Aerodynamic roughness parameters, namely the roughness length ( \({z}_{0}\) ) and zero-plane displacement height ( \(d\) ), are crucial for representing turbulent momentum exchange in urban boundary layer models. Numerous morphometric methods have been developed to estimate these parameters from wind-tunnel experiments and computational fluid dynamics (CFD) simulations. However, their quantitative performance evaluation for heterogeneous, real urban surfaces has been limited by a scarcity of observational turbulence data. This study addresses this gap by developing a comprehensive turbulence database from 852 large-eddy simulations (LES) over a wide range of real urban morphologies. This database, which includes explicitly resolved surface momentum fluxes ( \({\tau}_{0}\) ) and corresponding urban morphometric parameters, is used to evaluate six widely used morphometric methods. Specifically, we focus on evaluating the performance of these formulations in reproducing the surface friction velocity ( \({u}_{*}\) ), which governs the momentum exchange in Monin–Obukhov Similarity Theory (MOST). Results show that formulations accounting for building height variability (i.e., heterogeneous methods) generally outperform those based on homogeneous assumptions in reproducing \({\tau}_{0}\) . For instance, some heterogeneous formulations achieve high correlations (R > 0.8) but tend to overestimate momentum flux due to an over-representation of pressure drag. We find that the performance of all formulations degrades at both high and low extremes of heterogeneity, with none performing adequately in friction-dominated regimes. By analysing this extensive database, our study delineates the conditions under which urban morphology governs momentum exchange and identifies the critical limitations of existing methods. These findings provide a clear directive for developing improved parameterizations capable of accurately representing momentum fluxes over diverse urban landscapes.
