Transformation of the csu chill radar facility to

Pdf File 9,417.04 KByte, 22 Pages


by Francesc Junyent, V. Chandrasekar, V. N. Bringi, S. A. Rutledge, P. C. Kennedy, D. Brunkow, J. George, and R. Bowie

The CSU?CHILL radar can now make simultaneous polarimetric measurements at S and X bands to advance the understanding of precipitation processes.

T he Colorado State University?University of Chicago?Illinois State Water Survey (CSU?CHILL) National Radar Facility (Brunkow et al. 2000) located in Greeley, Colorado, is a research facility operated by CSU, under the sponsorship of the National Science Foundation (NSF) and CSU. The CSU?CHILL radar has recently gone through a major transformation to add support for simultaneous dual-wavelength (S and X bands), dual-polarization

AFFILIATIONS: Junyent, Chandrasekar, and Bringi--Department of Electrical and Computer Engineering, Colorado State University, Fort Collins, Colorado; Rutledge, Kennedy, Brunkow, George, and Bowie--Department of Atmospheric Science, Colorado State University, Fort Collins, Colorado CORRESPONDING AUTHOR: Francesc Junyent, Electrical and Computer Engineering, Colorado State University, 209 Campus Delivery, Fort Collins, CO 80525 E-mail:

The abstract for this article can be found in this issue, following the table of contents. DOI:10.1175/BAMS-D-13-00150.1

In final form 27 August 2014 ?2015 American Meteorological Society

(H and V) radar operation, as well as high polarization purity S- or X-band stand-alone operations (with S band being the portion of the electromagnetic spectrumcomprised inside the 2?4-GHz interval, X band being the portion inside the 8?12-GHz interval, and H referring to a linearly polarized electromagnetic wave with its electric field confined on the horizontal plane, and V referring to a linearly polarized electromagnetic wave with its electric field confined on the vertical plane). This transformation process started with the installation of a low-sidelobe dualoffset Gregorian antenna capable of supporting three different feeds (S band, X band, and simultaneous S and X band all with dual-polarization capability) and culminated with the development and installation of a separate X-band channel dual-polarization radar system. This work had multiple scientific and engineering motivations, ranging from the development of specific dual-wavelength techniques and X-band algorithms that could be validated with collocated S-band data to the observation of a variety of weather phenomena at high spatial resolution and with high polarization purity.

In general, multiple-wavelength radar systems rely on the fact that a given hydrometeor will have


JUNE 2015

| 975

a different scattering behavior depending on the wavelength at which it is illuminated, as the relative size of the scatterer with respect to the illuminating wavelength dictates the scattering regime. A small scatterer (i.e., one that is small compared to the illuminating wavelength) will operate in the Rayleigh regime, and one that has a size comparable to the wavelength (or larger) will enter the Mie regime. In addition, shorter wavelengths experience increasing attenuation when propagating through rain. All of these mechanisms are leveraged to infer something about the medium through which the radar waves are propagating.

Earlier dual-wavelength systems were motivated by hail detection and liquid water content estimation. The initial dual-frequency application appears to be hail detection based on multiple wavelength observations of a hailstorm in England (Atlas and Ludlam 1961). The underlying idea was that the scattering signal of the shorter wavelength radar would enter the Mie region before the signal of the longer wavelength radar would, when suitably large hailstones were being observed. In rain, both wavelengths would be in the Rayleigh regime (except for the larger raindrops over 3 mm in diameter), and with proper system calibration the two reflectivities should match (except for the impact of attenuation at the shorter wavelength). Estimation and removal of attenuation at the shorter wavelength (X band in this era) proved to be a difficult problem. This caused Eccles and Atlas (1973) to propose using the range derivative of the ZS/ZX power ratio (where DWR = ZS/ZX is defined as the dual-wavelength ratio, with ZS being the radar reflectivity at S band and ZX being the radar reflectivity at X band) to identify hail boundaries. It was thought that the location of a sharp negative d(DWR)/dr range derivative would accurately mark the far edge of the hail shaft. Additional experience with the derivative method resulted in rather confusing results. Jameson and Srivastava (1978) eventually showed that range variations in hailstone diameter and/or water coating could also cause significant fluctuations in the d(DWR)/dr derivative. From this point on, the derivative method was abandoned and the basic attenuation-corrected DWR was used for hail detection. One of the first major applications of this dualfrequency hail signal was in an analysis of hail growth in a National Hail Research Experiment (NHRE) storm using CP2 data (Jameson and Heymsfield 1980) (CP2 was a dual-wavelength S- and X-band research radar built and operated by the National Center for Atmospheric Research that initially used separate pedestals and antennas for each frequency). Rinehart

