Turbulent flux measurements are key to understanding ecosystem scale energy and matter
exchange, including atmospheric trace gases. While the eddy covariance approach has
evolved as an invaluable tool to quantify fluxes of e.g. CO2 and H2O continuously, it is
limited to very few atmospheric constituents for which sufficiently fast analyzers exist. High
instrument cost, lack of field-readiness or high power consumption (e.g. many recent
laser-based systems requiring strong vacuum) further impair application to other
tracers.
Alternative micrometeorological approaches such as conditional sampling might
overcome major limitations. Although the idea of eddy accumulation has already been
proposed by Desjardin in 1972 (Desjardin, 1977), at the time it could not be realized for trace
gases. Major simplifications by Businger and Oncley (1990) lead to it’s widespread
application as “Relaxed Eddy Accumulation” (REA). However, those simplifications (flux
gradient similarity with constant flow rate sampling irrespective of vertical wind velocity and
introduction of a deadband around zero vertical wind velocity) have degraded eddy
accumulation to an indirect method, introducing issues of scalar similarity and often lack of
suitable scalar flux proxies.
Here we present a real implementation of a true eddy accumulation system according to
the original concept. Key to our approach, which we call “Conditional Eddy Sampling”
(CES), is the mathematical formulation of conditional sampling in it’s true form of a direct
eddy flux measurement paired with a performant real implementation. Dedicated hardware
controlled by near-real-time software allows full signal recovery at 10 or 20 Hz, very fast
valve switching, instant vertical wind velocity proportional flow rate control, virtually no
deadband and adaptive power management. Demonstrated system performance
often exceeds requirements for flux measurements by orders of magnitude. The
system’s exceptionally low power consumption is ideal for the field (one to two orders
of magnitude lower compared to current closed-path laser based eddy covariance
systems). Potential applications include fluxes of CO2, CH4, N2O, VOCs and other
tracers.
Finally we assess the flux accuracy of the Conditional Eddy Sampling (CES) approach as
in our real implementation relative to alternative techniques including eddy covariance (EC)
and relaxed eddy accumulation (REA). We further quantify various sources of instrument and
method specific measurement errors. This comparison uses real measurements
of 20 Hz turbulent time series of 3D wind velocity, sonic temperature and CO2
mixing ratio over a mixed decidious forest at the “ICOS” flux tower site “Hainich”,
Germany.
Results from a simulation using real wind and CO2 timeseries from the Hainich site from
30 April to 3 November 2014 and real instrument performance suggest that the maximum
flux estimates error (50% and 75% error quantiles) from Conditional Eddy Sampling (CES)
relative to the true flux is 1.3% and 10%, respectively for monthly net fluxes, 1.6% and 7%,
respectively for daily net fluxes and 8% and 35%, respectively for 30-minute CO2
flux estimates. Those results from CES are promising and outperform our REA
estimates by about a factor of 50 assuming REA with constant b value. Results include
flux time series from the EC, CES and REA approaches from 30-min to annual
resolution. |