides variations in the neutral limit we find variations under convective conditions as well; \(c_1c_2^{2/3}\) ranges from 1.11 (SudMed—less dense) to 3.65 (Omseson).
The neutral limit of \(f_{{C_q}^2}\) under unstable conditions ranges between 2.3 (Omseson) and 4.7 (Transregio’08—sugarbeet). All datasets have a lower neutral limit for \(f_{{C_q}^2}\) compared to \(f_{{C_T}^2}\), which we reasoned to be an effect of the dimensionless group of \({C_q}^2\) that does not go to infinity, see Sect. 4.2. Moreover, comparison of the point cloud of \(f_{{C_q}^2}\) and \(f_{{C_T}^2}\) (unstable conditions) in Fig. 4 shows that the scatter is larger for \(f_{{C_q}^2}\), which implies that \({C_q}^2\) is more sensitive to nonlocal effects than \({C_T}^2\), as discussed in Sect. 2.1.
For stable conditions, for both \(f_{{C_T}^2}\) (upper right plot) and \(f_{{C_q}^2}\) (middle right plot), the largest differences exist for very stable conditions. Note that variability towards very stable conditions causes only small variations in the absolute values of the fluxes, as the fluxes are small under very stable conditions. The neutral limits for MOST functions of \(\varepsilon \) vary from 0.60 (SudMed—less dense and Transregio’08—sugarbeet) to 1.28 (BLLAST—grass) taking together the unstable and stable functions. Overall, MOST functions from all datasets imply that production by shear and dissipation of TKE are not in balance, as has been observed in previous studies.
Furthermore, we aim to identify plausible causes for variations of MOST functions in the literature. The potential causes that we are able to investigate here are: (1) the accuracy of z measurements, (2) a limited z/L range covered by the data, (3) a limited number of data points, (4) environmental influences such as surface type and humidity, and (5) nonlocal effects. Other factors that may affect the data are variable instrumentation and processing techniques, where the latter was covered by Braam et al. (2014), but which we are not able to identify in this study. For the analysis below we use data from the online Appendix, which includes \(c_1, c_2\) and \(c_1c_2^{2/3}\) for all individual fits, the number of data points and the z/L range of the data.
4.4.1 Accuracy of z Measurements
We observed in the analysis that when calculations were made with inaccurate values of z, the MOST functions fall outside the range of data points from other datasets. For example, the LITFASS’12—rye dataset was gathered in a period during which the crop was in its growing stage. Calculating the dimensionless groups with the average crop height over that period resulted in more scatter than when the actual crop height was used. Part of the variation that we find in Fig. 4 may therefore be a result of inaccurate z measurements. This may especially be the case for the SudMed datasets, which were gathered over an olive grove where the olive trees were separated by open spaces. The uncertainty in the effective displacement height over these open spaces induces an uncertainty in the further analysis. This height uncertainty would explain why the neutral limit of this dataset deviates from other datasets for all functions \(f_{{C_T}^2}, f_{{C_q}^2}\) and \(f_{{\varepsilon }}\). The same holds for the free convection parameter \(c_1c_2^{2/3}\) of the functions \(f_{{C_T}^2}\) and \(f_{{C_q}^2}\). This observation of deviating MOST functions with inaccurate z measurements is in line with Hartogensis et al. (2003) who did a sensitivity analysis of all parameters that link \({C_T}^2\) to H and showed that z is indeed the most sensitive parameter.
4.4.2 Limited z/L Range and Limited Number of Data Points Covered by the Data
When MOST functions are determined over a short z/L range or with a limited number of data points, they have a larger uncertainty and result in an inaccurate or invalid MOST function. For example, for \(f_{{C_T}^2}\) the LITFASS’12—rye dataset consists of the largest amount of data points and has the smallest uncertainty range for \(c_1\) for both stable and unstable conditions. In contrast, the Omseson dataset only covers a short z/L range and has a limited number of data points, leading to a function with higher uncertainty, which is in this case visible as a higher neutral limit (7.0) of \(f_{{C_T}^2}\) for unstable conditions and a value for \(c_1c_2^{2/3}\) (3.65) far from that found for other datasets. The same applies to the BLLAST—grass dataset that has a neutral limit of 1.28 for \(f_{{\varepsilon }}\) under unstable conditions, which is substantially higher than the neutral limit of other datasets. Note also that we already excluded a few datasets with less than 90 data points from Fig. 4, because we could not determine a proper function for these datasets. The argumentation about a limited number of data points is therefore not limited to the examples of datasets that we give here.