and Tuttle (1982) argued that pattern-matching artifacts had the potential to bias the hail signal. CP2 converted to the X-band antenna(s) being mounted on the S-band pedestal in the 1980s, and in this configuration, attenuation-corrected DWR power ratio data were included in several research results: correlations of Zdr [differential reflectivity, obtained as the difference in reflectivity (dBZ) at horizontal and vertical polarization] and dual-frequency indication of hail (Bringi et al. 1986b), hydrometeor identification in an evolving microburst [Microburst and Severe Thunderstorm (MIST) Project; Tuttle et al. 1989], and hail patterns in a range?height indicator (RHI) scan (Herzegh and Jameson 1992).

While dual-wavelength hail identification methods depended upon Mie scattering, efforts to estimate liquid water content (LWC) from differential attenuation at S and X bands worked best in all liquid conditions when Rayleigh scattering was present at both frequencies (Eccles and Mueller 1971). Eccles (1979) also proposed a method of using dualfrequency data to estimate rain rates. The effects of Mie scattering are significant enough that Gaussiat et al. (2003) used a three-frequency approach (S, Ka, and W bands) to improve LWC estimation in RHIs through rain showers. Frequency pairs shorter than the S?X combination used by CSU-CHILL have also been used in recent years for dual-frequency applications. As some examples, Vivekanandan et al. (2001) used X and Ka bands for the detection of supercooled water to identify aircraft icing hazards, and Chandrasekar et al. (2010) used Ka and Ku bands to document dual-frequency observations for Global Precipitation Measurement (GPM) ground validation.

The CSU?CHILL radar is designed to address the technical challenges posed by simultaneous dualwavelength measurements through the use of a singleantenna feed illuminating the same reflector at both wavelengths. To maintain appropriate performance at both wavelengths, the radar undergoes periodic characterization of end-to-end system gain using calibration sphere flights and through the continuous monitoring of transmitter and receiver performance.

MOTIVATION FOR TH E C SU ? C HILL X-BAND COMPONENT. The key motivating factor in the addition of the X-band component was the enhanced resolution achievable by the CSU? CHILL offset feed antenna at X band through its capability to accommodate different feedhorn units (S band, X band, and simultaneous S and X bands all with dual-polarization capability) to create singleand dual-wavelength, dual-polarization datasets. This

976 |

JUNE 2015

is a key feature of the CSU?CHILL radar and a dif- challenges disappear and the difference between reflec-

ferentiating factor from other weather radar systems. tivities at the different wavelengths can be attributed

However, this is also a departure from traditional to the wavelength sensitivity to different scattering

dual-wavelength radar systems with matched beams mechanisms. The dual-wavelength data collected so

and the associated techniques developed for those. far have many examples of differing S- and X-band

The main issue becomes the possibility of nonuni- reflectivities in gust fronts, boundary rolls, etc.

form beamfilling effects between the 1? beamwidth Finally, another significant motivating factor was

at S band and the 0.3? beamwidth at X band. In gen- the potential X-band stand-alone capabilities. The

eral though, it should be noted that it is possible to higher spatial resolution allows studies of small-scale

synthesize a larger 1? beam from the X-band higher- features, such as individual cumulonimbus clouds,

resolution data that matches the corresponding gust front circulations, and microbursts, while the

