321
Radar Polarimetry for Rain Estimation
Other radar measurements such as radar reflectivity and differential reflectivity are
also smoothed sometimes (e.g., over a 1-km range) before the rain estimation. Lee et
al. (1997) introduced a speckle filter technique, which could be used for radar applications (Cao et al. 2010). In general, the smoothing would lower the range/angular
resolution of radar measurements. However, it is useful to obtain a better rain estimation with a smaller variance.
13.3.3.2 Clutter Filtering
Ground clutter is attributed to the side-lobe effect of the radar antenna. The sidelobe effect is strong for low-elevation scanning. Clutter is usually observed in the
area close to the radar, but it is sometimes measured at a farther range due to the
effect of abnormal propagation. The signal of clutter is normally strong, as if there
were intense precipitation, and should be removed before the radar measurement is
applied for rain estimation.
To remove ground clutter, legacy radar systems usually apply various notch filters
such as finite/infinite impulse response (FIR/IIR) filters (Torres and Zrnić 1999;
Golden 2005). The latest clutter filtering techniques are mostly based on spectrum
analysis, for example, the Gaussian model adaptive processing (GMAP) algorithm
introduced by Siggia and Passarelli (2004). Spectrum-based clutter filtering can
reconstruct the weather signal when clutter and weather signals are mixed and
removed together. This kind of filtering ameliorates the deficiency of the conventional notch filters.
In practice, clutter identification is usually needed for efficient clutter filtering.
Compared with weather signals, the phase of clutter signals is normally more stationary. This is a key character of most ground clutter and can be utilized for identification. The typical algorithm is the clutter mitigation decision (CMD) system
introduced by researchers at the National Center for Atmospheric Research (NCAR;
Hubbert et al. 2009). Recently Moisseev and Chandrasekar (2009) have proposed
a new algorithm, applying dual-polarization spectral decomposition to identify the
clutter on this front.
13.3.3.3 Classification
Radar–rain measurements are usually contaminated by nonrain signals from snow,
hail, clutter, insect, bird, bat, and airplane. A common situation can be seen during
a convective–stratiform storm. The melting hail usually exists in the convective core
and results in a large reflectivity, which might be larger than 55 dB. This measurement could cause an unrealistic estimation of extremely intense rain. The similar
situation of melting hail/snow happens within the melting layer of stratiform. Above
the melting layer, the radar measures the graupel, ice crystal, dry snow, or hail.
Those radar measurements do not reflect the rain properties of the storm and would
again lead to an incorrect estimation of the rain.
The most advanced algorithms of hydrometer classification are currently based on
dual-polarization radar measurements. The fuzzy-logic scheme is the basis for most of
the algorithms, such as the radar echo classifier developed by NCAR (Vivekanandan
et al. 1999; Kessinger et al. 2003), the polarimetric hydrometeor classification algorithm developed by NSSL (Straka et al. 2000; Zrnić et al. 2001; Schuur et al. 2003),
Radar Polarimetry for Rain Estimation
Other radar measurements such as radar reflectivity and differential reflectivity are
also smoothed sometimes (e.g., over a 1-km range) before the rain estimation. Lee et
al. (1997) introduced a speckle filter technique, which could be used for radar applications (Cao et al. 2010). In general, the smoothing would lower the range/angular
resolution of radar measurements. However, it is useful to obtain a better rain estimation with a smaller variance.
13.3.3.2 Clutter Filtering
Ground clutter is attributed to the side-lobe effect of the radar antenna. The sidelobe effect is strong for low-elevation scanning. Clutter is usually observed in the
area close to the radar, but it is sometimes measured at a farther range due to the
effect of abnormal propagation. The signal of clutter is normally strong, as if there
were intense precipitation, and should be removed before the radar measurement is
applied for rain estimation.
To remove ground clutter, legacy radar systems usually apply various notch filters
such as finite/infinite impulse response (FIR/IIR) filters (Torres and Zrnić 1999;
Golden 2005). The latest clutter filtering techniques are mostly based on spectrum
analysis, for example, the Gaussian model adaptive processing (GMAP) algorithm
introduced by Siggia and Passarelli (2004). Spectrum-based clutter filtering can
reconstruct the weather signal when clutter and weather signals are mixed and
removed together. This kind of filtering ameliorates the deficiency of the conventional notch filters.
In practice, clutter identification is usually needed for efficient clutter filtering.
Compared with weather signals, the phase of clutter signals is normally more stationary. This is a key character of most ground clutter and can be utilized for identification. The typical algorithm is the clutter mitigation decision (CMD) system
introduced by researchers at the National Center for Atmospheric Research (NCAR;
Hubbert et al. 2009). Recently Moisseev and Chandrasekar (2009) have proposed
a new algorithm, applying dual-polarization spectral decomposition to identify the
clutter on this front.
13.3.3.3 Classification
Radar–rain measurements are usually contaminated by nonrain signals from snow,
hail, clutter, insect, bird, bat, and airplane. A common situation can be seen during
a convective–stratiform storm. The melting hail usually exists in the convective core
and results in a large reflectivity, which might be larger than 55 dB. This measurement could cause an unrealistic estimation of extremely intense rain. The similar
situation of melting hail/snow happens within the melting layer of stratiform. Above
the melting layer, the radar measures the graupel, ice crystal, dry snow, or hail.
Those radar measurements do not reflect the rain properties of the storm and would
again lead to an incorrect estimation of the rain.
The most advanced algorithms of hydrometer classification are currently based on
dual-polarization radar measurements. The fuzzy-logic scheme is the basis for most of
the algorithms, such as the radar echo classifier developed by NCAR (Vivekanandan
et al. 1999; Kessinger et al. 2003), the polarimetric hydrometeor classification algorithm developed by NSSL (Straka et al. 2000; Zrnić et al. 2001; Schuur et al. 2003),