4.4.3 Surface Variability
We could not find a relation between the MOST functions and environmental influences such as the surface type (e.g. vegetated or bare soil) or humidity of the site, which suggests that the MOST functions of \({{C_T}^2}, {{C_q}^2}\) and \({\varepsilon }\) can be used across different sites. For \(f_{{C_q}^2}\) we observed differences that are likely related to nonlocal effects induced by thermal and moisture heterogeneities of the surface. We discuss this below.
4.4.4 Nonlocal Effects
The Omseson and SudMed datasets were gathered over irrigated areas within very dry environments. This large thermal and moisture heterogeneity makes these datasets highly vulnerable to advection (Hoedjes et al. 2002, 2007). Figure 4 shows that it is also these two datasets for which the data points of \(f_{{C_q}^2}\) (unstable) are below that of other datasets. In Sect. 4.2 we showed that, when the Omseson and SudMed datasets were excluded, the scatter of the point cloud for \(f_{{C_q}^2}\) was reduced, but not for \(f_{{C_T}^2}\). This indicates that \({{C_T}^2}\) is less vulnerable to nonlocal effects than \({{C_q}^2}\).
The variation between functions from all individual datasets in Fig. 4 is of the same size as the variation between functions from the literature (Figs. 1, 2 and 3). We have now seen that MOST relations that are based on individual datasets of limited length can lead to biased results due to the limited number of data points, limited stability range, or nonlocal effects specific to that location. If the datasets that we marked as uncertain are removed from Fig. 4 (e.g. SudMed for its uncertainty in the displacement height, Omseson and BLLAST—grass for their short z/L range, Omseson for advection), this reduces the variation between the MOST functions to a smaller range than the variation between literature functions. Based on this analysis we can say that it is not sufficient to determine MOST functions from only one dataset.
4.5 SelfCorrelation
When applying MOST we have to be aware of selfcorrelation as \(\theta _{*}\) and \(u_*\) are included on both x and yaxes for \(f_{{C_T}^2}\) and \(f_{\varepsilon }\) respectively (Hicks 1978; Baas et al. 2006). When selfcorrelation dominates, we expect to observe a relation between the dimensionless group of \({C_T}^2\) and z/L when random values of \({C_T}^ 2\) are used to calculate the dimensionless group on the yaxis and z/L. The same holds for a relation between the dimensionless group of \({\varepsilon }\) and z/L when these groups are calculated with random values of \(\varepsilon \).
4.6 Sensitivity of Surface Fluxes to Varying MOST Functions
The bars in Fig. 6 express the adequacy of our MOST functions in capturing the relation between turbulence parameters and fluxes compared to measured fluxes. The fact that the fluxes from the eddycovariance method and MOST functions are not exactly the same indicates that the base functions for the MOST relations have difficulty covering all the variability over the wide z/L range. From the top left plot of Fig. 6 we can see that the relative difference between \(H_{\textit{MOST}}\) and \(H_{\textit{EC}}\) is more distinctive towards neutral stability (26 % difference) compared to free convection conditions (7 % difference). As the large difference only exists for small fluxes, the absolute difference between \(H_{\textit{MOST}}\) and \(H_{\textit{EC}}\) is small (see the absolute difference in the right plot). Also the difference between \(H_{\textit{MOST}}\) from the uncertainty range is small in absolute values (indicated by the error bars). The error bar in the left plot shows that the difference between the median from the upper and lower MOST function is 6 % for an average value of 315 W m\(^{2}\), and absolute difference of 17 W m\(^{2}\). For \(L_{v}E\) and \(\tau \) the relative differences are 4 % and 9 % at an average value of 351 W m\(^{2}\) and 0.18 N m\(^{2}\) respectively, and absolute differences of 13 W m\(^{2}\) for \(L_{v}E\) and 0.02 N m\(^{2}\) for \(\tau \). For stable conditions we find relative differences of the same order of magnitude as for unstable conditions, i.e., 4 % for both H and \(L_vE\) at an average flux of \(40\) W m\(^{2}\) for H and 48 W m\(^{2}\) for \(L_{v}E\). The relative differences that we find here are relatively small because of the narrow uncertainty of the overall MOST functions. However, the relative difference is larger when the MOST function variability from the literature is used.