S-band beam (as demonstrated later in Fig. 12) or that greater differential phase sensitivity should enhance

one can analyze the statistical properties of the multiple the analysis of winter storm patterns, such as crystal

X-band beams that fall inside the corresponding growth; summer storm patterns, such as the identi-

S-band beam (as shown in Fig. 10) as ways to deter- fication of electrification morphology (Caylor and

mine whether such nonuniform beamfilling effects Chandrasekar 1996); and general rainfall estimation

might be occurring.

applications, especially for lighter rain rates.

Another important motivating factor is the

investigation of attenuation correction schemes. With D E S C R I P TI O N O F TH E C S U ? C H I L L

X-band radar systems becoming an increasingly popular RADAR SYSTEM. The CSU?CHILL radar is

technology, there is a continued need to address the housed inside an inflatable radome and uses a 9-m

particular issues that arise in higher-frequency systems parabolic dual-offset reflector antenna (Bringi et al.

such as range?Doppler ambiguity coupling and signal 2011) mounted on an elevation over an azimuthal

attenuation. The availability of a dual-polarization positioner system as shown in Fig. 1. The antenna is

dataset at both S and X bands from coaxial beams is a illuminated using one of three interchangeable feeds,

new feature available to the research community that allowing the system to operate at S and X bands and at

should allow improvements to be made in the post- simultaneous S and X bands as required. This level of

processed X-band data fields and verified against their flexibility allows better tailoring of the radar perfor-

nonattenuated S-band counterparts (with CP2 being mance to the specific application and data purpose;

able to measure the linear depolarization ratio but no the characteristics of the combined dual-frequency,

copolar variables at X band; Atlas 1990).

dual-polarization feed, while good in its own metrics,

Another potential area of application of the dual- are not as refined as that of the single-frequency units.

wavelength, dual-polarization CSU?CHILL dataset is The resulting antenna beams are conceptually illus-

in developing the capability to independently identify trated in Fig. 2, showing the difference in beam sizes

Mie scattering effects in the X-band data. This may be due to the use of the same reflector and subreflector

accomplished by examining the X-band polarimetric apertures at the different wavelengths. Also, it is

variables and verifying against the historical com- worth noting the preferred location of sidelobes in

parison of the X- and S-band reflectivities. Reliable the higher elevation portion of the antenna pattern to

identification of Mie scat-

tering at X band versus

S band could be used to

estimate hydrometeor size.

The X- and S-band re-

flectivity difference could

also be used to identify

Bragg scattering (scattering

caused by sharp gradients

in the refractive index of

the atmosphere): in clear

air and/or first echo appli-

cations such as those dis-

cussed by Knight and Miller

(1993), the signal extinction Fig. 1. CSU?CHILL National Weather Radar Facility located near Greeley.


JUNE 2015

| 977

archive and display servers,

which are common to both

the S-band and X-band

systems. This allows

leveraging the same tools

for display and analysis of

both X- and S-band data

streams. Internet access

to the signal processor

computers allows remote

control and operation of

the radar through graphi-

cal user interfaces and/or

command line programs,

enabling overnight and

unattended system opera-

tion. Figure 4 illustrates the

architecture of the CSU?

Fig. 2. Illustration of CSU?CHILL radar coaxial antenna beams. The S-band antenna beam has a gain of 43 dB and a beamwidth of 1.0?, whereas the X-band antenna beam has a gain of 53 dB and a beamwidth of 0.3?.

CHILL dual-wavelength radar.

