Interpolation of missing radar wind profiler data

The interpolation of irregularly missing data often involves the assumption of random data gaps and a single metric for optimization. Boundary-layer radar wind profiler data is not suitable for that approach, because data gaps tend to become larger and more frequent with altitude and they can preferentially occur during specific parts of the diurnal cycle.

We have developed and implemented a set of Python algorithms that combines a novel data interpolation scheme with the ability to systematically explore tradeoffs among different choices available to the user. The interpolation involves successive application of a principal component filter, with shorter intervals filled using linear interpolation. For estimation of the error associated with the interpolation, the missing data mask is shifted in time while preserving its phase relative to the diurnal cycle. The resulting interpolation performance is a function of altitude, and the choice of an optimal interpolation scheme and an optimal number of principal components depends upon how performance at different altitudes is weighted.

An outline of the algorithm and examples of its application to profiler data to the Second Texas Air Quality Study (TexAQS-II) will be presented.