It is a common practice with SIGMET signal processors (and even from other radars such as research radars and operational data such as NEXRAD) to use the Doppler velocity spectrum in an attempt to filter ground clutter from other fields (which ideally has a near 0 velocity spectrum). Clutter (reflections off of objects, both stationary and moving targets such as mountains, aircraft, buildings, cars, etc.) is an annoyance because it contaminates the presentation of the radar images as well as causes errors in radar retrievals (velocity for severe weather interpretation, precipitation estimates, etc.). One idea is to remove the clutter by filtering the time series (pulse by pulse) by removing returns with near zero velocity. This is desirable because the remaining signal will be meteorological echo (which is usually moving towards or away from the radar). However, sometimes it’s not, and here are some examples where it can cause problems.
Here is an example from a NEXRAD PPI near Cleveland in widespread precipitation (snow in this case), showing the reduction in reflectivity factor near the radar where the velocity is near 0. You can see this effect nearly every time there is stratiform rain near a NEXRAD. The algorithm must have a range or altitude dependence, since the effect usually goes away after 15 km or so. However, if you’re trying to use reflectivity to estimate precipitation, the value of Z is missing some meteorological echo. It’s probably better than possibly including clutter (which will blow up your estimates of precipitation or mean particle size), but not ideal.
The problem is that this removes power from meteorological echo, and thus can bias Z and Zdr measurements in these regions. Here is an example of this in action: in this RHI scan a 0 isodop (an isodop is a surface of constant Doppler velocity) filter is able to remove low level clutter, but also removes valid data near the 0 isodop. When designing radar scanning strategies, the radar meteorologist must be aware of these settings for each sweep, and for research measurements it may be advisable to use polarimetric methods of QC rather than using the velocity spectra.
In this animation, you can see that in the “quality controlled” reflectivity (the image without the clutter and clear air echo aloft, there is power missing where the mean Doppler velocity is near 0 (especially below the freezing level in several broad horizontal regions), as shown via the black areas in the image below:
This issue impacts the spectral width estimate, making the spectrum artificially wide (biased high) in regions where the near 0 velocities are removed in the spectrum (keeping only the tails):
It also influences the differential reflectivity (Zdr), making the values negative since more power is preferentially removed from the horizontal reflectivity compared with the vertical reflectivity:
Also impacted is the differential phase, which has obvious artifacts due to the removal of phase shift spectra:
Several authors have released a paper on spectral time domain clutter filtering, including the following:
Nguyen, Cuong M., Dmitri N. Moisseev, V. Chandrasekar, 2008: A Parametric Time Domain Method for Spectral Moment Estimation and Clutter Mitigation for Weather Radars. J. Atmos. Oceanic Technol., 25, 83–92.
J. C. Hubbert, M. Dixon, S. M. Ellis, G. Meymaris. (2009) Weather Radar Ground Clutter. Part I: Identification, Modeling, and Simulation. Journal of Atmospheric and Oceanic Technology 26
Online publication date: 1-Jul-2009.
. Full Text
. PDF (2852 KB)
J. C. Hubbert, M. Dixon, S. M. Ellis. (2009) Weather Radar Ground Clutter. Part II: Real-Time Identification and Filtering. Journal of Atmospheric and Oceanic Technology 26
Online publication date: 1-Jul-2009.
. Full Text
. PDF (11134 KB)
and there is even a patent on the technique outlined in the first paper! It is not known at this time what specific filtering technique was used on the data displayed here. The impact is the following: The “quality controlled” reflectivity field, and other fields are impacted with biases, that must be identified and removed from quantitative retrievals. In addition, this quality control process deleted data in many of the other fields, which along with biases in the measured fields hampers the use of dual-polarization variables for quality control (including correlation coefficient and standard deviation of differential propagation phase). We have the uncorrected reflectivity, but we can’t effectively use the polarimetric variables to correct the This cannot be undone in recorded data, so unless one is careful, this issue can cause issues with your dataset.
Carey, L. D., S. A. Rutledge, D. A. Ahijevych, and T. D. Keenan, 2000: Correcting propagation effects in C-band polarimetric radar observations of tropical convection using differential propagation phase. J. Appl. Meteor., 39, 1405–1433. [Abstract]
Marks, David A., David B. Wolff, Lawrence D. Carey, Ali Tokay, 2011: Quality Control and Calibration of the Dual-Polarization Radar at Kwajalein, RMI. J. Atmos. Oceanic Technol.
, 181–196. doi: http://dx.doi.org/10.1175/2010JTECHA1462.1
Clearly improving time series/spectral analysis is an active area of research, so stay tuned for improved algorithms. Note that dual-pol QC methods aren’t perfect either, but that topic will be saved for another post.
Software credit: ARM-PyART, Argonne National Lab