392

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 4, NO. 3, JULY 2007

A Low-Power High-Resolution Broad-Band Radar Using a Pulse Compression Technique for Meteorological Application
Tomoaki Mega, Kazushi Monden, Tomoo Ushio, Kenichi Okamoto, Member, IEEE, Zen-Ichiro Kawasaki, and Takeshi Morimoto
AbstractA new high-resolution Ku-band Doppler radar for meteorological applications has been developed. With the new system design, the radar can accurately measure the radar reectivity factor with 4-m resolution over a range from 40 m to several kilometers for 100-mW power using a pulse compression technique. Details of the system design, signal processing algorithm, and data acquisition procedures are described. To demonstrate the accuracy of the system, the radar reectivity measurements are compared with the JossWaldvogel disdrometer measurements, and fairly good agreement is shown. The ability of the system to capture the backscattered signal and Doppler spectrum from rain volume at low altitude with high resolution is demonstrated for both convective- and stratiform-type rain events. Index TermsBroad-band radar (BBR), rain, weather radar.

I. I NTRODUCTION

AINFALL observations using weather radar have the major advantage that it is possible to observe precipitation widely in a short time. However, the rain rate obtained by weather radar does not necessarily correspond to that observed by ground-based rain gauges. One of the causes of this disagreement is the nonuniformity distribution of raindrops in a rain scattering volume. Another cause is that most types of radar cannot obtain a rain echo at low altitude, because the radar beam overshoots by several kilometers in height due to the Earths curvature or, in vertical pointing mode, the receiver is turned off while the pulse is being transmitted. Therefore, to identify how the reectivity prole changes near the ground and to estimate precisely the causes of the reectivity measurement errors, a radar capable of observing the reectivity from near the ground with high resolution and high precision is needed. In addition, our interest in the ne structure of the bright band, or melting layer, which is important for constructing the algorithm of spaceborne microwave radiometers, motivated us to develop a high-resolution radar.

Manuscript received August 22, 2006; revised January 9, 2007. This work was supported in part by the Strategic Information and Communications R&D Promotion Program (SCOPE) of the Ministry of Internal Affairs and Communications and in part by the fund of Core Research for Evolutional Science and Technology of the Japan Science and Technology Agency. T. Mega and K. Okamoto are with the Department of Aerospace Engineering, Osaka Prefecture University, Sakai 599-8531, Japan. K. Monden is with the Mitsubishi Electric Corporation, Tokyo 100-8310, Japan. T. Ushio, Z. Kawasaki, and T. Morimoto are with the Department of Information and Communications Technology, Osaka University, Suita 565-0871, Japan. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/LGRS.2007.895705

Pulse-based radars have been used extensively in the past for meteorological applications at microwave frequencies. They transmit pulses of short duration and high peak power at a single frequency and have a phase and duration detection circuit. A major drawback of such pulse radars is their coarse resolution, because it is difcult and expensive to generate pulses on the order of nanoseconds. Another drawback is that pulse radars cannot detect the targets at close range, because the receiver is turned off while the pulse is being transmitted. Although frequency-modulated continuous-wave (FM-CW) radars have been used for certain applications, linearity of frequency is critical for the range accuracy of FM-CW radar. On the other hand, in the pulse compression technique, frequency modulation is not limited to linear chirps; any kind of frequency modulation is acceptable for pulse compression radar. Step-frequency radars (vector network analyzer based) have been widely used especially for characterizing the response of the nonmeteorological targets. Although radar has the advantage of extracting both phase and magnitude information very accurately over a broad bandwidth, it takes a long time to record and process the signal, which is a disadvantage for meteorological applications. In contrast to these radar techniques, the broad-band pulse radar using pulse compression has the advantages of broadband, low peak power, fast data acquisition, and phase detection capabilities, but has the disadvantages of range sidelobe, large amount of data, and long processing time. However, with the recent advances of digital technology, it has become easier to adaptively modulate the transmitting signal, making it possible to reduce the range sidelobe signicantly. A pulse compression weather radar was reported by Tanner et al. [1]. The radar used 13.8-GHz frequency with 4-MHz bandwidth. We use a linear chirp for pulse compression; however, this linearity is not essential for pulse compression, and we could also use nonlinear frequency modulation. This is a difference between present radar and FM-CW radar. In this study, we developed a bistatic Ku-band broad-band radar (BBR), which can transmit and receive wide band (80 MHz) signals, to obtain good estimates of the radar reectivity factor at low altitude with high range resolution. The radar uses the pulse compression technique, which transmits a long modulated pulse and then compresses the received signal with a matched lter. Using this technique, we can acquire rainfall proles of high range resolution and high signal-to-noise ratio. The results show that the developed radar can observe the reectivity prole from 40 m to several kilometers by 100-mW transmitting power. In Section II, we describe the BBR. In Section III, the results of rainfall observations are compared with the data measured

