Averaging windows for estimating turbulent fluxes by airborne eddy-covariance over a heterogeneous oasis-desert system

Airborne eddy covariance (EC) systems can measure turbulent fluxes by rapidly sampling the atmospheric boundary layer over complex terrain at the regional scale, revealing turbulent structures across multiple spatial scales. Over a heterogeneous surface, however, their application is complicated by non‑stationary conditions, and by the need to choose averaging windows that balance mitigating random errors against preserving the imprint of land‑surface heterogeneity on the measured fluxes. In this study, we developed a wavelet-based adaptive method to determine averaging windows. Window boundaries along flight lines are delineated from the spectral wavelet analysis of the target covariances, so that the averaging windows explicitly follow the underlying turbulent structures. A central part of our evaluation is a systematic analysis of the relative random errors associated with these shorter, turbulence-guided windows. The results show that the adaptive windows keep relative random errors small, with median values of around 20% for latent heat flux (LE) and 15% for sensible heat flux (H), despite the substantially shorter averaging length. This suggests that they sample segments over which the turbulent statistics are more internally consistent. By reducing random error without lengthening the window, the method resolves the trade-off that forces fixed-window processing to sacrifice spatial resolution for statistical reliability, while it preserves the spatial imprint of land-surface heterogeneity on the fluxes. This is illustrated by many instances in the fragmented irrigated oasis of the study area. As an additional evaluation, we examined how the Bowen ratio derived from our flux estimates relates to footprint-weighted land-surface properties and found overall trends consistent with the expected controls by land surface properties. Overall, the proposed adaptive window approach enhances the reliability (in terms of random error) of airborne flux observations in resolving spatial patterns in surface heat fluxes, which supports a better understanding of the dominant energy exchanges between the land surface and the atmosphere.