Hartogensis et al. (2003) showed that the uncertainty of literature functions led to a 20 % difference in H in the free convection range. We have now brought this uncertainty down to 6 %, which is based on flux calculations with the uncertainty of MOST functions that we presented in Table 2. Although the errors that we find appear to be small, they are in the same order of magnitude as error estimates of other variables in flux calculations and should therefore not be ignored. It is already generally accepted that when fluxes are calculated, they should include an error estimate of the input parameters such as \({{C_T}^2}, {{C_q}^2}\) and \(\varepsilon \). For future studies we now recommend adding an additional uncertainty in the parameters \(c_1\) and \(c_2\). We suggest calculating fluxes with the overall functions that we presented in Table 2 and to calculate errors in these fluxes using the uncertainty range of MOST functions that were given in the same table.
5 Conclusions
Our main goal was to define robust MOST functions for \({C_T}^2, {C_q}^2\) and \(\varepsilon \), for which we used 11 field experiments carried out over various surface types and under various stability ranges. The data filtering used ensured that MOST assumptions are met. We combined all datasets to derive the MOST functions of \({{C_T}^2}, {{C_q}^2}\) and \({\varepsilon }\), with the resulting MOST functions summarized in Table 2. The MOST function for \({{C_T}^2}\) that we found for unstable conditions compares well with the MOST function that we derived through the Businger fluxprofile relations. The fact that these functions are similar gives confidence in the overall results. A difference between the dimensionless groups of \({{C_T}^2}\) and \({{C_q}^2}\) is observed for nearneutral conditions. That is, the dimensionless group of \({{C_T}^2}\) tends towards infinity for nearneutral conditions, in contrast to that of \({{C_q}^2}\). This difference is primarily driven by the unstable to stable transition. We observe that \({{C_q}^2}\), and thereby also \(f_{{C_q}^2}\), is more vulnerable to nonlocal effects than \({{C_T}^2}\). We investigated a parametrization of nonlocal effects in the combined dataset with \(R_{Tq}\), the skewness of q and also the Bowen ratio, but we could not relate any of these parameters to the dissimilarity between \({{C_T}^2}\) and \({{C_q}^2}\). We reason that this is because we have already minimized the influence of nonlocal effects with the data treatment. For \(\varepsilon \) we found a neutral limit well below 1, which implies that production by shear and dissipation of TKE are not in balance, as found in previous studies.
We also wished to identify if MOST functions are similar for different datasets, which is an indication of whether MOST functions are generally applicable across different sites and environmental conditions. As the effect of variable instrumentation and data treatment is eliminated, we were able to compare MOST functions for different datasets. This analysis showed that functions, determined over a short stability range, with a limited number of data points, or with inaccurate z measurements, have larger uncertainty and therefore can deviate from the overall MOST function. Hence, we conclude that these are likely reasons for variations of MOST functions in the literature in addition to variable instrumentation, and algorithms to estimate fluxes and \({C_T}^2, {C_q}^2\) and \(\varepsilon \). For MOST functions of \({C_T}^2\) and \(\varepsilon \) we found no environmental influence; for example, field experiments over a dry or moist surface did not result in significantly different MOST functions, which implies that MOST functions of \({C_T}^2\) and \(\varepsilon \) are indeed generally applicable. MOST functions of \({C_q}^2\) differed between datasets, which was related to surface heterogeneity, indicating that nonlocal effects likely have an influence on these datasets. Based on this we can say that it is not sufficient to determine MOST functions based on only one dataset, as has been done in previous studies.
We encourage further research that focuses on indicators for MOST violation and to develop a parametrization to correct for these conditions. A potential powerful approach is that of Maronga (2014) and Maronga et al. (2014) using LES modelling. Such an approach allows the quantification of the effect of nonlocal conditions on MOST relations. These studies can potentially provide MOST functions that correct for, e.g., the influence of nonlocal effects, such that MOST functions of \({C_q}^2\) can be applied under all conditions.
Lastly, we demonstrated the sensitivity of \(H, L_vE\) and \(\tau \) to the uncertainty of the overall MOST functions that we determined using subsample regressions. The errors caused by different MOST functions are in the same order of magnitude as error estimates for other variables in the flux calculations. We therefore recommend inclusion of the uncertainty of MOST functions into the uncertainty calculations of surface fluxes. Furthermore, to reach consensus on the MOST functions to be used for flux calculations we recommend using the MOST functions and their uncertainty range that we determined and that are summarized in Table 2.
