Extended Abstract (284K)11B.3
Data driven adaptive identification and suppression of ground clutter for weather radar
Svetlana M. Bachmann, Lockheed Martin MS2, Syracuse, NY; and M. D. Tracy
An innovative approach for the suppression of ground clutter contributions from weather radar data is presented. Most of the existing adaptive spectral clutter filters need a priore knowledge of clutter position, strength and its spectral characteristics. This knowledge is irrelevant for the presented approach because the identification and suppression are entirely data driven.
Judicious spectral analyses are used to construct a combination of phase-based and power-based identification schemes that pinpoint the exact spectral coefficients containing signatures that are primarily caused (or dominated by) ground clutter returns.
The phase-based identification scheme locates the contributions to a spectrum with characteristically small Doppler shifts, and consequently determines the spectral width for the suspected clutter component. The differential phase is determined from phase difference between complex spectral coefficients of two subsequences obtained by splitting data on two-poli-phase decompositions (subsequences with the even-indexed and odd-indexed portions of the original time series samples).
The power-based identification scheme locates the contributions to a spectrum with characteristically large complex amplitudes. For that, a small portion of equidistant samples of the original time series is used to construct a spectrum with a reduced span of Doppler velocities, with the same spectral resolution and presumably preserved clutter signature in terms of power spectral densities.
The suppression is performed only on the spectral coefficients identified by both power and differential phase as ground clutter. For the most part, the complex amplitudes of clutter spectral coefficients of the original spectrum are similar to those of the reduced-Doppler-span spectrum. Therefore, these can be used for the precise suppression via subtraction instead of the conventional notching.
The scheme is performed in the frequency domain and requires a sufficient number of samples for spectral processing. For each resolution volume three Doppler spectra are estimated: one complete spectrum, and two reduced-Doppler-span spectra for two-poli-phase decompositions. For dual polarization radar the amount of computations doubles. The resulting “clutter filter” can be characterized as extremely sensitive, non-symmetric, and non-identical-in-polarization-channels. The amount of filtering can be adjusted by changing values of thresholds of differential phase and power used for the phase-based and power-based identification. The presented suppression technique induces minimal disturbance to the spectral polarimetric fields and therefore allows for more accurate computation of polarimetric variables used for echo classification schemes. Presented approach can also be used to construct clutter maps indicating clutter power and intrinsic spectral width of clutter.
Presented approach was tested on data acquired with two radars: a rotating parabolic antenna of KOUN (S-band polarimetric prototype weather radar in Norman, OK) and a passive phased array of NWRT (National Weather Radar Testbed in Norman, OK).
Session 11B, Radar Applications - Session II - Part II
Wednesday, 14 January 2009, 4:00 PM-5:30 PM, Room 122BC
Previous paper Next paper