Monday, 10 February 2003: 10:00 AM
Quantitative Analysis of the SZ(8/64) Phase Code for the Mitigation of Range and Velocity Ambiguities in the WSR-88D
The range-velocity ambiguity problem can cause significant data loss in the lower elevation scans for the WSR-88D. The problem is the fact that the unambiguous range increases with increasing pulse repetition time (PRT), while the unambiguous, or Nyquist, velocity decreases with increasing PRT. Therefore, adjusting the PRT to obtain a desirable unambiguous range results in a small Nyquist interval and severe velocity folding. Conversely, a PRT that provides an adequate Nyquist interval delivers unacceptably short unambiguous ranges, resulting in inadequate area coverage and increased multiple trip contamination. One solution is to phase shift the transmitted pulses using a predetermined pattern called the SZ(8/64) phase code. The spectral moments from the multiple trip echoes can then be obtained by various recovery procedures. This approach has been shown to be a promising technique through simulation experiments and field studies on NCAR's S-Pol radar and NSSL's KOUN testbed radar. The evaluation of the moment recovery algorithms on experimental data has been largely qualitative because of a lack of baseline "truth" to compare with the algorithms' results. The radar cannot transmit both phase coded and non-phase coded data at the same time, thus limiting comparisons to data that is taken at different times, with different radars, or both. This makes precise comparisons impossible, especially if the weather is evolving rapidly.
To generate a truth field for quantitative analysis, two or more non-phase coded archive 1 (I&Q) datasets are phase coded and then overlaid to create a single dataset with multiple trip echoes. The moments of the original non-phase coded data are used as truth in evaluating the moment recovery algorithms. The advantages of such a procedure are that real data are used and no temporal or spatial adjustments must be made to compare the decoded data against the original "true" data. In this paper the method of creating these hybrid datasets is discussed as well as showing performance statistics for some of the moment recovery algorithms.