Footnotes
 1.
Appendices 1 and 2 can be found in webbased supplementary material.
Notes
Acknowledgments
The authors thank Miranda Braam, Bram van Kesteren and Arnold Moene for their valuable suggestions. We also thank the reviewers for their suggestions and critical comments, which helped to improve the manuscript. Furthermore, authors of papers in which the data have been presented are thanked for their contribution in collecting the data. Travel to the Tübingen Atmospheric Physics Symposium “Scintillometers and Applications” was supported by the Dutch Technology Foundation (STW).
Supplementary material
References
 Andreas EL (1988) Estimating \(C_n^2\) over snow and sea ice from meteorological data. J Opt Soc Am 5(4):481–495CrossRefGoogle Scholar
 Andreas EL (1989) Twowavelength method of measuring pathaveraged turbulent surface heat fluxes. J Atmos Ocean Technol 6:280–292CrossRefGoogle Scholar
 Baas P, Steeneveld GJ, Van De Wiel BJH, Holtslag AAM (2006) Exploring selfcorrelation in fluxgradient relationships for stably stratified conditions. J Atmos Sci 63:3045–3054CrossRefGoogle Scholar
 Beljaars ACM, Holtslag AAM (1991) Flux parameterization over land surfaces for atmospheric models. J Appl Meteorol 30:327–340CrossRefGoogle Scholar
 Beyrich F, Mengelkamp HT (2006) Evaporation over a heterogeneous land surface: EVA GRIPS and the LITFASS2003 experiment: an overview. BoundaryLayer Meteorol 121:5–32CrossRefGoogle Scholar
 Beyrich F, Leps JP, Mauder M, Bange J, Foken T, Huneke S, Lohse H, Ldi A, Meijninger WML, Mirnonov D, Weisensee U, Zittel P (2006) Areaaveraged surface fluxes over the LITFASS region on eddycovariance measurements. BoundaryLayer Meteorol 121:33–65CrossRefGoogle Scholar
 Beyrich F, Bange J, Hartogensis OK, Raasch S, Braam M, Van Dinther D, Gräf D, Van Kesteren B, Van Den Kroonenberg AC, Maronga B, Martin S, Moene AF (2012) Towards a validation of scintillometer measurements: the LITFASS2009 experiment. BoundaryLayer Meteorol 144:83–112CrossRefGoogle Scholar
 BouZeid E, Higgins C, Huwald H, Meneveau C, Parlange MB (2010) Field study of the dynamics and modelling of subgridscale turbulence in a stable atmospheric surface layer over a glacier. J Fluid Mech 665:480–515CrossRefGoogle Scholar
 Braam M, Moene AF, Beyrich F, Holtslag AAM (2014) Similarity relations for \(C_T^2\) in the unstable atmospheric surface layer: dependence on regression approach, observation height and stability range. BoundaryLayer Meteorol 153:63–87CrossRefGoogle Scholar
 Businger JA, Wyngaard JC, Izumi Y, Bradley EF (1971) Fluxprofile relationships in the atmospheric surface layer. J Atmos Sci 28:181–189CrossRefGoogle Scholar
 Crewell S, Bloemink H, Feijt A, García SG, Jolivet D, Krasnow OA, Van Lammeren A, Löhnert U, Van Meijgaard E, Meywerk J, Quante M, Pfeilsticker K, Schmidt S, Scholl T, Simmer C, Schröder M, Trautmann T, Venema V, Wendisch M, Willèn U (2004) The BALTEX Bridge Campaign: an integrated approach for a better understanding of clouds. Bull Am Meteor Soc 85(10):1565–1584CrossRefGoogle Scholar
 De Bruin HAR, Kohsiek W, Van Den Hurk BJJM (1993) A verification of some methods to determine the fluxes of momentum, sensible heat, and water vapour using standard deviation and structure parameter of scalar meteorological quantities. BoundaryLayer Meteorol 63(3):231–257CrossRefGoogle Scholar
 De Bruin HAR, Van den Hurk BJJM, Kohsiek W (1995) The scintillation method tested over a dry vineyard area. BoundaryLayer Meteorol 76:25–40CrossRefGoogle Scholar
 De Bruin HAR, Van Den Hurk BJJM, Kroon LJM (1999) On the temperaturehumidity correlation and similarity. BoundaryLayer Meteorol 93(3):453–468CrossRefGoogle Scholar
