A Comprehensive Error Analysis for Eddy Covariance Measurements

Despite the fact that eddy covariance measurements have become a standard tool in analysing ecosystem-atmosphere exchanges of energy and trace gases, reporting error estimates for eddy covariance flux data has not been a widespread practice in the past. Only recently has it become more common to report error estimates for eddy covariance data. Many of the reported error estimates are based on analysing the error made by gap filling incomplete eddy covariance data series and are for annual or seasonal total fluxes only. Although the error made by gap filling is an important source of uncertainty for eddy covariance flux estimates, other sources of systematic and random errors in eddy covariance data can also contribute significantly to eddy covariance flux errors. The impact of different error sources on eddy covariance flux data will also likely vary across different timescales (half hourly to annual).

Additional sources of errors can be grouped in the following way:

* Errors introduced by instrumentation: e.g. intrinsic accuracy/ precision of the instruments, angle of attack bias of sonic anemometers, sensor drift, etc.

* Errors introduced by setup: e.g. surface heterogeneity, advection, single point of sampling, variations in footprint

* Errors introduced by processing: e.g. spectral correction, gap filling, length of averaging interval

Some of these errors can be avoided or minimized by careful and thorough experimental design and instrument maintenance, others are unavoidable and can only be controlled to some extend. Many of these errors are often ignored when reporting eddy covariance measurement results. Past studies have examined the effect of some of the listed sources of errors on eddy covariance measurements. However, most of these studies focused on a single source of error.

We will present a comprehensive analysis of the effect of various random and systematic errors on eddy covariance flux measurements of CO2, CH4, latent heat, and sensible heat. We will also analyse the relative contributions of these errors to the total uncertainty of half hourly, daily, monthly, and annual fluxes.