High-resolution atmospheric profiling using combined spaced antenna and range imaging techniques

Transcription

1 RADIO SCIENCE, VOL. 39,, doi: /2003rs002907, 2004 High-resolution atmospheric profiling using combined spaced antenna and range imaging techniques T.-Y. Yu School of Electrical and Computer Engineering, University of Oklahoma, Norman, Oklahoma, USA W. O. J. Brown Atmospheric Technology Division, National Center for Atmospheric Research, Boulder, Colorado, USA Received 1 May 2003; revised 7 November 2003; accepted 19 December 2003; published 12 February [1] A novel approach to high-resolution wind profiling is presented using spaced antenna (SA) and range imaging (RIM) simultaneously. RIM was developed recently to improve the range resolution of pulsed radar by transmitting multiple frequencies. In this work, SA techniques are used to analyze RIM-synthesized signals and obtain high-resolution profiles of three-dimensional wind field. This high-resolution technique is termed RIM- SA. Simulation results demonstrate the feasibility of RIM-SA to measure wind shears embedded within the radar volume. The applications of RIM-SA in the atmospheric boundary layer and lower free troposphere are demonstrated using the multiple antenna profiler radar (MAPR) of National Center for Atmospheric Research (NCAR). Highresolution profiles using RIM-SA are shown to be consistent with profiles measured by a radiosonde and MAPR using a single frequency and short pulse. INDEX TERMS: 3360 Meteorology and Atmospheric Dynamics: Remote sensing; 3394 Meteorology and Atmospheric Dynamics: Instruments and techniques; 6924 Radio Science: Interferometry; 6974 Radio Science: Signal processing; KEYWORDS: wind profiling, spaced antenna, atmospheric radar Citation: Yu, T.-Y., and W. O. J. Brown (2004), High-resolution atmospheric profiling using combined spaced antenna and range imaging techniques, Radio Sci., 39,, doi: /2003rs Introduction Copyright 2004 by the American Geophysical Union /04/2003RS [2] High resolution in situ measurements have shown the atmosphere to be replete with distinct temperature, humidity, and wind shear layers at meter and submeter scales [e.g., Dalaudier et al., 1994; Muschinski and Wode, 1998]. Observations of these fine structures are important for the understanding of small-scale dynamics such as turbulent processes and atmospheric instabilities. However, in situ observations can only provide limited coverage in time and space. In contrast, remotesensing instruments can provide continuous vertical profiles of the atmosphere; however, the vertical resolution is often limited by the system bandwidth. For example, UHF/VHF wind profilers provide reliable measurements of backscattered power, vertical and horizontal wind from the boundary layer to the stratosphere. The vertical resolution of these profilers ranges typically from a hundred meters to several hundreds of meters. Although resolution can be increased by transmitting shorter pulses, it requires not only larger system bandwidth, but also more transmitted power in order to keep the sensitivity constant. In this work, a novel technique that uses long pulses and provides fine resolution of three-dimensional wind measurement is proposed and discussed. [3] Range imaging (RIM) has been recently developed to reconstruct the distribution of echo power within the radar volume in range by transmitting a few slightly shifted frequencies [Palmer et al., 1999; Luce et al., 2001]. RIM can be thought of as a generalization of frequency domain interferometry (FDI) developed by Kudeki and Stitt [1987] and Franke [1990]. Unlike FDI, RIM does not require prior knowledge of layer number and layer shape. As a result, fine details of atmospheric structure can be revealed within the radar volume in range. RIM can be postulated as an inverse problem given signals from multiple transmitting frequencies [Yu 1of13

