The separation distance between the two EC systems is less than 10 m, and thus they are expected to measure the same source area outside the flow distortion areas. At the same time the observed differences cannot arise from the post-processing as fluxes were calculated and processed in a similar manner. Some of the difference can still origi- nate from instrument drifting, but this would indicate non- directional dependence. As a result, the differences in the fluxes measured by the two systems very probably relate to the variation of the flux field caused by complex terrain. In past studies above vegetated ecosystems, the random uncer- tainty of flux measurements resulting from instrumental er- rors, heterogeneity of the surface and turbulence has been de- termined using the so-called two-tower approach (Hollinger and Richardson, 2005; Kessomkiat et al., 2010). Its assump- tion is that the two time series should be independent from each other and thus cannot be used in our case when the two systems are measuring the same footprint. We can however still calculate the RRE in order to get an understanding about the random uncertainties of our EC measurements. Of all studied vertical fluxes, the largest random uncertainties re- late to τ (medians between 23 % and 28 %) and the lowest to daytime H (medians 12 % and 13 %) (Fig. 6). For τ no systematic pattern between daytime and night-time is seen, whereas for the other fluxes nocturnal uncertainties tend to be larger when the scalar fluxes are small. For fluxes other than momentum, RREs from EC2 are slightly larger than those from EC1, whereas for τ it is vice versa. The RREs are of the same order of magnitude as observed at the semi-urban site in Kumpula and above vegetated ecosystems. In these, however, the RRE associated with τ tends to be the lowest contrary to our study (Finkelstein and Sims, 2001; Billes- bach, 2011; Nordbo et al., 2012b), which is because of the complex measurement location and source–sink distribution at our site.
18 Read more
Abstract. The energy balance of eddy-covariance (EC) mea- surements is typically not closed, resulting in one of the main challenges in evaluating and interpreting EC flux data. En- ergy balance closure (EBC) is crucial for validating and im- proving regional and global climate models. To investigate the nature of the gap in EBC for agroecosystems, we an- alyzed EC measurements from two climatically contrasting regions (Kraichgau – KR – and Swabian Jura – SJ) in south- western Germany. Data were taken at six fully equipped EC sites from 2010 to 2017. The gap in EBC was quantified by ordinary linear regression, relating the energy balance ra- tio (EBR), calculated as the quotient of turbulent fluxes and available energy, to the residual energy term. In order to ex- amine potential reasons for differences in EBC, we compared the EBC under varying environmental conditions and inves- tigated a wide range of possible controls. Overall, the varia- tion in EBC was found to be higher during winter than sum- mer. Moreover, we determined that the site had a statistically significant effect on EBC but no significant effect on either crop or region (KR vs SJ). The time-variable footprints of all EC stations were estimated based on data measured in 2015, complimented by micro-topographic analyses along the pre-
20 Read more
Even in the hypothetical situation of perfectly synchro- nized timestamps for wind and gas data, if the respective instruments’ sampling volumes have to be spatially sepa- rated to avoid presently intractable flow distortion issues in the anemometer, as is notably the case with open-path setups (see, for example, Wyngaard, 1988; Frank et al., 2016; Grare et al., 2016; Horst et al., 2016; and Huq et al., 2017), the corresponding time series will be affected by misalignment, possibly to varying degrees. Indeed, assuming the validity of Taylor’s hypothesis of frozen turbulence, wind and concen- tration data will be affected by a time lag (the time air takes to travel between the two sampling volumes), which will be further modulated by wind intensity and direction. Addition- ally, modification of turbulence structure intervening while air parcels transit through the dislocated instrument volumes may introduce further uncertainty in flux estimates (Cheng et al., 2017). In the case of co-located sensors (e.g., Hydra-IV, CEH; IRGASON, Campbell Scientific Inc.) this problem is not present but is replaced by the flow distortion issues men- tioned above and not addressed in the present study.
15 Read more
Before validating BETHY/DLR’s modelled NPP for NUTS- 1 regions across all of Germany, we performed a cross- check of BETHY/DLR GPP results with eddy covariance flux tower measurements provided by FLUXNET. Two tower sites were selected, one in the Hainich forest and one in the Tharandt forest. The Hainich tower is to the west of Jena, Germany, in a mixed deciduous beech forest, while the Tha- randt tower is south of Dresden in a coniferous forest. For both sites Level 4 data, providing information about GPP, are available for 2000 and 2001. GPP was calculated by sub- tracting the estimated ecosystem respiration from measured NEP) as described in Reichstein et al. (2005). These data were crosschecked against BETHY/DLR’s modelled GPP, as annual sums at each station (Table 3).
