Environmental Engineering Reference
In-Depth Information
but
W
dp
is extremely noisy and consequently K
dp
will be even noisier (see Equation 7.4). To decrease
the noise,
W
dp
is averaged for several kilometres
along the beam. Ryzhkov and Zrnic (1996) have
even suggested averaging in range over 2.5 km and
7.5 km approximately (17 and 49 gates, respec-
tively, with a resolution of 150m) for a threshold
in reflectivity of 40 dBZ. This obviously leads
to a considerable loss in resolution over the
conventional R(Z
h
) rainfall estimator. Brandes
et al. (2001) carried out an analysis between rain-
gauge observations and rainfall rates estimated
from K
dp
and Z
h
, and they found similar bias
factors and correlation coefficients between both
estimators, concluding that no obvious benefit is
obtained in using K
dp
to estimate rainfall rates
over using Z
h
from a well-calibrated radar.
Therefore, there is controversy whether or not
polarimetry is going to improve radar rainfall
estimates. Illingworth (2003) suggested that at the
2-km scale needed for an operational environ-
ment, the additional information provided by Z
dr
and K
dp
is not sufficiently accurate to improve
rainfall estimates. However, some improvement
in the precipitation estimates from polarimetric
radar measurements may be realized by not only
applying one particular rain estimator, but also
exploiting the attributes of the many different
polarimetric algorithms available depending on
the circumstance. Ryzhkov andGiangrande (2004)
proposed a 'synthetic' algorithm, which makes
use of different combinations of rain rate algo-
rithms depending on the rain rate estimated
using only the conventional R(Z) relationship.
They proposed the use of an algorithm of the form
R(R(Z
h
), Z
dr
) for low rain rates (R
into the operational domain is still to be proven on
the basis of sound operational experience.
In addition, polarimetric radar measurements
offer the possibility to classify hydrometeors
(Zrnic and Ryzhkov 1999; Vivekanandan
et al. 1999; Liu & Chandrasekar 2000; Zrnic
et al. 2001), which provides the possibility of
applying different rainfall estimators depending
on the classification. However, the operational
performance of such radars in practice is again
still to be established.
Problems Associated with the Estimation
of Precipitation
The estimation of precipitation using weather
radars is subject to a variety of error sources. In
the previous section the importance of the DSD
has been described in relating the reflectivity fac-
tor Z
h
(or any of the polarimetric variables Z
dr
, K
dp
)
to the rainfall rate R. However, uncertainties in
the knowledge of the DSD may not be the largest
source of errors in radar rainfall measurements
(Joss and Waldvogel 1990) and there are additional
errors that may require evenmore attention. Atlas
et al. (1984) concluded that the average deviation
in the rain rate estimation from reflectivity
measurements due to DSD variability would be
33%, whereas Doviak and Zrnic (1993) suggest
errors of 30-35%. However, Joss and Waldvo-
gel (1990) suggest that after averaging over space
and time, the errors in rainfall estimates due to the
variability of theDSDrarely exceed a factor of two.
Problems associated with the variation of the
vertical reflectivity profile of precipitation may
be one of the largest sources of error (Fig 7.1). As
the range increases from the radar, the radar beam
is at some height above the ground. The hydro-
meteors intercepted by the radar beam may then
be composed of raindrops, melting snowflakes,
snowflakes, hail, etc. This variability affects re-
flectivity measurements and the estimation of
precipitation is not representative of the rainfall
rate at the ground. This variation is due to the
growth or evaporation of precipitation, or change
of phase, in particular melting, where a layer of
<
6mmh
1
), of
the form R(R(K
dp
), Z
dr
) for medium rain rates (6
<
R
<
50mmh
1
), and the algorithm R(K
dp
)
for
>
50mmh
1
). Although the
algorithms proposed by Ryzhkov and Gian-
grande (2004) are slightly different from the rain-
fall rate estimators presented in this section, it is
clear that by exploiting the performance of differ-
ent relationships R(Z
h
), R(Z
h
, Z
dr
), R(K
dp
) and
R(K
dp
, Z
dr
), it may be possible to improve the
estimation of rainfall using dual-polarization
radars. However, the move from research radars
high rain rates (R