2 and Palmer, 2001]. Several algorithms, such as Fourier RIM, Capon RIM [e.g, Palmer et al., 1999; Chilson et al., 2003], MUSIC RIM [Luce et al., 2001], and Langunas- Gasull RIM [SmaÏni et al., 2002], demonstrated the utility of RIM for making high resolution observations of the backscatter profile of atmospheric structures. However, no Doppler information of these structures was shown, except as discussed briefly by Palmer et al. [1999] and Yu [2000]. Therefore, one of the goals for this work is to demonstrate applicability of RIM to obtain high-resolution profile of radial velocity using simulation and experimental data. [4] Furthermore, measuring horizontal wind at the same resolution as the resolution of echo power and radial velocity in RIM is of primary interest. On a conventional single-frequency radar, horizontal wind can be measured by either the Doppler beam swinging (DBS) or the spaced antenna (SA) techniques [e.g., Briggs, 1980; Larsen and Röttger, 1989]. A minimum of three beam directions including one at zenith and two off-vertical directions can be used to estimate the wind field for DBS. For SA, transmitting and receiving beams both point vertically. Backscattered signals are received at spatially spaced receivers and are used to estimate horizontal winds. For both DBS and SA, the vertical resolution of wind measurement is still determined by pulse width. In this work, a new technique, based on a combined use of RIM and SA (RIM-SA), is developed to improve the range resolution of horizontal wind meaurements. High-resolution profiles of echo power and vertical velocity are obtained simultaneously for each receiver using the RIM approach. As a result, RIM-SA has the potential to resolve not only fine reflectivity structures, but also fine-scale wind shears. [5] Experiments using combined multiple-receiver and dual-frequency techniques have been conducted by several authors [Stitt and Kudeki, 1991; Cohn and Chilson, 1995; Brown and Fraser, 1996]. In these experiments, signals from multi-receiver and dual-frequency were analyzed independently using SA and FDI, respectively. In other words, the resolution gained by FDI was not incorporated into SA analysis. In RIM-SA, SA is implemented on RIM improved resolution synthesized signals. Many SA algorithms have been proposed to measure horizontal winds. In this work, full correlation analysis (FCA) [Briggs, 1984], which has been widely used in many profilers, is employed and is termed RIM-FCA when signals from multiple frequencies are used. A different approach to high-resolution wind measurements was proposed by Palmer et al. [1995], where signals from dual-frequency were analyzed using imaging Doppler interferometry (IDI) technique [Adams et al., 1986]. Another application of combined multiple-receiver and multiple-frequency observations was the three-dimensional imaging of Yu and Palmer [2001]. [6] A description of RIM-SA and its limitation are presented in Section 2. High-resolution measurements of echo power, vertical velocity, and horizontal velocity are demonstrated using numerical simulations in Section 3. In Section 4, RIM-FCA is tested and verified using a 915 MHz boundary layer radar, the Multiple Antenna Profiler Radar (MAPR), of the National Center for Atmospheric Research (NCAR), Atmospheric Technology Division (ATD). 2. High-Resolution Profiling Using RIM and SA [7] In order to implement RIM-SA technique, a minimum of three spatially separated receivers and two shifted transmitting frequencies are needed. The radar beam points vertically for both transmitting and receiving in a RIM-SA configuration and typically, multiple frequencies are transmitted on a pulse-by-pulse basis. To simplify the problem, it is assumed that signals from multiple receivers and multiple frequencies are received simultaneously. The assumption is considered to be valid when signals from multiple frequencies are collected within the coherent time of the scattering medium. RIM-SA exploits the concept of RIM to generate synthesized time series at several subgates within one range gate for each receiver independently. These synthesized signals can be thought of as regular time series collected in the SA system using a finer resolution. A profile of horizontal wind is estimated using a SA algorithm on synthesized time series data from spaced receivers at every subgate. In practice, transmitting short pulses increases system complexity and the likelihood of interference [Skolnik, 2001]. On the other hand, RIM-based technique can provide improved resolution while a longer pulse is used High-Resolution Synthesized Time Series Using RIM [8] In general, signals from M frequencies and N receivers are considered. Let s ij (t) represent the signals from the ith receiver and the jth transmitting frequency at a given gate. In RIM, only signals from M frequencies at a given receiver i are of interest. The synthesized signals y i (t) at receiver i are defined as a weighted summation of signals from M frequencies and the i receiver [Palmer et al., 1999]. y i ðþ¼w t y s i ðþ t ð1þ where the column vector s i (t) consists signals from M frequencies, the dagger is a Hermitian operator, and the column vector w represents a weighting function. Note that the complex weighting function depends on transmitted frequencies and is designed to modulate 2of13