16 Read more
Abstract. Evaporation (E) and transpiration (T ) respond dif- ferently to ongoing changes in climate, atmospheric compo- sition, and land use. It is difficult to partition ecosystem-scale evapotranspiration (ET) measurements into E and T , which makes it difficult to validate satellite data and land surface models. Here, we review current progress in partitioning E and T and provide a prospectus for how to improve the- ory and observations going forward. Recent advancements in analytical techniques create new opportunities for partition- ing E and T at the ecosystem scale, but their assumptions have yet to be fully tested. For example, many approaches to partition E and T rely on the notion that plant canopy conductance and ecosystem water use efficiency exhibit op- timal responses to atmospheric vapor pressure deficit (D). We use observations from 240 eddy covariance flux towers to demonstrate that optimal ecosystem response to D is a reasonable assumption, in agreement with recent studies, but more analysis is necessary to determine the conditions for which this assumption holds. Another critical assumption for many partitioning approaches is that ET can be approximated as T during ideal transpiring conditions, which has been challenged by observational studies. We demonstrate that T can exceed 95 % of ET from certain ecosystems, but other ecosystems do not appear to reach this value, which sug- gests that this assumption is ecosystem-dependent with im- plications for partitioning. It is important to further improve approaches for partitioning E and T , yet few multi-method comparisons have been undertaken to date. Advances in our understanding of carbon–water coupling at the stomatal, leaf, and canopy level open new perspectives on how to quantify T via its strong coupling with photosynthesis. Photosynthe- sis can be constrained at the ecosystem and global scales with emerging data sources including solar-induced fluores- cence, carbonyl sulfide flux measurements, thermography, and more. Such comparisons would improve our mechanistic understanding of ecosystem water fluxes and provide the ob- servations necessary to validate remote sensing algorithms and land surface models to understand the changing global water cycle.
29 Read more
Based on MATLAB 7.6.0 (R2008a, The MathWorks, Inc, USA) we developed a custom analysis routine that retains sufficient mass spectral information after filtering the raw data (150 000 bins per single spectrum) piecewise. To reduce the analysis time, the routine utilizes 6 min sum spectra to perform necessary mass spectrometric corrections. To solve the challenges described above, the data processing routine consists of three sub routines: the time-of-flight to m/z con- version routine calculates a statistically most accurate mass scale (Sect. 4.1), the peak detection routine generates a peak list of all detected mass peaks (Sect. 4.2), and the signal anal- ysis routine utilizes the peak list for mass peak fitting, mass peak analysis and 10 Hz stick spectra calculation (Sect. 4.3). These 10 Hz stick spectra were subsequently used for eddy covariance flux calculations.
et al., 2014; Billesbach et al., 2014; Commane et al., 2015), this is the first time that QCLAS drift was characterised un- der field conditions. The Allan variance plot obtained by the manufacturer under laboratory conditions (Aerodyne Re- search, 2017) indicates white noise up to ca. 8 s integration time, which is quite comparable with Fig. 2. At longer inte- gration times, however, the Allan variance plot by the man- ufacturer exhibits a much smaller increase with increasing averaging time. Most likely, this is the result of less stable thermal conditions in our instrument hut compared to labo- ratory conditions. This finding highlights the importance of optimising the experimental set-up (minimisation of temper- ature variations, insulation of QCLAS, etc.) for minimising sensor drift in the first place. In order to explore the effects of the drift on flux estimates, the following eddy covariance flux calculations were conducted for three high-pass filtering sce- narios commonly used in the literature: (i) block averaging (BA), (ii) linear detrending (LD), and (iii) recursive filtering (RF) with a time constant of 50 s as determined from Fig. 2. In addition, we followed Wehr et al. (2017) and removed the measured instrument offset by linear interpolation between half-hourly background measurements, termed linear back- ground correction (LBC). The latter approach assumes the linear interpolation of half-hourly background measurements (median absolute COS change equal to 20 ppt) to success- fully represent any sensor drift, while LD and RF may re- move real flux in case of true trends in the ambient concen- tration time series.