1545-598X/$25.00 2007 IEEE

MEGA et al.: BBR USING PULSE COMPRESSION TECHNIQUE FOR METEOROLOGICAL APPLICATION

393

Fig. 2. Flow of pulse compression process.

transform (FFT) in the frequency domain. The output of the pulse compression Ppc (t) is dened as
Fig. 1. Block diagram of the BBR system.
Ppc (t) = FFT1 [Pr (f ) Pref (f )]

(1)

by other equipment. In Section IV, the initial observations for the convective- and stratiform-type rain events are analyzed. II. D ESCRIPTION OF THE BBR A. Conguration In this study, we developed a bistatic high-resolution and low-power radar system using a wide-band signal for the transmitting and pulse compression technique. The pulse compression technique has the advantage that high range resolution proles can be acquired by low transmitted power. However, the range sidelobe of a compressed signal may contaminate the neighboring rain echo. To overcome this disadvantage, we adopted an arbitrary waveform generator (AWG) for the signal source of our radar system and directly acquired the intermediate frequency (IF) signal by a high-speed analogto-digital (A/D) converter. We can thus exibly choose any transmitted waveform that results in low sidelobe by digital pulse compression. In this study, we used the chirp signal with the BlackmanHarris and raised-cosine window function for suppressing the range sidelobe [2]. Fig. 1 shows a block diagram of the BBR system. This system consists of transmitting and receiving Cassegrain-type antennas, transceiver, AWG, A/D converter, and GPS receiver. The diameters of the antennas are 60 cm. The chirp signal of the 20100-MHz frequency range in the pulse width of 128 s is generated by the AWG. The double phase lock oscillators in the transceiver upconvert this output to 15.7115.79 GHz, and this output is amplied by a solid-state amplier to about 100 mW. The echo received from precipitation is down-converted to an IF signal in the transceiver. The signal is sampled by the A/D converter (sampling ratio: 400 M samples/s, bit rate: 12 bits/sample, and available memory: 4 M samples) and is saved to a personal computer. The high-precision 10-MHz reference clock from the geographic positioning system (GPS) receiver synchronizes the internal clock of the AWG and the A/D converter. The GPS is also used for time synchronizing the other equipment such as the disdrometer. B. Signal Processing Procedure Fig. 2 shows the owchart of the pulse compression process in this letter. In the pulse compression processing, we calculate the cross correlation between the signal received from precipitation and the reference signal. The reference signal is presampled by directly connecting the transceiver output of the transmitting end to the input of the receiving end through the attenuator. In the cross correlation, we use a fast Fourier

(f ), which is called the matched lter, is the complex where Pref conjugate of a presampled transmitted signal Pref (f ), Pr (f ) is a received signal, and FFT1 is the processing of the inverse fast Fourier transform (IFFT). The IFFT algorithm is implemented in the pulse compression software. Before the received and reference signals are processed by the FFT, both the real data are converted to complex data consisting of inphase (I) and quadrature (Q) components using the Hilbert transform. In fact, the I component is the real data from the A/D converter and the Q component is the Hilbert-transformed data itself. For estimating the Doppler velocity of a raindrop, the successive data after the pulse compression in the same range bin is processed by the FFT. In this observation, the pulse repetition frequency is about 2.0 kHz and the number of pulses of the FFT is 128. The resolution of the Doppler velocity is 0.15 m/s in this conguration.

