Quantitative Effects of Clutter Highpass Filtering as Used by DWD

Clutter removal in weather radar images using digital highpass filtering in the time or spectral signal spaces has become a standard tool in routine networks, as operated in Germany by the German Meteorological Service (DWD). The DWD Weather Radar Network has been in full operation since 1999, featuring 12 high coherency Doppler and 4 non-Doppler Radars. One of its main goals is quantitative determination of precipitation amount.

While this has been effected by Z/R conversion and occasional offline gauge adjustment in the past, a line of "physical corrections" has been followed in the last few years, including calibration and hardware maintenance (see Mammen, this vol.), raw data quality tests (thresholding), clutter filtering, and, in future, VRP correction, prior to a realtime gauge adjustment (see Weigl, this vol.) as a last polish. Each one of those correction schemes constitutes a compromise of the desired benefit against detrimental sideeffects, depending on what is considered the "truth".

In the case of clutter filtering, (Log-) amplitude-, (Doppler-) time series - or (Doppler-) FFT-filters are applied in order to estimate the relative clutter power CCOR, which is quantitatively corrected for: corr. Z [dBZ]=uncorr. Z [dBZ] - CCOR [dB] and supplemented by decision filters to threshold for high CCOR, low Log-intensity (absolute and adaptive), relative IQ-intensity, (Doppler-)signal quality SQI, weather signal power WSP, micro clutter, and speckles.

This effectively cancels the clutter power in reflectivity measurement of precipitation and removes the bias toward zero in radial wind measurement.

However, adverse effects occur whenever the precipitation spectrum and IIR-filter notch overlap, which happens regularly with the radar beam tangential to the drop motion, giving -besides a decrease in signal quality SQI- rise to a loss of measured precipitation power and to a bias towards the folding boundaries in the radar estimate of mean radial velocity.

This drawback is largely mitigated if frequency domain filtering (FFT) can be performed instead of time series (PP/IIR) filtering which is not the case if high scan rates and dual PRF unfolding are required.

Introducing ground truth measurements by gauge adjustment of the clutter corrected data may also change the picture.

The quantitative improvement by clutter correction in radar precipitation amount and in radial wind, as well as the errors introduced by this technique, both depending on precipitation and scan parameters (mean velocity, spectral width; azimuth rate, PRF), will be assessed using calculations based on model and operational radar data.