Performance Assessment of the SZ(8/64) Phase Code to Separate Weather Signals in a Research WSR-88D

Abstract:

A distant squall line and a nearby isolated convective storm were located such that both appeared to be located at the same range (i.e., echoes from the respective storms are overlaid) when viewed by the NSSL research WSR-88D radar operating in the short PRT Doppler mode. Under these conditions, the normal WSR-88D signal processing yields highly censored data (i.e., in this case, about 90% of the data in a sector was censored). By phase coding the transmit pulse with the SZ(8/64) code, and processing the time-series data with newly developed algorithms to separate the trip-1 and trip-2 echoes, we have unraveled the velocity fields from the respective storms. Doppler velocity fields of overlaid echoes can be separated even when the weaker data field is 30-40 dB below the stronger one. Using the developed algorithms, it appears that almost all of the regions previously censored by WSR-88D have recoverable velocities. Reflectivity is more difficult to separate. When the spectrum width of the competing echo is very narrow, accurate reflectivity estimates can be obtained for signals over 40 dB below the level of the stronger overlaid echo. This effectively removes the effects of trip-1 ground clutter on most trip-2 power estimates.

Instances where separation of the two overlaid velocity fields is unsatisfactory are uncommon in the data we have examined. We observe a few instances where the spectrum width of a competing echo is very large, or has an unusual non-Gaussian shape, and the residual power of the out-of-trip echo which leaks into the currently processed trip may be only slightly less than it's unfiltered value. For most of the data analyzed, where spectrum widths and shapes are more typical, the competing echo power is suppressed by more than 20 dB. This is usually sufficient to produce good reflectivity estimates. Although the separation of echo power from overlaid echoes is more likely to be incomplete, the SZ filtering process substantially improves the echo power estimates for both trips. We are attempting to evaluate the accuracy of the separation in a variety of situations, and to identify the occasional instances when this method fails. Examples will be presented which demonstrate the performance obtained so far. Considerations will be discussed which appear to be important for implementing this method on the WSR-88D.