Both the S-band and

X-band portions of the

minimize returns from ground clutter. This is a result radar hardware contain provisions to maintain

of the dual-offset Gregorian reflector geometry of the accurate system calibration. The S-band receiver is

CSU?CHILL radar antenna. Traditional center-fed calibrated using an automated signal source injected

reflector antennas [such as those used in the National at the calibration plane, whereas the S-band trans-

Weather Service's network of operational Weather mitted power is measured at the same calibration

Surveillance Radar-1988 Doppler (WSR-88D) S-band plane using a power meter. The power meter is also

radars, also known as the Next Generation Weather used to measure the signal source output power

Radar (NEXRAD)] have no preferred sidelobe loca- and to remove any bias in the transmitter?receiver

tion and therefore cannot mitigate returns of ground calibration. The X-band receiver is calibrated using

clutter in that manner.

an onboard signal source, and a sample of the trans-

The X-band portion of the radar hardware is mitted pulse is passed through the calibrated receiver

mounted directly on the antenna structure (see to track the transmitted power on a pulse-to-pulse

Fig. 3) to minimize waveguide lengths and to avoid

the use of a waveguide rotary joint. The transmitter,

duplexer, and receiver subsystems share a single

enclosure (transceiver enclosure). A second enclosure

houses the data acquisition and timing generation

subsystems. The radar control and data streams share

a single Ethernet interface that is brought to the signal

processor system through a gigabit Ethernet-capable

slip-ring assembly. The corresponding S-band

portion of the radar hardware is located in a trailer

adjacent to the radome.

The user trailer houses the X-band and S-band

signal processor computers. The X-band signal

processor gathers the digitized complex voltage radar

data stream from the data acquisition system and

the position data stream from the motion control system and computes the Doppler spectrum moments according to the system configuration. The real-time

Fig. 3. CSU?CHILL radar X-band channel components: (a) the transceiver enclosure, (b) the data acquisition enclosure, and (c) the dual-frequency,

output of the signal processor is then passed to the data dual-polarization feed.

978 |

JUNE 2015

basis. Additionally, sphere calibrations using foil- compared to S band (mostly in the 0.99?1.00 range)

covered spheres suspended from free-flying balloons when operating in dual-frequency mode, probably

are performed routinely to check the overall system due to the lower quality of beam matching at the two


polarizations at X band. This is believed to be a side

The main characteristics of the CSU?CHILL radar effect of the dual-polarization, dual-frequency horn

both for the X-band component and the S-band com- illuminating the antenna system, as other X-band da-

ponent are listed in Table 1.

tasets collected with the single-frequency X-band horn

(as the one shown in Fig. 16) do behave as expected. The

DATA EXAMPLES. This section presents a few dp images show the expected increase after the high Z data cases illustrating the CSU?CHILL radar current areas with overall higher-phase excursions at X band

capabilities. These selected data cases include convec- (Fig. 7). The higher sensitivity of dp at X band makes tive summer storms, winter storms, and nonprecipi- some features, such as the potential vertical alignment

tiating echoes. A larger selection of data cases can of particles due to electrification (roughly at 30-km

be found on the CSU?CHILL webpage (chill range and 10-km height), easier to observe. One can under the "Articles" tab.

also see how in the area around 20 km in range where

the Z maximum is located, dp at X band goes through a Dual-frequency data for convective storms. A look at local maximum possibly due to backscatter differential

attenuation. During June 2013 the CSU?CHILL phase owing to scattering off melting ice particles and/

radar operated in simultaneous X-band and S-band or large raindrops (Tr?mel et al. 2013).

dual-frequency mode, collecting data on a number of The availability of simultaneous reflectivity mea-

convective storms during the CHILL Microphysical surements at X and S bands offers the possibility of

Investigation of Electrification (CHILL-MIE) project. using the unattenuated S-band measurement as a

Figures 5?7 show simultaneous data at X and S band constraint for the attenuation correction algorithm

for reflectivity (Z), copolar cross-correlation coefficient applied to the X-band data. To do so successfully, one

(hv), and differential propagation phase (dp) obtained must be certain that no residual bias exists between in an RHI scan through a convective storm on 29 June. the measured X- and S-band reflectivities (i.e., the

