324
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
represent rain physics much better than radar observations, which come from the
indirect scattering effect of raindrops. Therefore, the comparison between in situ
observations and radar–rain estimations could give an objective evaluation of the
rain estimator.
13.4.1 diSdRoMeteR
The disdrometer is an effective tool for rain microphysical study, because it can
measure DSDs. The traditional disdrometer is the impact type (e.g., Joss–Waldvogel
disdrometer), which is designed based on the measurement of raindrop momentum
(Tokay et al. 2001). Recent disdrometers apply the optical technique, for example,
the one-dimensional laser optical disdrometer (Parsivel disdrometer) and the twodimensional video disdrometer (2DVD; Kruger and Krajewski 2002). The disdrometer with the optical technique provides not only more accuracy but also additional
measurements of the shapes and falling velocities of the raindrops.
Figure 13.7 shows an example of a comparison between radar retrievals and dis3.7 shows an example of a comparison between radar retrievals and dis.7 shows an example of a comparison between radar retrievals and disdrometer observations on May 2, 2005. The disdrometer data were collected by a
2DVD deployed at ~28 km south of the radar. The disdrometer has a high resolution
(0.132 mm) and a sampling area of ∽100 cm 2 in measuring raindrops. It uses 41 size
bins with a bin width of 0.2 mm, indicating a range of 0–8.1 mm in diameter for
raindrop measurements. The radar data were collected by KOUN. The data have
been filtered by eliminating nonrain echoes, that is, using the threshold of correlation
coefficients larger than 0.9. The data also have been smoothed using measurements
at five range gates. The retrieval was based on the direct approach mentioned in the
previous section. Specifically, the retrieval applied the C-G DSD model with a constraint relation updated by Cao et al. (2008).
There are three different rain variables, R, D 0 , and N T , which are compared in
Figure 13.7. R is approximately proportional to the 3.67th moment of the DSD,
while N T is equivalent to the 0th moment of the DSD. In addition, D 0 is related
to the third-order distribution of DSD. The single-parameter DSD model, which
is intrinsically assumed by traditional R–Z relations, cannot provide reasonable
retrievals for all these rain variables. The C-G DSD model, using two parameters,
provides more flexibility in rain estimation. As the figure shows, the temporal
variation of radar retrieval matches the disdrometer observation well for all the
variables.
It is worth noting that the sampling volume difference should be considered during the validation of radar–rain retrieval. For example, the typical sampling volume
of a 2DVD within a 1-min interval is about 3–5 m 3 . For KOUN, however, the typical
sampling volume at 28 km is about 0.05 km 3 . If the inhomogeneity of rain is strong,
the sampling volume difference can cause a large difference in measurements by
two instruments. For the example shown in Figure 13.7, it was a stratiform precipita3.7, it was a stratiform precipita.7, it was a stratiform precipitation during 1100–1330 UTC. Most values of radar reflectivity were 25–35 dBZ. The
rain was less likely inhomogeneous, and the effect of sampling volume difference
is not remarkable for the comparison. This example demonstrates the application of
disdrometer in the validation of radar–rain retrieval. On the other hand, this example
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
represent rain physics much better than radar observations, which come from the
indirect scattering effect of raindrops. Therefore, the comparison between in situ
observations and radar–rain estimations could give an objective evaluation of the
rain estimator.
13.4.1 diSdRoMeteR
The disdrometer is an effective tool for rain microphysical study, because it can
measure DSDs. The traditional disdrometer is the impact type (e.g., Joss–Waldvogel
disdrometer), which is designed based on the measurement of raindrop momentum
(Tokay et al. 2001). Recent disdrometers apply the optical technique, for example,
the one-dimensional laser optical disdrometer (Parsivel disdrometer) and the twodimensional video disdrometer (2DVD; Kruger and Krajewski 2002). The disdrometer with the optical technique provides not only more accuracy but also additional
measurements of the shapes and falling velocities of the raindrops.
Figure 13.7 shows an example of a comparison between radar retrievals and dis3.7 shows an example of a comparison between radar retrievals and dis.7 shows an example of a comparison between radar retrievals and disdrometer observations on May 2, 2005. The disdrometer data were collected by a
2DVD deployed at ~28 km south of the radar. The disdrometer has a high resolution
(0.132 mm) and a sampling area of ∽100 cm 2 in measuring raindrops. It uses 41 size
bins with a bin width of 0.2 mm, indicating a range of 0–8.1 mm in diameter for
raindrop measurements. The radar data were collected by KOUN. The data have
been filtered by eliminating nonrain echoes, that is, using the threshold of correlation
coefficients larger than 0.9. The data also have been smoothed using measurements
at five range gates. The retrieval was based on the direct approach mentioned in the
previous section. Specifically, the retrieval applied the C-G DSD model with a constraint relation updated by Cao et al. (2008).
There are three different rain variables, R, D 0 , and N T , which are compared in
Figure 13.7. R is approximately proportional to the 3.67th moment of the DSD,
while N T is equivalent to the 0th moment of the DSD. In addition, D 0 is related
to the third-order distribution of DSD. The single-parameter DSD model, which
is intrinsically assumed by traditional R–Z relations, cannot provide reasonable
retrievals for all these rain variables. The C-G DSD model, using two parameters,
provides more flexibility in rain estimation. As the figure shows, the temporal
variation of radar retrieval matches the disdrometer observation well for all the
variables.
It is worth noting that the sampling volume difference should be considered during the validation of radar–rain retrieval. For example, the typical sampling volume
of a 2DVD within a 1-min interval is about 3–5 m 3 . For KOUN, however, the typical
sampling volume at 28 km is about 0.05 km 3 . If the inhomogeneity of rain is strong,
the sampling volume difference can cause a large difference in measurements by
two instruments. For the example shown in Figure 13.7, it was a stratiform precipita3.7, it was a stratiform precipita.7, it was a stratiform precipitation during 1100–1330 UTC. Most values of radar reflectivity were 25–35 dBZ. The
rain was less likely inhomogeneous, and the effect of sampling volume difference
is not remarkable for the comparison. This example demonstrates the application of
disdrometer in the validation of radar–rain retrieval. On the other hand, this example