C. Range Resolution Typically, a pulse-compressed waveform is written as f (t) = TB sin(Bt) Bt (2)

where T is the pulse width, B is the bandwidth, and the mainlobe width of the waveform is 1/B . Overall compression gain is given by B , and it becomes 40.1 dB when B is 80 MHz and pulse width is 128 s. Therefore, the ideal range resolution r of the radar using pulse compression can be written as r = c 2B (3)

where c is the speed of light, and the range resolution of our radar with bandwidth of 80 MHz is about 1.9 m. However, the range sidelobe is 13 dB down from the peak. Therefore, we use the output waveform of the signal source (AWG), which is the chirp signal multiplied by the BlackmanHarris or raised-cosine window function. As a result, the range sidelobe is suppressed to 125 dB (BlackmanHarris) and 16 dB (raised-cosine), respectively. However, the mainlobe width becomes broad and the range resolution degrades to about 4 m. This result was conrmed by a scattering experiment using the radar for a few metal plates as targets. III. C ALIBRATION AND E XPERIMENT C ONFIGURATION Most of the data discussed in this letter were obtained at the campus of Osaka Prefecture University, Sakai, Japan, with the BBR in vertical pointing mode. Therefore, the radar measured the time series of the radar reectivity prole with high space

394

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 4, NO. 3, JULY 2007

TABLE I OPERATING CHARACTERISTICS OF BBR

Fig. 4. Time series of Zm for every 2-s mean values and Ze for every 1-min mean values. The lower panel is Zm from the BBR, and the upper panel is Ze from the disdrometer. Altitude of Zm was 40 m.

Fig. 3. Scatter plot of Zm from the BBR and Ze from the disdrometer. The dashed line represents a 1 : 1 slope.

Fig. 5. Timeheight cross section of radar reectivity factor (in decibels of Z) observed by BBR on July 1, 2005.

and time resolution. Table I gives the operation characteristics of our radar. Two-second resolution data are obtained from the incoherent integration of 64 successive pulses and 1-s resolution data are obtained from the incoherent integration of 128 successive pulses. An initial observation was conducted on July 1, 2005. A microrain radar (MRR), which is a vertical pointing FM-CW radar at 24 GHz in 50 mW, was also installed 5 m away from the BBR as a reference. The MRR was operated at the resolution of 35 m in height and 60 s in time. We also used a Joss-Waldvogel disdrometer (JWD) instrument (Distromet Inc. RD-80) for the calibration and comparison of the reectivity above ground. The disdrometer estimates raindrop size distribution (DSD) and rain rate from the vertical momentum of an impacting raindrop. The results of the rainfall observation and the comparison with the equivalent radar reectivity factor Ze are provided by the disdrometer. Our radar was calibrated by using metal plates having known radar cross-sectional values. The results showed that the measured and calculated radar cross-sectional values agreed within 1 dB. This shows that our radar is well calibrated and stable for observing xed targets. However, Ulbrich and Lee [3] and Ulbrich and Miller [4] showed, by comparing the measured Z -value with a calculated Z -value by using the disdrometer, that even a well-calibrated radar like WSR-88D has a systematic offset error in rain observations. Gage et al. [5] showed that the serial reectivity measurements from a disdrometer can be a good calibration for a wind proler. Thus, the calibration of the precipitation radar by the disdrometer is useful to remove the systematic offset error. In this study, we used a similar method to calibrate our radar system. Fig. 3

shows scatter plots of the radar reectivity factor of the BBR, Zm, and Ze calculated from the disdrometer. In the calculation of Ze, the Mie scattering of spherical particles is assumed. The radar reectivity factor Ze is calculated from the DSD accumulated over each minute by the JWD. The radar data used here for comparison are the reectivities observed at the lowest usable height of about 40 m above ground. The attenuation coefcient (in decibels/kilometer) is about 0.042R1.13 for 16 GHz [6]. As the range between radar and rain scattering volume is 40 m, the effect of attenuation is negligibly small. It can be seen that they agree well from 15 up to 65 dBZ. The minimum detectable reectivity is about 15 dBZ. The correlation coefcient is 0.98. The root-mean-square difference between the BBR and the disdrometer observation is about 1.62 dBZ. The causes of the difference of 1.62 dBZ will be the differences of the location of the observed raindrops and of the rain scattering volume and so on. We will continue efforts to minimize the calibration difference in the future. The time series of Zm for every 2-s values and Ze for every 1-min values are shown in Fig. 4. This gure shows that the BBR and the disdrometer have almost the same reectivities. IV. A NALYSIS OF R EFLECTIVITY P ROFILE An example of the timeheight cross section of the reectivities observed simultaneously by the BBR and the MRR are shown in Figs. 5 and 6, respectively. The rain rate calculated from the disdrometer is also shown in the same temporal axis in Fig. 7. This event includes convective-type rainfall exceeding more than 80 mm/h at 18:38, and no bright band is observed. Although there are several differences between the MRR and the BBR in terms of frequency, power, modulation, and data

MEGA et al.: BBR USING PULSE COMPRESSION TECHNIQUE FOR METEOROLOGICAL APPLICATION

395

Fig. 6. Timeheight cross section of radar reectivity factor (in decibels of Z) observed by MRR on July 1, 2005.

