1 Introduction

Package source: Reddy package (Github)
Jupyter notebooks: jupyter notebooks (Github)
Any comments or issues? create an issue (Github)

Overview: Eddy-covariance (EC) measurements allow to measure turbulent fluxes directly and non-invasively over long periods of time, and thus represent the standard measurement method for turbulent exchange processes between atmosphere and land, vegetation, cryosphere or hydrosphere.
However, they require a thought-through technical setup and a special post-processing, before further analysis can be carried out. This gitbook provides an overview about the setup, post-processing and meteorological evaluation of EC measurements, which can be used for research and teaching. The structure of this gitbook is detailed in figure 1. All described functions are implemented in the R-package Reddy, which was specially developed to allow a reproducible and comprehensive analysis of EC measurements. Each chapter covers one topic and is also available as jupyter notebook, which can be downloaded here.

Figure 1: Overview of the topics in this gitbook and the functionality of the Reddy package

Installation of the Reddy package: The Reddy package can be installed directly from github:

        devtools::install_git("https://github.com/noctiluc3nt/Reddy")

Entering a function name function in an R terminal, shows the function and the performed calculations, such that the calculation and plotting procedure is fully comprehensible. The Reddy package depends on the libraries MASS (for kernel density estimation), pracma (basic linear algebra) and RcppRoll (C++ interface for accelerated data handling), which are automatically installed with Reddy.

1.1 Measurement setup and instrumentation

Setup: An EC setup has two main components: the sonic anemometer (for the wind components and temperature) and a gas analyzer (for water vapor, carbon dioxide, methane, …). The sonic usually consists of three pairs of transducers, that transmit and receive ultra-sonic sound waves whose propagation speed depends on wind speed, temperature and humidity. Therefrom, the three wind components and the sonic temperature, which is approximately the virtual temperature, can be derived. The gas analyzers utilize usually an absorption line in the near- or mid-infrared of the respective trace gas, to measure their number density (IRGA - infrared gas analyzer). The air is pumped from the inlet (with a flow rate of about 12 l/min, that is controlled by the flow module which contains a pump), through the inlet tube into the gas analyzer, where the actual measurement takes place. There are different types of infrared gas analyzer, the most common difference beeing between gas analyzers with closed or open measuring path. Further possible methods for gas analyzers are laser spectroscopy, mass spectroscopy, chromatography and chemiluminescence (often used for NOx) and also particle flux measurements can be combined with sonic measurements to derive particle fluxes.
The EC system can be mounted on a tower or mast with one or several measurement heights. For choosing the location and measurement height, the surface roughness and the flux footprint (i.e., the area where the flux originates from) are examined. Since the mast or tower disturb the EC measurements, the orientation of the system is chosen based on the prevailing wind direction(s), such that the main wind direction(s) are undisturbed by the tower structures, as exemplified in figure 3. The sonics have an internal orientation (and should face north as indicated on the sonic), and they should be levelled (however, their precise levelling is not essential as the wind is rotated in the post-processing anyway). There are several options for the inlet position: Usually it is placed below the sonic (see the setup in Kuivajärvi), but most importantly it should not interfere with the sonic measurements. A detailed description of high-standard EC site setup (following the standards of the Integrated Carbon Observation System, ICOS) can be found in Rebmann et al. (2018).

Figure 2: Eddy-covariance sites in different environments: urban (Helsinki, Finland), lake (Kuivajärvi, Finland), boreal forest (Sodankylä, Finland), and alpine tundra (Finse, Norway).

