Eddy Covariance in Tricky Spots: Importance of Flux Quality Control and Uncertainty quantification for Lake Surface Flux Measurement

The application of eddy covariance for long-term observation of surface-atmosphere exchanges of heat, momentum, water vapor, and other scalars is now well-established. However, accurate evaluation of long-term balances, patterns, and trends in these fluxes and relationships with drivers of these fluxes requires proper attention to quality control and uncertainty quantification. These criteria become even more important when flux towers are placed in "non-ideal" environments, that is, ones that suffer from heterogenous surface conditions, flow obstructions, topography, dense canopies or cold surfaces that encourage flow decoupling and drainage, weak turbulence, or extreme climates. Similarly, instrumentation and infrastructure specifics, such as tubing length, sensor separation, calibration protocol, and data stream time synchronization can impart significant uncertainty. Most attempts to identify flux quality and uncertainty have focused on more "ideal" sites, such as homogenous and flat grasslands or forests. Here, we present analysis based on application of a recently published flux processing, quality control, uncertainty, and footprint analysis framework (Mauder et al., 2013) to a very non-ideal site, on a building at the shoreline of a medium sized lake in the midwest US. Lakes suffer from being in low-spots and having surfaces that tend to be colder than the overlying air, which also tends to be saturated with water vapor from evaporation. Measurement platforms are also less secure than land-based towers. There is a growing body of literature and community interest in understanding in situ exchanges of carbon, water, and energy among aquatic systems, given their recently recognized and previously under-appreciated importance as active parts of terrestrial biogeochemical cycles. Our analysis reveals that the quality control framework, while significantly reducing the number of valid points, is quite effective at flagging unlikely fluxes, such as very large negative latent heat fluxes in daytime, or those caused by flow obstructions and also identifying sensor failure in icing conditions. The analysis also highlights how a previous intriguing finding of carbon uptake prior to ice out in the lake was likely not as real as we initially thought. We conclude that proper eddy covariance analysis frameworks allow the community to expand application of flux measurement to a larger range of "non-ideal" sites, where much of the major research needs for flux measurements are currently arising.