Eddy Covariance Frequently Asked Questions

Q What do I need to make basic eddy covariance measurements?

AAt the most basic level, eddy covariance requires a 3 dimensional sonic anemometer, high-speed gas analyzer, data logging device, power supply, and tower or tripod for mounting the instruments. LI-COR offers high-speed analyzers with on-board logging and sonic anemometers. Our instruments can be powered by solar panels almost anywhere in the world, and we provide application notes to walk you through selecting and sourcing your solar power system, tower or tripod, mounting hardware, lightning protection, and additional meteorological sensors.

Q Are there restrictions on where I can use the eddy covariance method?

A One of the first things to consider when looking at a potential eddy covariance site is infrastructure. You will need to access the site to 1) install your measurement tower or tripod, 2) provide power to your instruments, and 3) visit the site occasionally for system upkeep and data retrieval. LI-COR eddy covariance instruments can be operated without main line power or large generators, which greatly expands the number of potential study sites. Also, LI-COR's weatherized instruments do not require temperature regulated instrument sheds or protection from precipitation. In addition to practical issues related to infrastructure, the eddy covariance method relies upon some assumptions about the study area. The more your site deviates from these assumptions, the more potential error is introduced into your data. Site requirements are discussed in more detail in LI-COR's Brief Practical Guide to Eddy Covariance Measurements.

Q What sonic anemometers can I use with LI-COR gas analyzers?

A You can use any sonic anemometer with LI-COR gas analyzers, as long as the anemometer specifications meet the requirements of your study objectives. Selecting a sonic anemometer, however, requires careful consideration. Along with performance specifications, something to consider is integration with the gas analyzer and the data collection device. LI-COR provides easy-to-use weatherized cables for quickly connecting the Gill Instruments, Ltd. WindMaster® or WindMaster Pro and the Campbell® Scientific, Inc. CSAT3 directly to our high-speed CO₂/H₂O and CH4 analyzers and LI-7550 Analyzer Interface Unit (which provides USB data collection). Gill sonic anemometers can be purchased directly from LI-COR with gas analyzers. In addition, mounting hardware, configuration cables, and other accessories are available to make field installation and setup as easy as possible.

Q Why do we need coordinate rotation in eddy covariance data processing?

A One important assumption in the derivation of the eddy covariance equation is that mean vertical wind flow is null. To keep the mean vertical wind speed equal to zero, the sonic anemometer needs to be perfectly level. In reality, however, this is very difficult, if not impossible, to accomplish. Coordinate rotation is a mathematical way to meet this assumption.

Q Why do we consider the Webb-Pearman-Leuning (WPL) term in eddy covariance data processing?

A By definition, mixing ratio (µmol mol^-1) is used in the eddy covariance equation. But most gas analyzers measure molar density (mol m^-3) instead of the mixing ratio. Gas density will change with temperature and humidity. For example, when temperature increases, the mixing ratio of CO₂ does not change, but the CO₂ density decreases. The WPL correction is used to correct for the fluctuations of temperature and water vapor that affected the measured fluctuations in the density of CO₂, H₂O and other gases.

Q What is the Webb-Pearman-Leuning (WPL) term for Eddy Covariance flux? Do I need it?

A The WPL correction is actually a physical term (not a correction) to compensate for fast gas density effects on the gas flux, and it is needed for eddy covariance measurements. It was originally developed by Webb, E.K., G. Pearman and R.Leuning (1980) and published as 'Correction of flux measurements for density effects due to heat and water vapor transfer,' Quarterly Journal of Royal Meteorological Society, 106, 85-100.

It is used to compensate for the fluctuations of temperature and water vapor that affect the measured fluctuations in the density of CO₂, H₂O, and other gases. The WPL term consists of three parts: the sensible heat flux portion (thermal expansion), the latent heat flux portion (dilution), and atmospheric pressure fluctuations portion (usually assumed negligible). It is additive to the initial measured "raw" flux value, and may be significant in open-path systems (e.g., LI-7500, LI-7500A, LI-7700). It may also be important in closed-path systems (e.g., LI-6251, LI-6252, LI-6262, LI-7000), especially when the gas flux is small and the water vapor flux is large. When gas flux is small, it is possible for the WPL term to exceed the uncorrected flux value in both closed-path and open-path systems.

