Cross-Site Evaluation of Methods for the Estimation of Aerodynamic Roughness Parameters from Flux-Tower Data

FLUXNET is a vast network of more than 800 eddy covariance flux sites dispersing across all continents and most of the world's ecoregions. The global network provides valuable datasets of the direct and in situ measurements of fluxes and ancillary variables that are used across different disciplines and applications. Aerodynamic roughness parameters (i.e., roughness length, zero plane displacement height, aerodynamic canopy height) are one of the potential data products that are crucial for the applications of land surface and ecosystem modelling but have not yet been routinely generated in FLUXNET. This study aims to compare several available methods for the estimation of aerodynamic roughness parameters from single-level eddy covariance measurements and evaluate their feasibility and robustness for the application of FLUXNET datasets. Two approaches are adopted based on the surface-layer theory (i.e., logarithmic wind profile) and flux variance similarity (i.e., turbulent characteristics). Each approach is implemented by using a series of different estimation techniques, such as least-square regression, numerical iteration, and Markov chain Monte Carlo method. We run the tests across a broad range of ecosystem types ranging from tall-, short-canopy, to open water sites, from closed-canopy/homogeneous to open-canopy/heterogeneous sites, and also from evergreen to deciduous sites. Our findings show no single method dominates in terms of model performance and robustness across all sites. An ensemble average of all plausible estimates or only estimates from site-specific/pre-selected methods may be the most feasible approach. Alternatively, a semi-empirical approach based on the assumptions of presumably known relationships among aerodynamic roughness parameters may provide a robust and sufficiently accurate estimate.