3 signals from those frequencies to create a constructive interference at a given range. [9] The range brightness estimate (^B(r I )), which represents estimated echo power at a specific range r I, can be obtained by evaluating the autocorrelation function of y i (t) at temporal zero lag [Palmer et al., 1999]. In addition, Luce et al. [2001] has shown that the range brightness estimate is a convolution of the true brightness and a window function W(r I ). ^B ðr I Þ ¼ Br ð I Þ*Wðr I Þ ð2þ where * is a convolution operator. The window function W(r I ) is obtained by taking a wavenumber-range Fourier transformation of w l (dk), which is the correlation function of w and is analogous to the lag window defined in the spectrum estimation. As an example, a sinc-shaped W(r I ) is resulted when signals from multiple frequencies are equally weighed. [10] In previous work, the range brightness was estimated using the correlation matrix made up by s i (t) (e.g., (2) in Palmer et al. [1999]). Therefore, no synthesized signals (1) were needed. In this work, synthesized time series are generated by selecting appropriate ranges r I in the weighting function of (1). The synthesized time series inherit the high resolution provided by RIM and therefore, they can be thought of as signals obtained by a conventional single-frequency system operating in a high-resolution mode. A Doppler spectrum can be estimated at each subgate (r I ) by squaring the time-frequency Fourier transform of y i (t). Echo power, mean radial velocity, and spectrum width can be obtained by estimating the first three moments of a spectrum [Woodman, 1985] High-Resolution RIM-SA [11] The same procedure described in section 2.1 is implemented independently for N receivers. Consequently, SA is applied to synthesized signals at subgates. A range distribution of horizontal wind within the radar volume can be obtained. This combined use of RIM and SA, is termed RIM-SA. [12] There are various techniques to estimate winds from observations made by SA radars [e.g., Briggs, 1984; Briggs and Vincent, 1992; Holloway et al., 1997]. The technique used here is the Full Correlation Analysis (FCA) technique of Briggs [1984]. The technique examines the auto and cross correlation functions of signals received at the separated antennas. The time lag of the peak of the cross correlation function between two antennas gives the average time for the atmosphere to drift past the antennas and thus the apparent velocity of the wind. Corrections are made to account for factors such as time variability (e.g. induced by turbulence) in the signals, to produce a so-called true 3of13 velocity which approximates the wind in the atmosphere. The studies of Cohn et al. [1997] and [2001] reported good agreement between FCA winds produced by MAPR and anemometers on a 300 m tower, as well as with rawinsondes. [13] Note that the performance of RIM is determined by W(r I ) as shown in (2). Palmer et al. [1999] defined two window functions for Fourier and Capon RIM. The Fourier window function has a fixed pattern after the transmitter frequencies are selected, while Capon window function is adaptive to B(r I ) to suppress artificial effects. Palmer et al. [1999] and Luce et al. [2001] showed that Capon RIM is better than Fourier RIM at resolving fine structures including thin layer or multiplelayers. [14] For RIM-SA, the wind estimate ^V at r I can be represented in the following form. ½ ^V ðr I Þ ¼ Br ð IÞVðr I ÞŠ*Wðr I Þ þ e Br ð I Þ*Wðr I Þ ð3þ where V(r I ) is the true velocity and can be either zonal or meridional wind. The first term on the right of (3) represents a smoothed version of true velocity after the synthesized signals (1) are generated. The true velocity at this stage can be obtained when (1) V(r I )isa constant or (2) W(r I ) approaches a delta function. An additional error e produced in the SA analysis [e.g., Zhang et al., 2003] is included in (3). 3. Simulation Results 3.1. Model Description [15] Although sophisticated and realistic models [e.g., Muschinski et al., 1999] exist, a simulation model developed by Holdsworth and Reid [1995] was employed because of its robustness and relative simplicity. This model was originally developed for studies of FCA wind measurements, and has been modified to investigate imaging techniques [Palmer et al., 1999; Yu et al., 2000; Yu and Palmer, 2001]. Signals from multiple receivers and multiple frequencies are generated simultaneously by the summation of individual scatterers within a radar volume at a given sampling time. The phase and amplitude of a scatterer are determined by a two-way path and weighting function in both angle and range directions, respectively. Raw time series data are generated when scatterers are advected through the radar volume by a background wind and turbulent motion. A random noise sequence can be added to raw signals to achieve a desired signal-to-noise ratio (SNR). [16] In the present simulation, a radar configuration based on the MAPR system was used. Four receivers were located at four corners of a square with a length