One way to visualize the WPL process is by simply imagining a surface that has an actual zero flux and is covered with continuously warmed air of constant gas concentration. In this scenario, an instrument would measure a flux simply because of volume expansion, which resulted from the warming air.

A more detailed way to visualize the WPL term is to imagine the process at a high frequency scale, for example, 10 Hz. If a CO₂-inert surface is warm and wet, then high-frequency updrafts in the vertical wind speed, w', would be a little warmer and little wetter than downdrafts, because this indicates transport of the heat and water up from the surface into the atmosphere. So then, for CO₂ flux, updrafts would have slightly lower CO₂ density than downdrafts, This high-frequency process could create an appearance of CO₂ uptake when there is no actual CO₂ flux, just because the surface is warm and wet, or both.

Click here for more details on this and other eddy covariance applications:

Q Do I need to measure the water vapor concentrations when measuring CO₂ concentrations and computing eddy covariance CO₂ flux with data from closed path analyzers, such as the LI-6251, LI-6252, LI-6262, or LI-7000?

A Yes. There are two important reasons to have water vapor measured simultaneously with gas fluxes when using any fast closed-path trace-gas analyzer:

1. A reasonable estimate of water vapor concentration inside the cell is needed to determine the mean gas concentration. If the measured water concentration is incorrect, it can affect the mean gas concentration, and ultimately, the computed flux.

2. A good estimate of closed-path water vapor flux is also needed for the Webb-Pearman-Leuning (WPL) latent heat flux term (Webb et al., 1980). This term is needed to compute correct flux, it is additive to the initial gas flux value and is quite significant in both open-path systems and closed-path systems. WPL consists of three parts: sensible heat flux portion (thermal expansion), latent heat flux portion (dilution), and atmospheric pressure fluctuations portion (usually assumed negligible).

For intake tubes with length-to-inner diameter ratios from 1000:1 to 500:1, the sensible heat flux portion of the WPL term is usually considered negligible, at least in the case of normal CO₂ and H₂O fluxes. However, the water vapor (or latent heat flux) portion of the WPL term is still required.

Water vapor flux for the WPL latent heat flux term must be measured in the same sampling cell as the closed-path gas concentration. Actual closed-path water vapor fluctuations are attenuated in the intake tube, and as a result, water vapor flux measured inside the cell is significantly lower than the ambient flux. Thus, using ambient water vapor flux for WPL latent flux correction should be avoided.

Alternatives, such as Nafion tubes and chemical dryers are also undesirable. They can lead to increased power requirements and maintenance demands, and to a decreased frequency response of the system. Effects of the drying arrangements on the gas flux frequency attenuation may also be difficult to predict and correct due to the high variability and site-specific nature of such arrangements, and due to lack of experimental research on the topic.

In summary, measuring the water vapor concentration in fast closed-path gas analyzers is the most reliable way to ensure accurate data.

Q I see that you offer both closed path and open path CO₂/H₂O analyzers for eddy covariance measurements. Which is more appropriate for me?

A There are two things to consider when answering this question: availability of electrical power and local rainfall. In general, open path analyzers have lower power requirements than closed path analyzers. Open path analyzers are ideal in scenarios when the available power is limited.

Regarding rainfall, if a study sight has light, frequent precipitation (e.g. fog, mist, light rain) open path analyzers will miss more data than closed path analyzers. If precipitation events are always heavy, you won't see much difference between the two. With open path CO₂/H₂O analyzers, the signal is lost as soon as liquid water enters the optical path. Sonic anemometers, however, can continue to deliver reliable data until the acoustic sensors become covered with liquid water. So, in conditions of light mist, for example, many sonic anemometers will continue to deliver good data, while an open path CO₂/H₂O analyzer will not. In conditions such as these, a closed-path analyzer would be better. Alternatively, if heavy rain is the main form of precipitation at the site, the anemometer will drop out just about as quickly as an open path analyzer. Recovery is typically quicker for the anemometer than an open path analyzer. A closed path analyzer will have an advantage under these conditions, because you will have complete data as soon as the anemometer recovers.

Q How important is signal synchronization when collecting high frequency data for eddy covariance measurements?

A It is not important that data records be time-aligned in raw data sets. If the delay is known and fixed, time-aligning records is a simple process in post processing of data.

