Geoscience Reference
In-Depth Information
4.5 QPE Summary in terms of bias and scatter in dB
For the three mini-radar sites discussed in the previous paragraphs, Table 5 provides a
summary of the QPE evaluation in terms of Bias in dB (as a function of the number of
considered hours). When all the wet and dry hours are considered, we have the so-called
“overall” Bias: it includes False Alarm events (mainly caused by ground clutter contamination)
and Missing Detection (caused by beam overshooting, see Sec. 1.2 and, in some cases,
attenuation) events; it is certainly the most resistant and complete definition of Bias. It is
obviously a measure of the mean error and says nothing about the error dispersion around the
mean. For this purpose, there is the Scatter, which will be presented in the Table 6.
Fig. 18. As for Fig. 16, but for but for 41 samples with both R and G 0.4 mm/h.
While the “overall” Bias (1
st
column) includes both wet and dry periods, the other two Bias
(2
nd
and 3
rd
column) are conditional upon rain: the “Wet-Gauge” Bias considers only hours
where the gauge amounts are larger than 0.4 mm/h; finally, the definition of “wet-wet” Bias
reduces further the number of hours used in the calculus: in such case, only hours with both
radar and gauge amounts larger than 0.4 mm/h are used.
Site “Overall”
Bias
“Wet-Gauge”
Bias
“Wet-Wet”
Bias
Palermo radar
0.9 dB
(144 h)
1.1 dB
(48 h)
0.7 dB
(38 h)
Bisacquino radar
1.0 dB
(96 h)
1.2 dB
(37 h)
0.9 dB
(29 h)
Torino radar
0.3 dB
(126 h)
0.4 dB
(63 h)
+0.5 dB
(41 h)
Table 5. QPE evaluation summary in terms of Bias for three mini-radar sites. The number of
hours of each data set are given in parentheses.
