Trade-Offs in Flux Disaggregation: A Large-Eddy Simulation Study

Airborne flux measurements allow us to quantify the surface–atmosphere exchange over heterogeneous land surfaces. While often applied to regional-scale fluxes, it is also possible to infer component fluxes emanating from different surface patches from the measurement via disaggregation strategies. Here, we emulate flux disaggregation strategies by conducting an ensemble of virtual flight measurements within a set of large-eddy simulations over idealized surface heterogeneities and under different flow regimes. The resulting patch surface fluxes are compared with the prescribed patch surface fluxes in the simulation. To calculate fluxes along the flight legs, we apply traditional eddy-covariance and space–frequency (wavelet) methods. We show that the patch fluxes are captured best with the space–frequency method, where the disaggregation error is almost invariant of the segment length. For the eddy-covariance method, however, the error strongly depends on the segment length, with largest random and systematic errors for shorter segments. Furthermore, we determine a trade-off between a permissible disaggregation error and a sufficient resolution of the heterogeneous surface signals. Among our set-ups, an optimal segment length is determined to be 3–4 km for the eddy-covariance method, while with the space–frequency method even shorter segment lengths of a few hundreds of metres can be chosen, which enables sufficient isolation of signals from surface patches and the resolution of small-scale surface heterogeneity.