4 Figure 1. Range distribution of (a) normalized echo power and (b) vertical velocity estimated using both Fourier and Capon RIM at a SNR of 20 db. Modeled values and estimates are indicated by solid and dashed lines, respectively. Note only results from 1.95 km to 2.07 km are shown, where the modeled power is strong. of m. Four transmitting frequencies of , , , and MHz were selected to have nonredundant frequency spacing in order to improve the resolution capability of RIM, as suggested in Palmer et al. [2001]. The radar volume was centered at an altitude of 2.0 km and was defined by a 1.33 ms pulse and a 9 half-power beam width (HPBW). Horizontal and vertical winds were modified to be a function of altitude in order to demonstrate RIM-SA High-Resolution Measurement of Echo Power and Vertical Velocity [17] It has been shown that the atmospheric structure can be imaged within the radar volume in range using RIM [e.g., Palmer et al., 1999; Luce et al., 2001]. The first part of the simulation focused on demonstrating that a high-resolution profile of radial velocity can be obtained using synthesized signals (1). The model profile includes a range weighting function and a layer structure centered at 2.02 km with a width of 0.12 km. In the simulations, the vertical wind was increased linearly with altitude from 0.9 ms 1 to 0.9 ms 1 within the radar volume, while zonal and meridional winds were 20 ms 1 and 0 ms 1, respectively. Raw time series data were obtained at a sampling interval of 16.7 ms for the four receivers and four frequencies simultaneously. The aliasing velocity was approximately 4.9 ms 1. Synthesized time series (1) were generated every 10 m in range using both Fourier and Capon weighting functions for four receivers independently. These synthesized data were then Fourier transformed and squared to obtain power spectrum density (PSD) at 20 subgates within 200 m range gate. The range distribution of echo power, mean radial velocity, and spectral width were obtained by estimating the first three moments of these PSD at 20 subgates. [18] The resulting normalized power and vertical velocity from the first receiver are shown in Figure 1 for both Fourier and Capon methods. Similar results were obtained from the other three receivers. The model profiles of normalized power and vertical velocity are denoted by a solid line. Negative vertical velocity indicates downward motion. Generally speaking, varia- 4of13

5 Figure 2. (a) Normalized spectra estimated using raw time series data of four transmitting frequencies (i.e., with no RIM processing). All four spectra are centered close to zero velocity, while a linearly increasing vertical velocity was used in simulation. (b) Normalized spectra (in linear scale) at 20 subgates calculated using Capon RIM synthesized time series data. Note that variation of power and vertical motion in range can be identified easily in Figure 2b. tions of normalized power and vertical velocity within a 200 m gate can be grossly reconstructed using both Fourier and Capon RIM. Results are degraded toward both boundaries of the gate, where return powers are weak due to the range weighting function and layer structure. Therefore, only results at regions in which the normalized echo power is larger than 0.5 are shown (1.95 km 2.07 km). In practice, the problem can be alleviated by faster range sampling such that contiguous radar volumes overlap in range, and selection of those results with higher power in the overlapping regions. In Figure 1b, it is evident that the variation of vertical velocity is well determined by Capon RIM. [19] An example of the spectra used to obtain results in Figure 1 are shown in Figure 2b for the Capon method. Vertical velocity estimates are denoted by filled triangles. The normalized spectra of signals from four frequencies before RIM processing are shown in Figure 2a in linear scale. The mean radial velocity for the four spectra is approximately 0.08 ms 1. Conventional techniques would just produce this