3 Screening out weather

3.1 using correlation coefficient

Screening precipitation based on correlation coefficient is arguably the most simple and established approach for screening out weather in cases where dual-polarization data is available. It is based on removing data above a certain \(\rho_{/HV}\) tresholds, most commonly 0.95.

# We will screen out the reflectivity areas with for which the
# correlation coefficient is > 0.95 (indicating precipitation)
# We do so by redefining parameter DBZH, replacing pixels with NA values
# whenever RHOHV > 0.95
my_ppi_clean <- calculate_param(my_ppi, DBZH = ifelse(RHOHV > 0.95, NA, DBZH))
# plot the original and cleaned up reflectivity:
map(my_ppi_clean, param = "DBZH", zlim = c(-20, 40), map="osm")

3.2 using MistNet

You can also use MistNet for screening out weather when * the radar operates at S-band wavelengths * you only have single polarization data available (specifically, the three basic quantities radial velocity VRADH, reflectivity DBZH, and spectrum width WRADH).

# apply the MistNet model to the polar volume file and load it as a polar volume (pvol):
my_pvol <- apply_mistnet(my_pvolfiles[1])
# mistnet will add additional parameters to the
# elevation scans at 0.5, 1.5, 2.5, 3.5 and 4.5 degrees
# let's extract the scan closest to 0.5 degrees:
my_scan <- get_scan(my_pvol, 0.5)
# plot some summary info about the scan to the console:
my_scan

MistNet adds several new parameters:

  • WEATHER: a probability (0-1) for the weather class
  • BIOLOGY: a probability (0-1) for the biology class
  • BACKGROUND: a probability (0-1) for the background (empty space) class
  • CELL: the final segmentation, which is based on the WEATHER parameters of all five MistNet elevation scans. CELL also includes a border calculated by a region growing approach. CELL values equal 1 for the additional border, and 2 for the originally segmented precipitation area by MistNet.
# as before, project the scan as a ppi:
my_ppi <- project_as_ppi(my_scan)
# plot the probability for the WEATHER class
plot(my_ppi, param = 'WEATHER')
# plot the final segmentation result:
# plot the probability for the WEATHER class
plot(my_ppi, param = 'CELL')
# let's remove the identified precipitation area (and additional border) from the ppi, and plot it:
my_ppi_clean <- calculate_param(my_ppi, DBZH = ifelse(CELL >= 1, NA, DBZH))
map(my_ppi_clean, param = 'DBZH')

Exercise 4: Calculate and plot a ‘cleaned up’ PPI for the radial velocity that includes only the segmentation by MistNet and not the additional border. Remember that the segmentation by Mistnet has a CELL value 2, and the border has a CELL value 1.