13 Read more
Besides data-driven approaches, combining the EC tech- nique with modeling approaches to reliably estimate the con- tribution of heterogeneously distributed sources and sinks of specific trace gases to the net flux can be achieved (Goeckede et al., 2006; Massman and Ibrom, 2008; Vesala et al., 2008), which will allow for deployments of EC towers in more com- plex environments than ever before (e.g., small-scale multi- crop systems as often used in organic farming). Furthermore, deployments of EC towers within or below the canopy as well as at ecosystem edges (Rogiers et al., 2005; Kirton et al., 2009) bear great potential in studying soil processes at the ecosystem scale. Still, each application in heterogeneous ter- rain involves complex micrometeorological conditions and thus requires a careful interpretation of the measured fluxes. Modeling activities may not only focus on the source distri- bution within a specific area but may also be carried one step further by simulating net trace gas emissions from the soil for areas where it is impossible to measure trace gas fluxes with experimental approaches. Such process-based biogeochemi- cal models (e.g., DAYSCENT, DNDC, and PaSim, to name only a few of them) can be validated in similar ecosystems with in situ eddy covariance flux measurements beforehand and then applied at the landscape scale.
19 Read more
Abstract. Fast response optical analyzers based on laser absorption spectroscopy are the preferred tools to measure field-scale mixing ratios and fluxes of a range of trace gases. Several state-of-the-art instruments have become commer- cially available and are gaining in popularity. This paper aims for a critical field evaluation and intercomparison of two compact, cryogen-free and fast response instruments: a quantum cascade laser based absorption spectrometer from Aerodyne Research, Inc., and an off-axis integrated cavity output spectrometer from Los Gatos Research, Inc. In this paper, both analyzers are characterized with respect to preci- sion, accuracy, response time and also their sensitivity to wa- ter vapour. The instruments were tested in a field campaign to assess their behaviour under various meteorological con- ditions. The instrument’s suitability for eddy covariance flux measurements was evaluated by applying an artificial flux of CH 4 generated above a managed grassland with otherwise
13 Read more
We include two different approaches to identify a subset of measurements per eddy-covariance site in which the ecosys- tem was unstressed and provide the results for both methods. A first approach is based on soil moisture levels. For those sites where soil moisture measurements were available, the maximal soil moisture level for each site was determined as the 98th percentile of all soil moisture measurements. We then split up the dataset of each site in five equal-size classes based on evaporation percentiles. For each class, days having soil moisture levels belonging to the highest 5 % of soil mois- ture levels within each class were selected as unstressed days, but only if the soil moisture level of these selected days was above 75 % of the maximum soil moisture. This guaranties the sampling of unstressed evaporation during all seasons.
24 Read more
second- to fourth-order moments in atmospheric measure- ments was derived rigorously by Lenschow et al. (2000) and applied to EC fluxes by Mauder et al. (2013). Lenschow et al. (2000) derived the method to estimate the instrumental random noise variance σ n, f 2 from the auto-covariance func- tion of the measured turbulent record close to zero-shift, en- abling us to determine the respective error for each half-hour flux averaging period under field conditions. In this study, the auto-covariance is linearly extrapolated to lag zero using the auto-covariance values at lags 1. . .5 (at 10 Hz frequency sampling rate) and the difference between this extrapolation and the observed auto-covariance value at lag zero (i.e. the variance of the time series) is the variance related to instru- mental noise. The lag interval from 0.1 to 0.5 s was chosen as a compromise between accuracy and precision of the vari- ance estimate. This method relies on the property of the noise that it is not correlated with the true signal variation. In the following, the noise variance estimate obtained according to the Lenschow et al. (2000) approach is denoted by σ ˆ n, f 2 and the respective flux error according to Eq. (13) by δ ˆ
19 Read more
tosynthesis and ecosystem respiration – photosynthetically active radiation, water table depth, and soil temperature) . However, out of the in situ measurements only air tempera- ture and precipitation were used for developing the RF model for flux upscaling since gridded data products of the other po- tentially important drivers were not readily available and/or the data for the other drivers were missing from several sites. The 30 min averaged flux data were acquired from 21 sites and daily data were provided for 4 sites. The flux time series were quality filtered by removing fluxes with the worst qual- ity flag (based on 0,1,2 flagging scheme, Mauder et al., 2013) and with friction velocity below a site-specific threshold (if friction velocity and threshold were available for the site). After filtering, daily medians were calculated if the daily data coverage was above 29 out of 48 half-hourly data points (daily data coverage at a minimum of 10 data points for sites without a diel pattern in CH 4 flux) and no gap filling was