averaged velocity. Figure 2b shows that RIM can be used to determine much of the variation of echo power, vertical velocity, and spectral width in range within the radar volume High-Resolution Measurement of Horizontal Winds [20] The focus of this simulation is to demonstrate that high-resolution measurements of horizontal wind can be achieved using RIM-SA. The parameters were the same as those used previously except that the vertical wind was zero and the horizontal wind was designed to be a function of altitude. The meridional wind linearly increases with altitude from 5 ms 1 to 5 ms 1, while the profile of zonal wind is modeled by a constant velocity of 12 ms 1 minus a Gaussian function centered at 2.0 km with a maximum of 10 ms 1. This special profile of wind is designed to demonstrate that RIM-SA can reconstruct different types of wind variations. The model winds are indicated by solid lines in Figure 3. The 5of13

6 Figure 3. (a) Zonal and (b) meridional components of high-resolution wind field estimated using both Fourier and Capon RIM-FCA. zonal and meridional components estimated by standard FCA, i.e., without RIM, are approximately 6.7 ms 1 and 0.86 ms 1, respectively. These FCA winds are single estimators of motion averaged over the radar volume. On the other hand, a range distribution of horizontal winds can be obtained using Fourier and Capon RIM-FCA, as shown in Figure 3. The Capon method seems to be more successful than the Fourier method at reproducing the wind profiles and this is due to the better resolving capability of Capon RIM. Therefore, only results of Capon RIM-FCA are shown and discussed from now on. [21] Results in Figures 1 and 3 have demonstrated that the range resolution, originally defined by transmitted pulse width, is now effectively improved using RIM analysis. In practice, the system bandwidth required for RIM-SA implementation is the sum of a maximum frequency spacing and the bandwidth used by a transmitted pulse. For example, a 2.75 MHz bandwidth is occupied when a 1.33 ms pulse and 2 MHz frequency span are used. This system bandwidth is equivalent to the one used to transmit a 0.36 ms pulse (approximately range resolution of 54.5 m). Can RIM-SA provide a better resolution than through transmitting short pulses given an available system bandwidth? In order to answer this question, the horizontal wind measured by a 50-m pulse is derived, which is the sum of the model velocity profile (Figure 3) weighted by a normalized power distribution within a 50 m range gate. These 50-m winds are denoted by triangles in Figure 3. In Figure 3b, the 50-m pulse and Capon RIM- FCA have comparable results when the velocity variation is relatively smooth. On the other hand, Capon RIM-FCA can better resolve the wind variation between km and km than the case that 50-m pulses were used, as shown in Figure 3a. It is evident that a resolution better than 50 m could be achieved using RIM-SA. [22] Measures of the error in RIM-FCA, defined by j^v Vj, depend on the interplay between the brightness, the velocity, and the window function. Furthermore, the Capon window function is a function of the brightness as well. Two examples of error analysis for Capon RIM- FCA are presented using simulations. In the first example, velocity profiles and transmitted frequencies are the same as those used previously. However, a single layer located at 2.02 km with various widths was simulated. The error of Capon RIM-FCA is shown in Figure 4 in terms of layer width and range. A contour line at a error 6of13

7 Figure 4. Error of (a) zonal wind and (b) meridional wind estimates using Capon RIM-FCA at a SNR 0f 20 db. The white line denotes a contour line at an error value of 1 ms 1. Regions within or between contour lines can be thought of as regions of reliable estimation of velocity. value of 1 ms 1 is superimposed to indicate regions of reliable estimation. The value of 1 ms 1 is arbitrary, different values can be assigned according to desired accuracy. Generally speaking, the range of reliable wind estimation is extended when the layer width was increased from 20 m to 80 m. More number of reliable winds were obtained when the wind was linearly increased within the range gate, as shown in Figure 4a. A slight degradation of zonal wind measurement at approximately 2.0 km is mainly caused by the smoothing effect described in (3). The second example is presented when double layers exist within one range gate. Two layers both has a width of 20 m and were located symmetrically within a 200 m gate. Results of RIM-FCA error are shown in Figure 5 as a function of layer separation. Figure 5a has shown that reliable RIM-FCA estimations were obtained at regions close to the center of both layers, where the strongest returns occur. On the other hand, this effect is less significant when the variation of wind is relatively smooth. 