Experimental evaluation of flux footprint models

- Four flux footprint models were evaluated based on artificial tracer experiments.
- Most evaluated models match observations approximately over a wide stability range.
- Most evaluated models overestimate the distance of the footprint's peak location.
- Model quality is sensitive to changes in surface roughness upwind of the flux tower.

In this study we experimentally evaluate analytical flux footprint models, as well as models based on Lagrangian stochastic particle dispersion. For this purpose, we conducted tracer experiments at a grassland site in southern Germany. An artificial tracer was released continuously over a number of flux-averaging intervals from a surface source. The flux contributions from the tracer source were measured by eddy covariance and compared to those predicted by footprint models. Furthermore, an additional eddy covariance measurement tower was used to evaluate the along-wind distribution of footprint models, as well as to analyze to what extent a forest edge upwind of the measurement tower affects model performance. Additionally, we quantify footprint model uncertainty resulting from the random error of input parameters.Our measurements show that all evaluated models match observations roughly, but tend to underestimate the value of the footprint maximum, and overestimate its distance. The analysis of stability dependence of model performances indicates that one model, based on simulation outputs of a Lagrangian stochastic model, clearly underestimates observations for near neutral to stable conditions, while no clear stability dependence could be identified for the performance of the other models.As expected, model performance is sensitive to an abrupt change in surface roughness and sensible heat flux at a forest edge in the near upwind fetch of the measurement tower. Using the local apparent roughness length (derived from measured wind speed and friction velocity) only slightly or negligibly improved model performance compared to the use of a constant local roughness length (determined from local surface characteristics). Thus we confirm experimentally that footprint estimates and related data quality assessments should be handled with care at sites with inhomogeneities in surface roughness.