 Finkelstein PL, Sims PF (2001) Sampling error in eddy correlation flux measurements. J Geophys Res 106(D4):3503–3509CrossRefGoogle Scholar
 Frenzen P, Vogel CA (2001) Further studies of atmospheric turbulence in layers near the surface: scaling the TKE budget above the roughness sublayer. BoundaryLayer Meteorol 99(2):173–206CrossRefGoogle Scholar
 Graf A, Schuttemeyer D, Geiss H, Knaps A, MollmannCoers M, Schween JH, Kollet S, Neininger B, Herbst M, Vereecken H (2010) Boundedness of turbulent temperature probability distributions, and their relation to the vertical profile in the convective boundary layer. BoundaryLayer Meteorol 134(3):459–486CrossRefGoogle Scholar
 Hartogensis OK, De Bruin HAR, De Van, Wiel BJH (2002) Displacedbeam small aperture scintillometer test. Part II: CASES99 stable boundarylayer experiment. BoundaryLayer Meteorol 105(1):149–176CrossRefGoogle Scholar
 Hartogensis OK, Watts CJ, Rodriquez JC, De Bruin HAR (2003) Derivation of an effective height for scintillometers: La Poza experiment in Northwest Mexico. J Hydrometeorol 105(1):149–176Google Scholar
 Hartogensis OK, De Bruin HAR (2005) MoninObukhov similarity functions of the structure parameter of temperature and turbulent kinetic energy dissipation rate in the stable boundary layer. J Hydrometeorol 4:915–928CrossRefGoogle Scholar
 Hartogensis OK (2006) Exploring scintillometry in the stable atmospheric surface layer. Wageningen University, Wageningen 227 ppGoogle Scholar
 Hicks BB (1978) Some limitations of dimensional analysis and power laws. BoundaryLayer Meteorol 14:567–569CrossRefGoogle Scholar
 Hill RJ, Ochs GR, Wilson JJ (1992) Measuring surfacelayer fluxes of heat and momentum using optical scintillation. BoundaryLayer Meteorol 58(4):391–408CrossRefGoogle Scholar
 Hill RJ (1997) Algorithms for obtaining atmospheric surfacelayer fluxes from scintillation measurements. J Atmos Ocean Technol 14(3):456–467CrossRefGoogle Scholar
 Hoedjes JCB, Zuurbier RM, Watts CJ (2002) Large aperture scintillometer used over a homogeneous irrigated area, partly affected by regional advection. BoundaryLayer Meteorol 105:99–117CrossRefGoogle Scholar
 Hoedjes JCB, Chehbouni A, Ezzahar J, Escadafal R, De Bruin HAR (2007) Comparison of large aperture scintillometer and eddy covariance measurements: can thermal infrared data be used to capture footprintinduced differences? J Hydrometeorol 8:144–159CrossRefGoogle Scholar
 Högström U (1988) Nondimensional wind and temperature profiles in the atmospheric surface layer: a reevaluation. BoundaryLayer Meteorol 42:55–78CrossRefGoogle Scholar
 Högström U (1990) Analysis of turbulence structure in the surface layer with a modified similarity formulation for near neutral conditions. J Atmos Sci 47(16):1949–1972CrossRefGoogle Scholar
 Högström U (1996) Review of some basic characteristics of the atmospheric surface layer. BoundaryLayer Meteorol 78:215–246CrossRefGoogle Scholar
 Kaimal JC, Finnigan JJ (1994) Atmospheric boundary layer flows—their structure and measurement. Oxford University Press, New York, USA 289 pGoogle Scholar
 Kanda M, Moriwaki R, Roth M, Oke T (2002) Areaaveraged sensible heat flux and a new method to determine zeroplane displacement length over an urban surface using scintillometry. BoundaryLayer Meteorol 105:177–193CrossRefGoogle Scholar
 Katul GG, Sempreviva AM, Cava D (2008) The temperaturehumidity covariance in the marine surface layer: a onedimensional analytical model. BoundaryLayer Meteorol 126(2):263–278CrossRefGoogle Scholar
 Katul GG, Konings AG, Porporato A (2011) Mean velocity profile in a sheared and thermally stratified atmospheric boundary layer. Phys Rev Lett 107(26):268502CrossRefGoogle Scholar
