The separation of noise and signal components in Doppler RADAR returns [presentation]

Knowing the correct noise value in a Doppler radar return is essential for (a) computing moments with good data quality (Ivic and Torres 2010), (b) optionally censoring (i.e. setting to missing) data which contains noise only and (c) contributing to a data quality metric (Friedrich et. al 2006, Osrodka et. al 2010). The receiver noise, however well calibrated a radar may be, will drift over time. Clutter, weather and water vapor emit radiation at all wavelengths, and these emissions will add to the thermal noise. This is especially problematic at shorter wavelengths, such as Kaband and W-band. Radar moments in noise-only regions have well-known statistical properties. For stationary radars (such as vertically pointing instruments) it is possible to compute these statistics from a single gate over time. For a scanning radar we need to consider the statistics from a number of adjacent gates, substituting variability in space for variability in time. In this paper we present a method to identify noise regions in data from a scanning radar, utilizing the known behavior of returned power and radial velocity in noise. We show that the method is applicable to radars over a range of frequencies, from S-band to Ka-band. We demonstrate that the method can robustly identify noise regions, allowing us to compute the noise on a ray-by-ray basis. The radar moments for individual beams can therefore be effectively adjusted.