Comparative assessment of methods to retrieve fine-scale atmospheric boundary layer structures from airborne measurements

Measurement platforms based on either manned or remotely piloted light aircrafts are ideal instruments for collecting detailed observations of the atmospheric boundary layer structure, especially over complex terrain. In order to retrieve the 3D fine-scale features of these structures from airborne measurements, suitable interpolation techniques are required to remap such observations over high-resolution regular grids. In particular, a residual kriging (RK) technique is proposed by the authors for this application. RK performance is assessed – both in terms of likelihood of the resulting fields and interpolation accuracy – in comparison with other techniques previously applied in the literature, namely inverse distance weighting, exponentially-weighted inverse squared distance and natural neighbor methods. Observations of potential temperature at a mountain valley cross section, collected during three flights of an instrumented motorglider, are adopted as test-bed dataset. A critical comparison of the re-gridded 3D meteorological fields, as well as an extensive cross-validation analysis, is/are carried out to evaluate the different interpolation techniques. We find substantial differences across the performances of the four methods, with RK showing the best results. However, better estimates are observed for the other methods when a residual approach, i.e. the decomposition of the target variable into a dominant drift and a residual term, is adopted as in RK. Moreover, we examine the characteristics of each method, discussing advantages and disadvantages offered by RK, and provide further evaluations about the suitability of the different interpolation techniques and about the choice of the flying strategy.