27 Read more
is rapid warming during the springtime transition. Given an initial value, the integral of dT (blue curve) gives the trajectory of T. Before lag 0, dT bounces around its mean value of zero, so the integral of dT should also bounce around the initial T value, which corresponds to the quasi-steady winter state. During the early springtime transition, a pulse of strong positive dT and EHF cause the rapid warming. After this stage, dT is constantly above zero and has much less variance. The persistently positive dT is mainly due to the fact that solar heating starts to weigh in, and the reduced variance is due to the less varying EHF. The variance of the zonal average of the eddy heat flux at 70N is calculated as following: first calculate the variance of EHF within an 11-day moving window; the variance value is recorded at the center point of the window; then the 30- year climatology of the variance obtained in last step is calculated. The climatology of EHF variance is shown in Fig.3.7 as the blue curve. The three dots are calculated using a similar method for each perpetual simulation. The variance of the perpetual simulations follows the trend in the control simulation: highest variance appears in the winter, and it decreases with time and reaches its minimum in the summer. However, the decreasing trend is not gradual, the variance decreases in a stepwise manner. This sudden decrease in the variance of the eddy heat flux is part of the reason that the springtime transition is abrupt.
128 Read more
Dry deposition, which is the process by which pollutants are transported from the atmosphere to the earth’s surface without precipitation (Seinfeld and Pandis, 2006), is an im- portant component of atmospheric deposition. This process is estimated to account for up to 50 % of total atmospheric deposition in the United States (EPA, 2010; Wesely and Hicks, 2000). Despite this sizable contribution to total at- mospheric deposition, there is a shortage of direct measure- ments of dry deposition in the US. Because of this measure- ment shortage, improving deposition models is crucial. Ad- ditionally, understanding deposition and emission rates is an important piece of estimating atmospheric concentrations in the planetary boundary layer for climate and weather mod- els. Efforts to improve deposition models are ongoing (Say- lor et al., 2014; Zhang et al., 2003; Brook et al., 1999; Pleim et al., 2013), and models estimate flux well under some con- ditions, but fluxes determined by different models and obser- vations can vary by a factor of 2 to 3 (Schwede et al., 2011; Wu et al., 2011; Flechard et al., 2011). Direct dry-deposition measurements are needed to improve and validate models for a variety of ecosystems and environmental conditions.
14 Read more
Investigation revealed that one major difference of class C2 was hidden in the despiking processing step. In fact, TK3 implements the robust statistical method of Mauder et al. (2013) based on median absolute deviation (MAD), while the Gaussian statistical method of Vickers and Mahrt (1997) was used in EddyPro. This difference explained most of the observed scatter. Visual inspection and analysis of flux vari- ances showed that for those scattering points EddyPro re- sults were the implausible ones, giving variances > 10 times larger than those of TK3, which instead fell into plausible ranges. Clearly, for those cases the despiking algorithm of Vickers and Mahrt (1997) – at least with the default settings used – was ineffective to remove large spikes in the raw data, which compromised the flux values. Since the newer despik- ing method of Mauder et al. (2013) proved to be more effec- tive in our case, the same algorithm was also implemented in EddyPro. Hence, the scatter was largely eliminated in the second round for all observed fluxes. Changes in the source code of EddyPro were required to make this possible, and the new implementation will be available to EddyPro users as an alternative despiking method in a forthcoming release. In the second round, the agreement between quality flags also slightly improved because of the improved comparability of the two packages after this modification. It is to be noted that the despiking method of Vickers and Mahrt (1997) is highly customizable. Therefore, we could have followed a different strategy and try and fine-tune that method in EddyPro until results matched satisfyingly. However, because of the sound- ness and simplicity of the MAD method, it was deemed ap- propriate to implement it in EddyPro and propose it as an option to its users.
