Initial Results
Introduction
Currently, we are working on the bright band identification algorithm using Python 3 (Objective 1). Although it may be easy to find the bright band by eyes, doing this identification by programming is still necessary. Because it would be time-consuming to analyse a one-year dataset manually, and human eyes may ignore some important details.
One of the methods, generally adopted to identify and correct the bright band signal, is to compute the vertical profile of reflectivity (VPR) by analysing the operational radar data (Illingworth and Thompson, 2011). Figure 2 shows the standard profile of VPR. The basic idea of the VPR method is that the bright band location is where the VPR peaks.
Another method is to analyse the Doppler radial velocity measured by the radar. As the Copernicus is a vertically pointing radar, the fall velocity of a particle is its radial Doppler velocity. According to literature, such as Battan (1973), the fall velocity of the hydrometeors increases in the melting layer as ice melts to liquid water and becomes denser. Hence, the bright band should be located at where the velocity is rapidly increasing (the snow is melting) (as shown in Figure 3). The bright band top is where the increasing/melting starts, and the bright band bottom is where the increasing/meting stops.
Figure 3: Sample plot of height profile of Doppler velocity for vertically pointing radar during a bright band event. $V_s$ is the velocity of snow and $V_r$ is the velocity of rain. The bright band height is near 4.5 km. (Described in detail by Sarma et al. (2016).)
In practice, however, these vertical profiles are variable in space and time. The video below shows the time series of SNR (signal to noise ratio), Radar Reflectivity Factor ($dBZ$), Spectral Width (SPW, enlarged 10 times here) and Doppler velocity ($v$) during a bright band precipitation on 12th March 2017 at CFARR. Note that the bright band height in this case is around 2000 m. More of this video can be found here (begins at 2:00). This longer version also contains another case - the Storm Doris (23rd February 2017).
Although the vertical profiles of 4 variables are shown here, we currently only focus on $dBZ$ and $v$. Because the profiles of SNR and SPW are essentially the same as the profiles of $dBZ$ and $v$ respectively (Pfaff et al., 2014; Emory et al., 2014).