328
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
performed, the rainfall rate estimated from contaminated Z H and Z DR would be much
larger than 100 mm h –1 . In situ measurements in the figure, however, show that this
is not the case. Radar retrievals in Figure 13.9 demonstrate that radar data quality
control (i.e., using radar measurements [classified as rain] from an adjacent area to
interpolate over a hail-contaminated region) can provide a reasonable rain estimate
for the contaminated region.
13.5 CONCLUSIONS
This chapter addresses radar polarimetry for rain estimation, including its basis and
methods. Compared with rain estimation based on single-polarization radar measurements, dual-polarization radar observations can improve rain estimation with
a better representation of DSD variability. The study presents two major methods
for polarimetric radar rain estimation, including the empirical method and the DSD
retrieval method. For the empirical method, three estimators, R(Z h ,Z dr ), R(K DP ), and
R(K DP ,Z dr ), are commonly used. These estimators have their own advantages and disadvantages. This chapter emphasizes the DSD retrieval method, because it is more
flexible in obtaining rain variables than the empirical method. Besides the direct
approach, two other recently proposed approaches are discussed. The Bayesian and
variational approaches optimize the use of polarimetric radar measurements. The
Bayesian approach applies the historical statistical information of rain, while the
variational approach uses the spatial information of rain. Both approaches can utilize
multiple observations, making the algorithms extendable for more variables. Thus
far, there are still issues to be worked out for these two approaches. The major issues
include the accuracy of radar forward models and the quality of data. However, the
concept of optimization of radar observations would be meaningful in rain estimation. In this context, a two-order C-G DSD model was applied to illustrate the three
approaches of DSD retrieval. It is worth noting that these approaches are also applicable for other DSD models if the necessary revisions are made.
It can be concluded that, first, the characterization of rain/hail/snow microphysics should be studied further. Any progress on the modeling of DSD and/or radar
variables could strengthen the estimation/retrieval algorithms. In addition, classification of hydrometeors is helpful for obtaining the correct rain estimation. Second,
rain estimation would benefit from higher quality radar data. The data quality could
be improved either from the upgrade of radar hardware or applying proper quality
control algorithms. Third, rain estimation would be improved with the optimal use
of multiple radar observations. Minimization of the error effect would be the goal
of this approach. In any circumstances, radar polarimetry is a promising way for
accurate rain estimation.
ACKNOWLEDGMENT
The work was supported by the National Science Foundation under Grant ATM0608168. The authors thank Drs. Terry Schuur and Edward Brandes for providing
radar and disdrometer data.
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
performed, the rainfall rate estimated from contaminated Z H and Z DR would be much
larger than 100 mm h –1 . In situ measurements in the figure, however, show that this
is not the case. Radar retrievals in Figure 13.9 demonstrate that radar data quality
control (i.e., using radar measurements [classified as rain] from an adjacent area to
interpolate over a hail-contaminated region) can provide a reasonable rain estimate
for the contaminated region.
13.5 CONCLUSIONS
This chapter addresses radar polarimetry for rain estimation, including its basis and
methods. Compared with rain estimation based on single-polarization radar measurements, dual-polarization radar observations can improve rain estimation with
a better representation of DSD variability. The study presents two major methods
for polarimetric radar rain estimation, including the empirical method and the DSD
retrieval method. For the empirical method, three estimators, R(Z h ,Z dr ), R(K DP ), and
R(K DP ,Z dr ), are commonly used. These estimators have their own advantages and disadvantages. This chapter emphasizes the DSD retrieval method, because it is more
flexible in obtaining rain variables than the empirical method. Besides the direct
approach, two other recently proposed approaches are discussed. The Bayesian and
variational approaches optimize the use of polarimetric radar measurements. The
Bayesian approach applies the historical statistical information of rain, while the
variational approach uses the spatial information of rain. Both approaches can utilize
multiple observations, making the algorithms extendable for more variables. Thus
far, there are still issues to be worked out for these two approaches. The major issues
include the accuracy of radar forward models and the quality of data. However, the
concept of optimization of radar observations would be meaningful in rain estimation. In this context, a two-order C-G DSD model was applied to illustrate the three
approaches of DSD retrieval. It is worth noting that these approaches are also applicable for other DSD models if the necessary revisions are made.
It can be concluded that, first, the characterization of rain/hail/snow microphysics should be studied further. Any progress on the modeling of DSD and/or radar
variables could strengthen the estimation/retrieval algorithms. In addition, classification of hydrometeors is helpful for obtaining the correct rain estimation. Second,
rain estimation would benefit from higher quality radar data. The data quality could
be improved either from the upgrade of radar hardware or applying proper quality
control algorithms. Third, rain estimation would be improved with the optimal use
of multiple radar observations. Minimization of the error effect would be the goal
of this approach. In any circumstances, radar polarimetry is a promising way for
accurate rain estimation.
ACKNOWLEDGMENT
The work was supported by the National Science Foundation under Grant ATM0608168. The authors thank Drs. Terry Schuur and Edward Brandes for providing
radar and disdrometer data.