Fig. 9. Vertical prole of radar reectivity factor and mean Doppler velocity with bright band at 9:43 in Fig. 8. Fig. 7. Time series of rain rate observed by the disdrometer at the ground on July 1, 2005.

Fig. 8. Timeheight cross section of radar reectivity factor observed by BBR on January 14, 2006.

processing, the overall patterns of reectivity proles between the BBR and the MRR look similar except from 18:38 to 18:40 above 600 m, during which a strong rain attenuation effect is observed at the MRR. The attenuation coefcients of the MRR and the BBR are about 12 dB/km and about 4.8 dB/km for a rain rate of 80 mm/h, respectively. For example, reectivities of the MRR and the BBR at 18:36 and 595 m are 37.4 and 45.9 dBZ, respectively. The attenuation values estimated by the rain rate at the surface are about 14.2 and 5.7 dB, respectively, and thus the corrected reectivities of the MRR and the BBR are about the same, i.e., 52 dB. It is clearly shown that the observations by the BBR reveal the rain structure in much more detail than the MRR, and this small-scale structure tends to be obscured in the MRR observations, indicating that the BBR observations are well suited for ne-scale measurements of rainfall at low altitude. An increase in the rainfall is similarly seen from 18:40 to 18:45 in the lower layer from the BBR and the MRR observations. Similarly, this may be largely attributable to the rainfall attenuation. However, it is difcult to correct the attenuation by rainfall in general, and if a part of it is not due to rainfall attenuation, it would be found that it strengthens toward the ground, as discussed in [7]. Another example of the BBR measurement is shown in Fig. 8 for the stratiform-type rain on January 14, 2006. In this 75-min observation, many tilted rain echoes are observed. Although the horizontal distribution of precipitation cannot be seen in this vertical pointing observation, it is possible to make some

Fig. 10. Vertical prole of radar reectivity factor and mean Doppler velocity with bright band at 9:44 in Fig. 8.

speculations. If we assume no vertical wind velocity and no horizontal advection of precipitation systems, the tilted rain echo would be due to the descent of raindrops toward the ground. From the slope estimation of the tilted echoes, the terminal velocity of rain echo can be estimated. For example, the relatively strong tilted echo at 9:44 has fallen from 2000 m in height to the ground over about 130 s, giving a fall velocity of about 15 m/s. This terminal velocity can also be estimated by the Doppler spectrum measurements. Figs. 10 and 12 show the Doppler spectrum and the reectivity factor at 9:44 calculated from the phase change of the 128 successive pulses. Below the melting layer at 2000 m, the maximum Doppler velocity is distributed from 6 to 8 m/s, which is not consistent with the slope estimation of 15 m/s. The classical estimation assumes no horizontal advection and no vertical wind velocity, and the rain event analyzed is a stratiform rain event which does not have a strong updraft or downdraft wind. It is possible that the apparent terminal velocity of 15 m/s from the classical estimation would be due to the advection of tilted rain streaks in the stratiform rain system. As is clearly seen in Fig. 8 on January 14, 2006, the bright band appeared during the observation. High-resolution radar measurement is very suitable for researching the bright band.

396

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 4, NO. 3, JULY 2007

Fig. 11. Vertical prole of Doppler spectrum of bright band of Fig. 9.

Fig. 12. Vertical prole of Doppler spectrum of bright band of Fig. 10.