4. Experimental Results [23] The MAPR system was originally modified from a conventional 915 MHz boundary layer profiler (BLR) 7of13 and can provide boundary layer measurements on high temporal and range resolutions of 30 s and 50 m, respectively [Cohn et al., 1997, 2001]. The antenna has an aperture of 2 m by 2 m which is divided into four panels for receiving. The one-way half power beamwidth is 9 and the peak power of the system is 4.5 kw. Recently, MAPR has been further modified to transmit a maximum of four frequencies generated by independent frequency synthesizers. Preliminary RIM results using MAPR were presented and discussed by Yu et al. [2002]. The feasibility of RIM-SA is demonstrated using MAPR during two experiments. In the first experiment, RIM-SA was tested by alternating two modes with different range resolutions. RIM-SA was further verified by comparing high-resolution results with wind profiles measured by a radiosonde Verification Using a Single Frequency and Short Pulse [24] The first experiment conducted on April 27, 2002 was designed to test and verify high-resolution techniques. MAPR was located in the compound outside the NCAR Foothills Laboratory in Boulder, Colorado, and operated at two modes from 00 UTC to 03 UTC. In the first mode, a single frequency of 915 MHz and a range

8 Figure 5. Same as Figure 4, but for a double-layer case. In the simulations, two layers with the same width of 20 m were separated from 0 m to 135 m every 15 m. The location of the center of the layer is denoted by a white dashed line. resolution of 100 m were used. The data collected in the first mode were processed using a standard FCA to produce profiles of echo power, vertical and horizontal winds. Since MAPR has shown its robust and reliable measurements in this mode [Cohn et al., 1997, 2001], these 100-m-resolution profiles would be considered as a reference. This mode is termed standard mode. In the second mode, a coarse range resolution of 300 m and four frequencies of , , , and MHz were used, and this mode is defined as RIM-SA mode. Signals from four frequencies and four receivers were RIM-FCA processed to produce echo power and wind field every 100 m within the 300 m range gate. The two modes were alternated every two minutes throughout the experiment. The profiles obtained using RIM-FCA are compared with those profiles obtained in standard mode. 50% oversampling in range was employed in RIM-SA mode in order to mitigate errors toward boundaries of the gate. [25] In RIM-SA mode, four frequencies were transmitted on a pulse-by-pulse basis and signals from the same frequency were collected and coherently integrated. The number of data points and inter pulse period are 256 and 63 ms for both modes. A data sampling time of ms and ms is resulted for the standard and RIM-SA modes after 68 and 264 integrations, respectively. Aliasing velocities of approximately 4.92 ms 1 and 4.78 ms 1 were obtained for the two modes. The correlation functions used in FCA and RIM-FCA were obtained after six incoherent integrations. As a result, each estimate of wind and echo power has a temporal resolution of approximately 30 s. [26] Results from standard and RIM-SA modes both at 100 m range resolution are shown in Figures 6a and 6b, respectively. Echo power, vertical velocity, zonal and meridional winds are shown from top to bottom, respectively. Results from standard mode and Capon RIM-FCA analysis are denoted by subscript of S100 and C, respectively. Missing data correspond to samples were rejected by FCA [Briggs, 1984]. Stronger echo powers are generally coincident with downward motions, indicating these echoes were caused by precipitation. However, precipitation did not reach the ground except during the first half hour of the experiment. Surface anemometer and low level winds from MAPR showed veering from northerlies to north-easterlies. South-westerlies aloft strengthened as the precipitation developed. As shown in Figure 6, results from standard FCA and RIM-FCA are qualitatively consistent for echo power, vertical velocity, and horizontal wind. Note that Capon method 8of13