2 of the original value; the cutoff frequency can also be calculated as the inverse of the response time divided by 2π (see Sect. 4.2). The residence time in the sampling line decreases with increasing flow rate, which also reduces mixing and other line effects that cause a “smearing” of the sample in the manifold. The response time approached ∼ 0.25 s at the highest flow rates that were achievable with our pump/mass flow controller/tubing con- figuration. With decreasing sample residence time in the manifold, the cutoff frequency increases accordingly, result- ing in reduced loss of the measureable flux from high fre- quency attenuation. As the theoretical residence time in the reaction chamber was estimated at 0.04 s (see above), these results show that the response time is primarily determined by the sampling manifold and not by replacing the sample volume inside the reaction chamber. For the different system configurations used thus far, response times on the order of ∼ 0.25–0.40 s have been achieved (cutoff frequencies of 0.4 to 0.6 Hz). As shown below, this resolution allowed us to capture most (>90%) of the ozone flux from the 18 m-high inlet under typical ocean conditions.
15 Read more
based studies (2, 4, 24), shown at the bottom of this table. Part of this difference is due to the inclusion of the Chukchi Sea in the present study, which accounts for 29.4% of the ESAS area and had markedly lower sea-air fluxes than the Laptev or East Siberian seas, the two seas that the earlier three studies based their results solely on. Inclusion of low flux observations from the Chukchi Sea in the earlier measurement-based studies could have reduced pan-ESAS areal flux estimates in all of them. The flux values in (24) were reported in Tg- C- CH 4 year −1 and have been converted to Tg CH 4 year −1 here in Table 3. Extrapolating the
11 Read more
The eddy covariance (EC) method is one of the most es- tablished techniques to determine the exchange of water, en- ergy, and trace gases between the land surface and the atmo- sphere. On the basis of the covariance between vertical wind speed and water vapor density, the EC method calculates the vertical moisture flux (and therefore ET) in high spatial and temporal resolution with relatively low operational costs. The size and shape of the measurement area (EC footprint) vary strongly with time (Finnigan, 2004). Under conditions of limited mechanical and thermal turbulence the EC method tends to underestimate fluxes (Wilson et al., 2001; Li et al., 2008). Energy balance deficits are on average found to be be- tween 20 and 25 % (Wilson et al., 2001; Hendricks Franssen et al., 2010), and therefore latent heat flux or actual evap- otranspiration estimated from EC data shows potentially a strong underestimation. The energy balance closure problem can be corrected by closure procedures using the Bowen ra- tio. However, this is controversially discussed, especially be- cause not only the underestimation of the land surface fluxes but also other factors like the underestimation of energy stor-
17 Read more
perature. We interpret this as an inertia effect of the ther- mohygrometer. So, if the thermohygrometer complex has a higher thermal mass than the ambient air, the temperature measurements taken by the thermohygrometer are attenuated in the high-frequency range. As the attenuation effect was not found in the relative humidity measurements, we assume that the relative humidity measurements were independent of temperature measurements, and therefore relative humidity was not attenuated in the same way as air temperature. Sub- sequently, relative humidity fluctuations were conserved and could be used for the calculations of the water vapour mole fraction. In general, the deviation from the mean is of higher interest than the mean itself for the EC method (Baldocchi, 2014). As long as the relative humidity fluctuations are con- served in the calculations of the water vapour mole fraction, a plausible covariance between the water vapour mole fraction and the vertical velocity can be calculated.
20 Read more
Eddy-covariance data were post-processed analogously for the 99 m tower and for the WSMA turbulence measurements. The software package TK3 (Mauder and Foken, 2011) was used to process the tower EC data, applying the raw data treatments and flux corrections as put forward in Foken et al. (2012). (i) The raw data were screened for spikes us- ing the algorithm of Hojstrup (1993). Visual inspection re- vealed that neighbouring spikes in the IRGA data were not detected by the algorithm. The original algorithm uses av- erage and standard deviation criteria with low break down points for small sample sizes (Rousseeuw and Verboven, 2002). After substituting the criteria with the median and the median absolute deviation, the spikes were efficiently re- moved. (ii) The time delay due to separation between the ver- tical wind measurement and adjacent sensors was determined and corrected by maximizing their lagged correlation. (iii) To correct for potential misalignment, the USA-1 wind mea- surement was rotated into the streamline coordinate system using the planar-fit method by Wilczak et al. (2001). (iv) The temperature variance as well as the sensible heat flux was cal- culated using the crosswind correction by Liu et al. (2001). (v) The formulations by Webb et al. (1980) were used to cor- rect the latent heat flux for density fluctuations.
19 Read more