A Simple Method for Separating Weather from Non-Weather echoes on Dual-Polarization Radars

Researchers at the Wildlife branch of Environment and Climate Change Canada dreams of using weather radar networks to monitor bird migrations and warn for their associated hazards. To that end, we have been tasked to devise ways to isolate biological echoes from annoying weather echoes. With sea clutter and occasionally chaff, biological echoes are the dominant non-weather echoes in these days of clutter-filtered radar data; the problem of isolating biological echoes is hence largely tantamount to identifying non-weather echoes at low levels. A key constraint for this exercise was that whatever method would be easy to implement for non-radar experts and that it would work both on Canadian and US radars, and that it would work better than currently implemented algorithms that encounter difficulties in some migration conditions (e.g., Stepnanian el al. 2016).

What we found was that instead of using correlation and differential reflectivity as separate inputs, it was extremely advantageous to combine them using the simultaneous transmit and receive derived circular Depolarization Ratio (SDR) of Melnikov and Matrosov (2013). SDR estimates the non-sphericity of targets and as such is very useful in segregating the generally nearly spherical meteorological targets (low SDR) from non-spherical non-meteorological targets (high SDR). Using SDR on Canadian and US radars, including cases that cause difficulties to the operational target ID approach used on WSR-88Ds, we correctly identified precipitation and non-precipitation pixels with a 96% accuracy, compared to 75-85% for tests using texture of PhiDP, Zdr, or Doppler velocity. Most of the incorrectly classified precipitation pixels come from the bright band, high reflectivity areas (probably hail), and second trip echoes etc., and tend to be sprinkled, as a result of which they can be greatly reduced with a simple despeckling algorithm. The simplicity of this algorithm, its applicability across different radar bands, and the fact that it does not require training or tuning makes it particularly useful to non-experts looking for an easy solution to separate weather from non-weather echoes.