Identifying and Managing Data Validity Challenges with Automated Data Checks in the AmeriFlux Flux Measurement Network

Abstract:

AmeriFlux is a network of sites managed by independent investigators measuring carbon, water and heat fluxes. Individual investigators perform many data validity checks. Network-level data validity checks are also being applied to increase network-wide data consistency. A number of different types or errors occur in flux data, and while corrections have been developed to address some types of errors, other error types can be difficult to detect. To identify errors rapidly and consistently, we have developed automated data validity checks that rely on theoretical limits or relationships for specific measured variables. We present an example of a data validity check that is being developed for the friction velocity u_*. The friction velocity is a crucial variable used to identify when low turbulent mixing in the atmospheric boundary layer invalidates eddy covariance measurements of fluxes. It is measured via sonic anemometer and is related to the wind speed WS, the measurement height relative to the canopy height, and the surface roughness, through the log law. Comparing independent measurements of WS and u_* can help identify issues related to the sensor but doesn’t take into consideration changes in the canopy (e.g. due to leaf emergence). The u_* data check proposed relies on recent work comparing multiple methods for determining the aerodynamic roughness length z₀ and zero plane displacement d (Graf, A., A. van de Boer, A. Moene & H. Vereecken, 2014, Boundary-Layer Meteorol., 151, 373-387). These methods, each of which is most robust across a different atmospheric stability range, yield multiple estimates for z₀ and d at daily resolution. We use these multiple estimates for z₀ and d, as well as half-hourly wind speeds and Obukhov length scales and their uncertainties to generate a predicted u_* and a tolerance around this predicted value. In testing, this check correctly identified as invalid u_* data known to be erroneous but did not flag data that could look anomalous but instead reflect real changes in the vegetation canopy. This and other validity checks both manual and automated are serving to increase accuracy and inter-comparability of all data within the AmeriFlux and FLUXNET networks preceding the upcoming release of a new global data set of fluxes.