Can Data Mining Help Eddy Covariance See the Landscape? A Large-Eddy Simulation Study

Eddy-covariance fluxes serve as an essential benchmark for Earth system models and remote sensing data. However, two challenges prevent model-data intercomparisons from fully utilizing eddy-covariance fluxes. The first challenge stems from the differing and variable spatial representativeness of the eddy-covariance measurements, or footprint bias and transience. The second originates from the phenomenon of a non-closed energy balance using eddy-covariance measurements, hypothesized to result from unaccounted mesoscale flows or under-sampling of hot spots by flux towers, among others. Previous studies have suggested that these two problems can be mitigated by either building multiple towers or by applying space-time rectification approaches, such as the environmental response function (ERF) approach. Here we ask: (1) How many eddy-flux towers do we need to sufficiently rectify location bias, close the energy budget, and sample the regional domain? (2) Can an advanced space-time rectification approach reduce the tower density, while still adequately sampling the regional flux domain? Furthermore, (3) How accurately can the ERF approach retrieve the surface-flux variation? To answer these questions, we used data from a large-eddy simulation of atmospheric flows above a heterogeneous surface as captured by an ensemble of virtual tower measurements. We calculated eddy-covariance fluxes by spatial and spatio-temporal methods. The spatial eddy-covariance method captured 89% of the prescribed total surface energy flux with about one tower per 15 km(2), while the spatio-temporal method required only one tower per 40 km(2) to capture 95% of surface energy. To capture 97% of energy, applying the ERF approach further reduced the required tower density to one tower per 200 km(2), as a result of space-time rectification and incorporating mesoscale flows. This approach also enabled retrieving the surface spatial variation of the sensible heat flux. The results provide a reference for informing and designing future observation systems based on flux tower clusters, and scale-aware data products.