Looking at the Z images, one can see that qualitatively measured data are properly calibrated). To assess that

there is very good agreement on the storm features potential bias as shown in Fig. 8, the data from the top

and reflectivity levels at the higher elevation portion portion of the previously shown Z scan are selected

of the scan (above 10? elevation angle; Fig. 5). Below (over 10? in elevation in Fig. 5), and a two-dimensional

that, the X- and S-band measurements differ due to histogram of Z values at X and S band between 10 and

the presence of signal at-

tenuation at X band beyond

the higher reflectivity area

in the 20-km range. The

hv images [not corrected for the signal-to-noise ratio

(SNR)] reveal that the area

of higher Z is collocated with

a local dip in the hv values (around 20 km in range,

extending from 0 to 2 km in

height) possibly due to some

combination of depolarizing

and Mie scattering effects of

what could be melting grau-

pel or hail mixed with rain.

(Fig. 6; Bringi et al. 1986a).

Another feature that is read-

ily apparent is the generally

slightly reduced correlation

values at X band (mostly in

the 0.98?0.99 range) when Fig. 4. CSU?CHILL dual-frequency radar architecture.


JUNE 2015

| 979

Table 1. CSU?CHILL dual-frequency radar system main specifications.


Antenna Reflector type Feed type Polarization Gain (dBi) Beamwidth (?) Sidelobe level (dB) Cross-polarization level (dB) Scan type Scan rate

Transmitters Frequency Type Power max Transmit modes Duty cycle (%) PRF max (kHz)

Receivers Sensitivity (dual-wavelength mode) Noise figure (dB) Dynamic range (dB) Digitizer bits Range sampling (m)

Signal processing and products Processing modes Polarization processing Data products

S band

X band

8.5-m dual-offset Gregorian parabolic Scalar, symmetric orthomodal transducer (OMT) Linear H and V 43 1.0 <-33 <-35 PPI (360? sector), RHI, fixed pointing, vertically pointing <18? s-1

53 0.3 <-36 <-35

2.725 GHz Dual Klystron 1 MW Single polarization, simultaneous, alternating 0.16 1.25

9.41 GHz ? 30 MHz Magnetron 25 kW Simultaneous 0.16 2.00

-15 dBZ, 10 km 3.4 80 14 30 ?150

-15 dBZ, 10 km 4.0 90 14 1.5?192

Pulse pair, spectral clutter filter, second-trip suppression, dual-Doppler velocity unfolding Hydrometeor ID, attenuation correction, KDP estimation Z, ZDR, V, W, HV, NCP, DP, KDP, SNR, (LDR, CX for S-band only)

35 dBZ is created. The data points in the histogram are probably mostly ice particles, although supercooled liquid water cannot be ruled out. All histogram data points below a threshold (at least four occurrences for this particular dataset) are eliminated, so that rare occurrences (points affected by attenuation, dissimilar scattering mechanisms, etc.) do not bias the sample. After that, the mean value of each one-dimensional histogram of X-band reflectivity values per S-band reflectivity bin is obtained (shown as plus sign points in Fig. 8), and a regression line is fit through the resulting mean values. The obtained regression line shows that the two Z datasets are well correlated (with slope m = 1.0006) and that there is negligible bias (with offset n = -0.0040). Additionally, the standard deviation is calculated and found to be slightly over 2 dB.

Once established that the X and S band datasets are well correlated and unbiased, attenuation correction can be attempted. As previously mentioned, the S-band measurement can be used as a constraint for the attenuation correction algorithm applied to the X-band data. This is expressed in mathematical form in Eq. (1) below, where ZHS (r) is horizontal polarization reflectivity at S band, ZHX(r) is horizontal polarization reflectivity at X band, A(r, ) is specific attenuation as defined in Eq. (7.143) in Bringi and Chandrasekar (2001), r is range, and is the total attenuation (dB) at the maximum range considered:

min f () = ZHS (r)-[ZHX (r)+ A(r,)] .(1)

980 |

JUNE 2015

Download Pdf File