9 Figure 6. Comparison of results using (a) standard FCA and (b) Capon RIM-FCA. Echo power, vertical, zonal and meridional velocities are shown from top to bottom, respectively. Both the standard results and the Capon results are at 100 m resolution, however the Capon results were produced from observations made at 300 m resolution. RIM-FCA produces similar results to standard FCA for echo powers and three-dimensional wind fields. These observations were made in Boulder, Colorado, on April 27, of13

10 Figure 6. (continued) does not provide the absolute power estimates, only their relative values are estimated. [27] Profiles of horizontal wind obtained by standard FCA and RIM-FCA are shown in Figure 7. Zonal and meridional winds are shown on the top and bottom panels, respectively. A two-minute period of RIM-SA mode is labeled on the top of each plot. The median of RIM-FCA winds over this period of time is denoted by a solid line with circles. Each standard FCA wind estimate is the median over two two-minute periods, i.e., the periods 10 of 13

11 Figure 7. Profiles of wind estimated by standard FCA and Capon RIM-FCA at 100 m resolution. Zonal and meridional velocities are shown on the top and bottom panels, respectively. 11 of 13 before and after the RIM-SA mode. Although there is more variability in the RIM-FCA profiles than in the standard mode profiles, the trends are very consistent indicating that RIM-FCA does produce reasonable subgate winds Verification Using a Radiosonde Data [28] The second data set was collected on June 6, 2002, during the International H 2 O project (IHOP). MAPR was located in the Oklahoma panhandle area ( N, W) and operated in a single RIM-SA mode. All parameters in this mode were the same as those used in the first experiment except a 200 m range resolution was used. A Vaisala radiosonde RS80-15GH was launched at 1739 UTC at the MAPR site. Profiles of zonal and meridional wind measured by the radiosonde are indicated by a solid line in Figures 8a and 8b, respectively. It took approximately seven minutes for the sonde to reach a height of 1.5 km above the ground. During this time the radiosonde drifted approximately 4 km from MAPR. MAPR data collected from 1739 UTC to 1746 UTC were RIM-FCA analyzed. RIM-FCA winds were estimated every 20 m with a temporal resolution of 49 s. Mean profiles of RIM-FCA winds over this seven-minute period are denoted by dashed lines. The standard deviation of RIM-FCA estimates are indicated by the length of the error bars. It is clear that the wind profiles produced by RIM-FCA are consistent with those of the radiosonde given the separation and different sampling strategies of the two systems. For example, a sinusoidal variation of zonal velocity was observed by both radiosonde and RIM-FCA at altitudes from 0.4 km to 1.0 km. Additionally, a shear of approximately 3.5 ms 1 over 70 m range at an altitude of 1.1 km was shown in the sonde s meridional profile. A similar shear feature was observed using RIM-FCA (dashed line) with a 200 m pulse. The results demonstrate the potential of RIM-FCA to improve the range resolution of pulsed radar. 5. Conclusions [29] RIM is a recent development that improves the range resolution of conventional pulsed radars by transmitting multiple frequencies. Earlier studies demonstrated the usefulness of RIM for making fine scale measurements of atmospheric backscatter. In this work, simulations demonstrate that Doppler information in these fine scale structures can be revealed from RIM analysis. The synthesized time series time series at

12 Figure 8. Comparison of (a) zonal and (b) meridional wind measured by high-resolution RIM- FCA technique and a radiosonde. The mean profile and temporal variation of RIM-FCA wind over a seven-minute period are denoted by a dashed line and the length of error bar, respectively. subgates are the weighted sum of signals from multiple frequencies. Each synthesized time series can be thought of as time series collected in a finer range resolution. As a result, a Doppler spectrum can be obtained at each subgate and therefore, the echo power, radial velocity, and spectral width can be estimated by the first three moments of the spectrum. [30] The concept of synthesized time series was further exploited to estimate horizontal wind fields by combining signals from spatially separated receivers. A profile of horizontal wind was obtained within each range gate by applying a SA technique to the synthesized time series from the various receivers. As a result, not only fine structures but also dynamical processes such as vertical shear, which are otherwise averaged over the radar volume, can be detected using RIM-SA. It has shown that the performance of RIM-SA depends on transmitted