A Novel Bandstop Regression Filter with Application to Range–Velocity Mitigation for Weather Radar Data

Abstract Regression filters are typically used to either extract the trend of a signal (low-pass filter) or to estimate the high-frequency content of a signal by subtracting the estimated trend from the original signal (high-pass filter). However, it is possible to also use a regression least squares fit as a bandstop (or bandpass) filter. The technique for doing this is described. The new regression bandstop filter offers distinct advantages over traditional finite impulse response (FIR), infinite impulse response (IIR), and spectral-based filters, especially for shorter-length sequences, that can result in improved data quality. The advantages of the new regression bandstop filter are demonstrated via processing of SZ(8/64) phase-coded data. The SZ(8/64) is a technique used by Next Generation Weather Radar (NEXRAD) for extending the unambiguous velocity range. Specifically, the regression bandstop filter is used to suppress the strong-trip echo (i.e., most power) so that the velocity of the overlaid weak-trip echo can be estimated. Significance Statement The novel bandstop filter presented here shows how a high-pass regression filter can also be used as an arbitrary-frequency bandstop filter. For applications where bandstop filtering is required for short-length time series, this new filter has several significant advantages over the more traditional filters, such as finite impulse response (FIR), infinite impulse response (IIR), and spectral filters: no filter initialization, no ringing, no phase delay, and no attenuating window functions. One application is in SZ(8/64) phase coding, where a bandstop filter is required to eliminate the strong-trip echo. The new bandstop filter can significantly reduce measurement standard deviations.