A modelling approach to estimate the friction velocity threshold for Eddy Covariance measurements in urban areas

The Eddy Covariance (EC) technique has been widely used to quantify gas exchanges between ecosystems and the atmosphere. Recently, the application of this method in urban areas has gained attention within the scientific community, aiming to understand, measure, and track the gas exchanges in densely populated areas. Among all the EC-related parameters, friction velocity (u*) is commonly used to identify non-turbulent (and thus unreliable) fluxes, calculating a threshold for filtering the data. However, standard approaches to calculate the u* threshold in urban areas cannot be applied.

Taking advantage of the data available from the FLUXNET Data System, we developed a modelling approach to estimate the u* threshold in urban sites. A set of predictive variables related to site physical and meteorological characteristics, u* values, and distribution indices (kurtosis, skewness) were tested within a multiple linear regression on non-urban sites. The relation was then applied to urban sites, calculating “synthetic” u* threshold values.  

Preliminary results show that u* has the highest predictive capacity, while the other variables add only a relatively small contribution in improving the model accuracy. In addition, the choice of site-related physical characteristics should be carefully evaluated according to their different behaviour in urban and non-urban sites. Since it is not possible to retrieve reference u* threshold values for urban sites, model validation is implemented only for non-urban sites.

Although without performing a calibration directly on urban sites, the proposed modelling approach represents a precious refinement in estimating EC fluxes in urban areas, allowing to generate u* threshold values where not possible with standard approaches. Moreover, the new FLUXNET Data System launched in December 2025 ensures a robust model calibration, providing a larger dataset compared to the previous data release.