Innovative Cross-Correlation Method for Determing Three-Dimensional Wind

Wind velocity can be measured by using either the Doppler method or interferometry technique. The Doppler method is routinely used to obtain radial velocity, while the interferometry technique has the potential of obtaining transverse velocity. In the past, the interferometry technique for wind measurement has been interpreted as the motion of the interference pattern associated with the motion of scatterers. The transverse velocity is half the pattern velocity for a spaced-antenna technique. However, the motion of the scatterers can be directly described using correlation and de-correlation (basic statistical property of a random field). As the sample volume (or receiver) follows the motion of the randomly distributed scatterers, correlation is maximum; otherwise, the correlation of the scattered wave signal decreases. Correlation is minimal when the movement of sample volume and motion of scatterers are in opposite directions. Therefore, the ratio of the forward and the backward correlation functions is directly related to the scatterer velocity.

In this paper, we describe a cross-correlation method for estimating 3-D wind fields. The proposed approach combines the Doppler method and interferometry technique for wind retrieval, and uses both the phase and magnitude of cross-correlation function. The phase is used for determining radial velocity while the ratio of cross-correlation function is for estimating transverse velocity. The logarithm of the cross-correlation ratio linearly depends on time lag and the slope is used to obtain transverse velocity. The velocity estimation is optimized by using multi-level correlation processing, which combines received signals from individual receivers to increase signal/noise ratio and then calculates the correlation function. The method is illustrated by processing the NCAR Multiple Antenna Profiler (MAPR) data for velocity retrieval. The MAPR is 915 MHz wind profiler and its receive-antenna is divided into 4 vertically pointing sub-arrays that allow correlation processing. The velocity is retrieved at the 1st, 2nd, and 3rd level. The consistency and advantages of the current technique are discussed by comparing the results with those obtained using existing methods.