One of the Doppler spectra of echoes between 0 and 2000 m altitude measured by the BBR at 9:43 on January 14, 2006 is shown in Fig. 9. This shows the most common prole of the bright band in this event in terms of reectivities and the Doppler spectrum. As is well known, the reectivity prole clearly shows the V-shaped signature that is associated with the melting process. Just above the bright band, the reectivity increases rapidly with decreasing height, and in the lower part of the melting layer, the reectivity drops off sharply where the rain particles collapse to form small raindrops. However, some observations do not show these common features. In Figs. 10 and 12, there is an interesting feature between 1200 and 2200 m in height. Fabry and Zawadzki [8] showed the similar double or multiple bright bands during the transition between winter and summer from long-term observations in Montreal, Canada, using an X-band and Wind proler radar system and speculated that snowakes melt in the rst bright band, refreeze between the two bright bands, and melt again in the second bright band. In our observation, the rst bright band appeared at about 2000 m in height and the Doppler velocity peaked at the almost the same altitude. Below the bright band, the reectivity decreases several decibels of Z for several hundred meters in height and then a second bright band appears again at 1400 m altitude. During these regimes, there is no change in the Doppler velocity with a slight increase toward the second bright band. The physical process involved here could be related to refreezing with another measurement such as temperature measurement and so on. However, the BBR is clearly effective for research of the bright band. V. C ONCLUSION In this study, we developed a BBR, and an initial experiment was conducted in Osaka, Japan, to study the small-scale variability of rainfall at low altitude. The BBR uses a low-

power (100 mW) 80-MHz broad-band signal for transmitting with arbitrary modulation and digital pulse compression technique for detecting targets. The radar system can observe the backscattered signal from rain volume with high resolution from near the ground. The range resolution was on the order of 4 m with 1-s temporal resolution. Calibration was successfully done within 1.62 dBZ using the DSD measured by the JWD, and the correlation coefcient between the radar reectivity factor observed at 40 m above ground and calculated from the DSD was 0.98. Two case studies were presented here, which demonstrate the BBR performances. In the rst case, a ne-scale measurement by the BBR was shown in contrast with the conventional radar system for a convective-type event and the attenuation effect was discussed in the comparison of the MRR and the BBR. In the second case, a stratiform-type rain event was analyzed. The rain streaks were clearly observed by the BBR system, and a classical estimation of rainfall velocity using the slope of the rain streaks showed a disagreement with a Doppler spectrum estimation, which suggests the existence of advection. A bright band of the stratiform-type rain was observed. A vertical radar reectivity prole and a Doppler prole of rain with a bright band were observed. A double-peak bright band feature was also observed and analyzed. In these case studies, the radar was operated in the vertically pointing mode. The system can rotate and make 3-D observations. This radar will be useful for studying the inhomogeneity observed in rain. It would be worth adding polarization capabilities to this radar for measuring the distribution of raindrop size. This function can be installed now, and the results of using this function will be examined in another paper. ACKNOWLEDGMENT The authors would like to thank H. Hanado, K. Nakagawa, and T. Iguchi for lending the equipment necessary to pursue this research. R EFERENCES
[1] A. Tanner, S. L. Durden, R. Denning, E. Im, F. K. Li, W. Ricketts, and W. Wilson, Pulse compression with very low sidelobes in an airborne rain mapping radar, IEEE Trans. Geosci. Remote Sens., vol. 32, no. 1, pp. 211213, Jan. 1994. [2] A. Nashashibi, K. Sarabandi, R. De Roo, P. Frantzis, and F. Ulaby, An ultrafast wide-band millimeter-wave (MMW) polarimetric radar for remote sensing applications, IEEE Trans. Geosci. Remote Sens., vol. 40, no. 8, pp. 17771786, Aug. 2002. [3] C. W. Ulbrich and L. G. Lee, Rainfall measurement error by WSR-88D radars due to variations in Z-R law parameters and the radar constant, J. Atmos. Ocean. Technol., vol. 16, no. 8, pp. 10171024, 1999. [4] C. W. Ulbrich and N. E. Miller, Experimental test of the effects of Z-r law variations on comparison of WSR-88D rainfall amounts with surface rain gauge and disdrometer data, Weather Forecast., vol. 16, no. 3, pp. 369374, 2001. [5] K. S. Gage, W. L. Clark, and C. R. Williams, Determining reectivity measurement error from serial measurements using paired disdrometers and prolers, Geophys. Res. Lett., vol. 31, no. 23, L23107, 2004. [6] R. Meneghini and T. Kozu, Spaceborne Weather Radar. Norwood, MA: Artech House, 1990. [7] R. D. Palmer, B. L. Cheong, M. W. Hoffman, S. J. Frasier, and F. J. L. Dekker, Observations of the small-scale variability of precipitation using an imaging radar, J. Atmos. Ocean. Technol., vol. 22, no. 8, pp. 11221137, Aug. 2005. [8] F. Fabry and I. Zawadzki, Long-term radar observations of the melting layer of precipitation and their interpretation, J. Atmos. Sci., vol. 52, no. 7, pp. 838851, 1995.