It is, however, important that the sampling frequencies be matched between the different instruments used to make eddy covariance measurements. For data collection at 10 Hz, you would want both the sonic anemometer and gas analyzer(s) to collect 10 data points for every one second of measurement time. This however, is not a given as one second in time as measured by the gas analyzer's clock may be longer or shorter than one second in time as measured by the sonic anemometer's clock. This clock asynchrony can lead to slight differences in the instruments' true sampling frequencies as compared to some reference clock, which ultimately may introduce additional error in flux measurements.

There are several strategies for dealing with clock asynchrony between instruments. LI-COR Bioscience's eddy covariance gas analyzers use an IEEE standard timing protocol (Precision Time Protocol or PTP), which allows analyzers that are networked together to interact to maintain precise time synchronization with each other. In systems using an SDM capable datalogger to collect the data, the SDM command can be used to trigger a data collection event at the same time by all the instruments in the system, or it may simply retrieve the most recent data record from the instruments. In the later case the datalogger serves as a "pseudo-master" clock because the sampling frequency is still driven by the instrument and data collected this way will exhibit a jitter that will be twice the fundamental sampling frequency of the instrument.

Q I know what I want my output frequency rate to be, but what should I set the bandwidth to?

A In most applications of Eddy Covariance, measuring the high-frequency portion of turbulent transport of flux (the significant portion of the transport) requires data collection rates of about 5 to 10 Hz. There are also a few applications which may require 20 Hz measurements, such as airborne and ship-borne applications, urban and complex terrain studies, and measurements located extremely close to the ground or vegetation.

Regarding the bandwidth setting and its relationship to the frequency output setting, there are a couple thoughts to consider.

The more common and traditionally accepted method (Method 1) is to select a bandwidth that is half the output frequency. For example, a frequency output rate of 10 Hz should have a bandwidth rate set to 5 Hz. This follows the well-known Nyquist Frequency Theorem, also known as Nyquist—Shannon sampling theorem, and is designed to prevent signal 'aliasing' during high-frequency sampling. The theorem has been verified numerous times since its introduction in 1928, and presently is widely used in experimental physics, electronics, and other areas dealing with sensitive, fast data sampling. It should be mentioned that anything beyond the selected bandwidth will be eliminated by the Digital Signal Processing (DSP) filter.

The other method (Method 2) is to set the bandwidth equal to the output frequency. For example, when both are set to 10 Hz. This will cause the signal will be aliased. Specifically, the portion of high-frequency contributions will fold onto the lower-frequency contributions. Since flux calculations are 'integrated' measurements, this method may provide correct raw covariance over given time period, similar to the Method 1, although it is not entirely clear if high-frequency uncertainties folded onto lower frequency signal would increase low-frequency uncertainties.

In regards to Eddy Covariance measurements, both methods should generally provide a similar raw covariance over a given time period, but there are some practical benefits to using Method 1 in eddy covariance studies. With eddy covariance measurements, setting the bandwidth the same as the output rate (Method 2) would introduce aliasing into measured values. When the bandwidth is set to half the output rate (Method 1), the aliasing is avoided.

However, there may indeed be occasions where the expected spectral feature and/or storage limits would justify (for example) a 10 Hz output and a 10 Hz bandwidth arrangement, even if it implies aliasing.

When data is not aliased, the main benefit to a researcher is the ability to perform quality checks on the spectral and co-spectral data. Two key quality checks often performed with eddy covariance measurements are; (i) the comparison of spectra and co-spectra to the Kaimal, Massman or other turbulent transfer models, and (ii) checking the correctness of co-spectra-based frequency response corrections.

These checks are important for quality control of measurements at different frequencies. The aliasing may mask or distort high-frequency and middle-frequency portions of spectra and co-spectra, and it may be difficult to diagnose the potential issues in these portions of the data.

In addition, if data are collected with 'forced aliasing' (by Method 2), such data are not recoverable, and cannot be put back to pre-aliased conditions. One could not go back with aliased data and investigate the two key quality checks. So, while either method may be okay for computing raw covariance over given time period, the spectral and co-spectral considerations listed above may be important for choosing the best method for setting the bandwidth frequency.

Kaimal, J. C., Wyngaard, J. C., lzumi, Y. and Cote, O. R. 1972. Spectral characteristics of surface-layer turbulence. Quarterly Journal to the Royal Meteorological Society. 98: 563-589.
Massman, W. J. 2000. A simple method for estimating frequency response corrections for eddy covariance systems. Agricultural and Forest Meteorology. 104: 185-198.