Calibration and Maintainance: The EC system requires regular maintainance, in particular the gas analyzers. The gas analyzers are calibrated regularly in-field, and more rarely, intensively by the manufacturer. For this prupose, reference measurements are carried out with gas cylinders containing a fixed amount of the gas, e.g. zero-gas or 450 ppm CO\(_2\). For water vapor, usually only the zero-gas calibration is performed, since fixed water vapor concentrations are difficult to maintain. For other trace gases, the in-field calibration is performed with a zero-gas and another fixed concentration, while the manufacturer calibrates for several fixed gas concentrations. The optics of the gas analyzer have to be cleaned regulary depending on the environmental conditions, whereby open-path gas analyzers require more maintainance than closed-path gas analyzers. Additionally, the inlet filters should be cleaned or replaced regulary as well as the sampling tube. The sonic requires less maintainance, however, the measurements are sensitive to the distance of the transducer pairs, so their distance should be checked.

Figure 3: Components of an eddy-covariance system (demonstration setup)

1.2 Data sources

FLUXNETs: There are several coordinated FLUXNETs, which standardise eddy-covariance measurements regionally or globally and provide post-processed flux data. For example:

  • FLUXNET: FLUXNET combines several regional networks to a global flux dataset. A list of regional networks can be found here and a list of stations here. The most recent global flux data set is FLUXNET2015 (Pastorello et al. 2020).
  • ICOS: ICOS (Integrated Carbon Observation System) is a European network of different station types (Atmosphere, Ecosystem, Ocean), which observe carbon concentrations and fluxes. The atmospheric stations are listed here and data can be downloaded from the data portal.

Other packages for processing of eddy-covariance data: Multiple other packages for eddy-covariance data processing exist with different applications and different degrees of specialization as well as in different languages. The main focus of these packages, however, is the post-processing of raw eddy-covariance data (often as wrapper of manufacturer software) and long-term ecosystem monitoring studies, as summarized below. In this regard, Reddy R fills a gap, as it additionally considers turbulence theoretical and meteorological applications and allows for a fully customized post-processing and analysis of eddy-covariance data.

  • EddyPro® Fortran: Post-processing of eddy-covariance data from the manufacturer LI-COR Biosciences.
  • ONEFlux C (“Open Network-Enabled Flux processing pipeline”): Post-processing of (half-)hourly eddy-covariance data used to create the FLUXNET2015 dataset (Pastorello et al. 2020).
  • REddyProc R: Post-processing of (half-)hourly eddy-covariance measurements.
  • openeddy R: Post-processing of eddy-covariance data, aligned with REddyProc.
  • RFlux R: GUI for post-processing of eddy-covariance raw data by calling EddyPro®.
  • eddy4R R (Metzger et al. 2017) : Family of several R-package for eddy-covariance post-processing in a DevOps framework.
  • icoscp Python: Access to data from ICOS (Integrated Carbon Observation System) data portal.
  • flux-data-qaqc Python (Volk et al. 2021) : Post-processing of eddy-covariance measurements to derive daily or monthly evapotranspiration in a energy balance framework.

References

Metzger, S., D. Durden, C. Sturtevant, H. Luo, N. Pingintha-Durden, T. Sachs, A. Serafimovivh, et al. 2017. “eddy4R 0.2.0: a DevOps model for community-extensible processing and analysis of eddy-covariance data based on R, Git, Docker, and HDF5.” Geosci Model Dev 10: 3189–3206. https://doi.org/10.5194/gmd-10-3189-2017.

Pastorello, G., A. Trotta, E. Canfora, and others. 2020. “The FLUXNET2015 dataset and the ONEFlux processing pipeline for eddy covariance data.” Sci Data 7 (225). https://doi.org/10.1038/s41597-020-0534-3.

Rebmann, C., M. Aubinet, H. Schmid, and others. 2018. “ICOS eddy covariance flux-station site setup: a review.” Int Agrophys 32: 471–94. https://doi.org/10.1515/intag-2017-0044.

Volk, J., J. Huntingtion, R. Allen, F. Melton, M. Anderson, and A. Kilic. 2021. “flux-data-qaqc: A Python Package for Energy Balance Closure and Post-Processing of Eddy Flux Data.” Journal of Open Source Software 6 (66). https://doi.org/10.21105/joss.03418.