Averaging, Detrending, and Filtering of Eddy Covariance Time Series

Data from sensors in an eddy covariance system are routinely processed to remove trends and to produce fluctuations and means. Historically this has been seen to be a relatively straightforward task and the methods are well known. Such re-processing can result in the loss of real signal since the detrending and averaging methods act as high-pass filters. We review the main methods used to separate the active, turbulent transport that we treat as eddy flux from the slower, deterministic atmospheric motions and instrument drift. We discuss the advantages and disadvantages of various algorithms used in averaging, detrending and filtering and conclude that the best method is likely to be dependent on site conditions and data processing system in use. We recommend the use of the ogive to determine the optimal averaging period at any site. We illustrate outstanding issues with data from a number of FLUXNET sites. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.