Quantitative evaluation of turbulence reconstruction algorithms for flux estimation based on the characteristics of atmospheric turbulence

This study evaluates turbulence reconstruction algorithms based on the characteristics of atmospheric turbulence using large-eddy simulation (LES). Three idealized boundary-layer regimes (convective, stable, and neutral) under homogeneous surface conditions, together with a near-realistic full-day case incorporating land surface model, large-scale forcing, and nudging modules are examined. The LES is configured under four nested regions, combined with high-resolution data from an observational experiment at Horqin Sandy Land in July 2022, which generates a hybrid dataset validated via spectral analysis and contribution test to preserve resolved turbulence structures while compensating subgrid-scale limitations. From the idealized simulations, single-point eddy covariance flux calculations prove to be consistent with horizontal averages in magnitude with high temporal variability and systematic overestimation due to contamination by non-turbulent motions. These disadvantages are mitigated by the reconstruction algorithms derived from three distinct aspects of turbulence characteristics: the properties of transport excel in stable/neutral conditions but risk overcorrection in convective regimes; fractal dimension methods aggressively isolate turbulence in strongly unstable layers; anisotropy-based approaches demonstrate robustness across stratification types except rare neutral conditions. Full-day simulations confirm diurnal patterns of flux overestimation and algorithm efficacy, with context-dependent selection reducing flux overestimation from 40%–90% to around 20%. These results highlight the potential of turbulence characteristics in turbulence reconstruction and turbulent transport estimation, advocating further exploration concerning the efforts in the parameterization schemes of the atmospheric boundary layer.