 Katul GG, Li D, Chamecki M, BouZeid E (2013) Mean scalar concentration profile in a sheared and thermally stratified atmospheric surface layer. Phys Rev E 87(2):023004CrossRefGoogle Scholar
 Kohsiek W (1982) Measuring \({C_T}^2, {C_q}^2\), and \({C_{Tq}}\) in the unstable surface layer, and relations to the vertical fluxes of heat and moisture. BoundaryLayer Meteorol 24:89–107CrossRefGoogle Scholar
 Kolmogorov AN (1941) The local structure of turbulence in an incompressible viscous fluid for very large Reynolds numbers. Dokl Akad Nauk 30:299–303Google Scholar
 Li D, BouZeid E, De Bruin HAR (2012) MoninObukhov similarity functions for the structure parameters of temperature and humidity. BoundaryLayer Meteorol 145:45–67CrossRefGoogle Scholar
 Lothon M, Lohou F, Pino D, Couvreux F, Pardyjak ER, Reuder J, VilàGuerau de Arellano J, Durand P, Hartogensis O, Legain D, Augustin P, Gioli B, Faloona I, Yagüe C, Alexander DC, Angevine WM, Bargain E, Barrié J, Bazile E, Bezombes Y, BlayCarreras E, van de Boer A, Boichard JL, Bourdon A, Butet A, Campistron B, de Coster O, Cuxart J, Dabas A, Darbieu C, Deboudt K, Delbarre H, Derrien S, Flament P, Fourmentin M, Garai A, Gibert F, Graf A, Groebner J, Guichard F, Jimenez Cortes MA, Jonassen M, van den Kroonenberg A, Lenschow DH, Magliulo V, Martin S, Martinez D, Mastrorillo L, Moene AF, Molinos F, Moulin E, Pietersen HP, Piguet B, Pique E, RománCascón C, RufinSoler C, Saïd F, SastreMarugán M, Seity Y, Steeneveld GJ, Toscano P, Traullé O, Tzanos D, Wacker S, Wildmann N, Zaldei A (2014) The BLLAST field experiment: BoundaryLayer Late Afternoon and Sunset Turbulence. Atmos Chem Phys 14(20):10931–10960CrossRefGoogle Scholar
 Lüdi A, Beyrich F, Mätzler C (2005) Determination of the turbulent temperaturehumidity correlation from scintillometric measurements. BoundaryLayer Meteorol 117(3):525–550CrossRefGoogle Scholar
 Mahrt L (1999) Stratified atmospheric boundary layers. BoundaryLayer Meteorol 90:375–396CrossRefGoogle Scholar
 Maronga B (2014) MoninObukhov similarity functions for the structure parameters of temperature and humidity in the unstable surface layer: Results from highresolution largeeddy simulations. J Atmos Sci 71:716–733Google Scholar
 Maronga B, Hartogensis OK, Raasch S, Beyrich F (2014) The effect of surface heterogeneity on the structure parameters of temperature and specific humidity: a largeeddy simulation case for the LITFASS2003 experiment. BoundaryLayer Meteorol 153:441–470CrossRefGoogle Scholar
 Mauder M, Foken T (2004) Documentation and instruction manual of the eddy covariance software package TK2 Arbeitsergebn. University of Bayreuth, Abt Mikrometeorol, vol. 26, p 42. ISSN 1614–8916Google Scholar
 Mauder M, Cuntz M, Drüe C, Graf A, Rebmann C, Schmid HP, Schmidt M, Steinbrecher R (2013) A strategy for quality and uncertainty assessment of longterm eddycovariance measurements. Agric For Meteorol 169:122–135CrossRefGoogle Scholar
 Meijninger WML, Beyrich F, Luedi A, Kohsiek W, De Bruin HAR (2006) Scintillometerbased turbulent fluxes of sensible and latent heat over a heterogeneous land surface: a contribution to LITFASS2003. BoundaryLayer Meteorol 121(1):89–110CrossRefGoogle Scholar
 Moncrieff JB, Massheder JM, De Bruin HAR, Ebers J, Friborg T, Heusinkveld B, Kabat P, Scott S, Soegaard H, Verhoef A (1997) A system to measure surface fluxes of momentum, sensible heat, water vapor and carbon dioxide. J Hydrol 188–189:589–611Google Scholar
 Monin AS, Obukhov AM (1954) Osnovnye zakonomernosti turbulentnogo peremesivanija v prizemnom sloe atmosfery (Basic laws of turbulent mixing in the atmosphere near the ground). Trudy geofiz inst AN SSSR 151:163–187Google Scholar
