At the Finnish Meteorological Institute we have started to test estimates of POD of ground level precipitation applying three different methods:
(1) The visibility of precipitation (V) can be estimated as a function of range (r) applying the measurement geometry of radar beam, minimum detectable dBZ and high resolution measured vertical profiles of reflectivity (VPR) from the polar radar data at close ranges to each radar. By using Gaussian beam convolution of the VPR at all ranges applying the known lowest elevation angle we obtain a single value for the maximum distance of detection. Probability of precipitation detection at each range is obtained by repeating the convolution procedure for an ensemble of VPRs. The ensemble members can be obtained from the time series of actual measurements in a network of radars, using climatological VPRs or by generating simulated VPRs from the measured VPRs.
(2) The actually observed POD can be quantified at the ranges where overlapping radar pairs measure the same precipitation area with the lowest elevation PPI. The close range radar (1) measures almost at the ground level (which represents well the actual precipitation) diagnosing the area of precipitation A(1) whereas the distant radar (2) detects only part of the precipitating area A(2). The ratio A(2)/A(1) is a measure of POD at the average range r, which is the distance from radar 2 to the center of the area of comparison.
(3) The observed POD as a function of range can be estimated also from the ratio f = A(p)/A(np) where A(p) is the area of precipitation and A(np) the area of no precipitation in a circular range belt r2 - r1 from a radar. If the precipitation coverage fraction (f) is horizontally homogeneous the decrease of f as a function of range will measure the quantity POD(r).
When POD(r) is presented as a quantitative shade underlay on an operational precipitation product the users immediately recognize at least semi-quantitatively the detection probability of precipitation in each pixel of the composite image. We present real examples how widely POD(r) can vary from day to day.