 Monin AS, Yaglom AM (1975) Statistical fluid mechanics of turbulence, vol 2. MIT Press, Cambridge, 874 ppGoogle Scholar
 Obukhov AM (1946) Turbulentnost’ v temperaturnoj  neodnorodnoj atmosfere (Turbulence in an atmosphere with a nonuniform temperature). Trudy Inst Theor Geofiz AN SSSR 1:95–115Google Scholar
 Obukhov AM (1960) O strukture temperaturnogo polja i polja skorostej v uslovijach konvekcii (Structure of the temperature and velocity fields under conditions of free convection). Izv AN SSSR, ser Geofiz, pp 1392–1396Google Scholar
 Obukhov AM (1971) Turbulence in an atmosphere with a nonuniform temperature. BoundaryLayer Meteorol 2:7–29CrossRefGoogle Scholar
 Pahlow M, Parlange MB, PortéAgel F (2001) On MoninObukhov similarity in the stable atmospheric boundary layer. BoundaryLayer Meteorol 99(2):225–248CrossRefGoogle Scholar
 Petrás I, Bednárová D (2010) Total least squares approach to modeling: a Matlab toolbox. Acta Montan Slovaca 15(2):158–170Google Scholar
 Renno NO, Abreu VJ, Koch J, Smith PH, Hartogensis OK, De Bruin HAR, Burose D, Delory GT, Farrell WM, Watts CJ, Garatuza J, Parker M, Carswell A (2004) MATADOR 2002: a pilot field experiment on convective plumes and dust devils. J Geophys Res 109:E07001CrossRefGoogle Scholar
 Roth M, Salmond JA, Satyanarayana ANV (2006) Methodological considerations regarding the measurement of turbulent fluxes in the urban roughness sublayer: the role of scintillometry. BoundaryLayer Meteorol 121:351–375CrossRefGoogle Scholar
 Stull RB (1988) An introduction to boundary layer meteorology. Kluwer academic publishers, Dordrecht, 666 ppGoogle Scholar
 Thiermann V, Grassl H (1992) The measurement of turbulent surfacelayer fluxes by use of bichromatic scintillation. BoundaryLayer Meteorol 58(4):367–389CrossRefGoogle Scholar
 Van de Boer A, Moene AF, Graf AF, Schüttemeyer D, Simmer C (2014) Detection of entrainment influences on surfacelayer measurements and extension of Monin–Obukhov similarity theory. BoundaryLayer Meteorol 152(3):19–44CrossRefGoogle Scholar
 Van de Wiel BJH, Ronda RJ, Moene AF, De Bruin HAR, Holtslag AAM (2002) Intermittent turbulence and oscillations in the stable boundary layer over land. Part I: a bulk model. J Atmos Sci 59:942–958CrossRefGoogle Scholar
 Van Kesteren B, Hartogensis OK (2010) Analysis of the systematic errors found in the Kipp & Zonen largeaperture scintillometer. BoundaryLayer Meteorol 138(3):493–509CrossRefGoogle Scholar
 Van Kesteren B, Hartogensis OK, Van Dinther D, Moene AF, De Bruin HAR (2013) Measuring \({\rm H}_2{\rm O}\) and \({\rm CO}_2\) fluxes at field scales with scintillometry: part I—introduction and validation of four methods. Agric For Meteorol 178–179:75–87CrossRefGoogle Scholar
 Vickers D, Mahrt L (1997) Quality control and flux sampling problems for tower and aircraft data. J Atmos Ocean Technol 14:512–526CrossRefGoogle Scholar
 Wilczak JM, Oncley SP, Stage SA (2001) Sonic anemometer tilt correction algorithms. BoundaryLayer Meteorol 99:127–150CrossRefGoogle Scholar
 Wyngaard JC, Coté OR (1971) The budgets of turbulent kinematic energy and temperature variance in the atmospheric surface layer. J Atmos Sci 28:190–201CrossRefGoogle Scholar
 Wyngaard JC, Izumi Y, Collins SA Jr (1971) Behaviour of the refractive index structure parameter near the ground. J Opt Soc Am 15:1177–1188Google Scholar
 Wyngaard JC (1973) On surface layer turbulence. In: Haugen DA (ed) Workshop on micrometeorology. American Meteorological Society, Boston, pp 101–149Google Scholar
Copyright information
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.