frequencies, reflectivity, and wind within a range gate. In this study, RIM-FCA was first demonstrated using simulation data for both Fourier and Capon methods. Given a system bandwidth, RIM-SA can offer better resolution than the use of short pulses. RIM-FCA measurements were then verified using observations made by MAPR and in situ data. Profiles of echo power and three-dimensional winds obtained in standard mode (100 m) are consistent with these high-resolution profiles obtained using 300 m range resolution and RIM-FCA. RIM-FCA was further verified using simultaneous radiosonde measurement. A study to further compare RIM-SA with other techniques is planned. [31] Acknowledgments. Some work was done when T.-Y. Y was a postdoc at ASP/ATD NCAR. NCAR is sponsored by the National Science Foundation. Thanks to Robert Palmer, Phillip Chilson, and Steve Cohn for useful discussions and suggestions. Thanks are also due to Mike Susedik, Gary Granger, and Charles Martin for modifying the MAPR system. W.O.J.B. and MAPR received support from the the Environmental Meteorology Program of the Department of Energy under the VTMX (Vertical Transport and MiXing) program. References Adams, G., J. Brosnahan, D. Walden, and S. Nerney (1986), Mesospheric observations using a 2.66-MHz radar as an imaging Doppler interferometer: Description and first results, J. Geophys. Res., 91, of 13

13 Briggs, B. (1980), Radar observations of atmospheric winds and turbulence: A comparison of techniques, J. Atmos. Terr. Phys., 42, Briggs, B. (1984), The analysis of spaced sensor records by correlation techniques, in MAP Handbook, vol. 13, pp , Sci. Comm. on Sol.-Terr. Phys. Secr., Univ. of Ill., Urbana. Briggs, B., and R. A. Vincent (1992), Spaced-antenna analysis in the frequency domain, Radio Sci., 27, Brown, W. O. J., and G. J. Fraser (1996), Frequency domain interferometry on spaced antenna MF radar, Radio Sci., 31, Chilson, P. B., T.-Y. Yu, R. G. Strauch, A. Muschinski, and R. D. Palmer (2003), Implementation of range imaging on the Platteville 915-MHz tropospheric profiler, J. Atmos. Oceanic Technol., 20, Cohn, S. A., and P. B. Chilson (1995), NCAR Workshop on Multiple-Receiver and Multiple-Frequency Techniques for Wind Profiling, Bull. Am. Meteorol. Soc., 76, Cohn, S. A., C. L. Holloway, S. P. Oncley, R. J. Doviak, and R. J. Lataitis (1997), Validation of a UHF spaced antenna wind profiler for high-resolution boundary layer observations, Radio Sci., 32, Cohn, S. A., W. O. J. Brown, C. L. Martin, M. S. Susedik, G. Maclean, and D. B. Parson (2001), Clear air boundary layer spaced antenna wind measurement with the Multiple Antenna Profiler (MAPR), Ann. Geophys., 19, Dalaudier, F., D. Sidi, M. Crochet, and J. Vernin (1994), Direct evidence of sheets in the atmospheric temperature field, J. Atmos. Sci., 51, Franke, S. J. (1990), Pulse compression and frequency domain interferometry with a frequency-hopped MST radar, Radio Sci., 25, Holdsworth, D. A., and I. M. Reid (1995), A simple model of atmospheric radar backscatter: Description and application to the full correlation analysis of spaced antenna data, Radio Sci., 30, Holloway, C. L., R. J. Doviak, S. A. Cohn, R. J. Lataitis, and J. S. VanBaelen (1997), Cross correlations and cross spectra for spaced antenna wind profilers: 2. Algorithms to estimate wind and turbulence, Radio Sci., 32, Kudeki, E., and G. R. Stitt (1987), Frequency domain interferometry: A high-resolution radar technique for studies of atmospheric turbulence, J. Geophys. Res., 92, Larsen, M. F., and J. Röttger (1989), The spaced antenna technique for radar wind profiling, J. Atmos. Oceanic Technol., 6, Luce, H., M. Yamamoto, S. Fukao, D. Helal, and M. Crochet (2001), A frequency domain radar Interferometric Imaging (FII) technique based on high resolution methods, J. Atmos. Sol.-Terr. Phys., 63, Muschinski, A., and C. Wode (1998), First in situ evidence for coexisting submeter temperature and humidity sheets in the lower free troposphere, J. Atmos. Sci., 55, Muschinski, A., P. P. Sullivan, D. B. Wuertz, R. J. Hill, S. A. Cohn, D. H. Lenschow, and R. J. Doviak (1999), First synthesis of wind-profiler signals on the basis of large-eddy simulation data, Radio Sci., 34, Palmer, R. D., X. Huang, S. Fukao, M. Yamamoto, and T. Nakamura (1995), High-resolution wind profiling using combined spatial and frequency domain interferometry, Radio Sci., 30, Palmer, R. D., T.-Y. Yu, and P. B. Chilson (1999), Range imaging using frequency diversity, Radio Sci., 34, Palmer, R. D., P. B. Chilson, A. Muschinski, G. Schmidt, T.-Y. Yu, and H. Steinhagen (2001), SOMARE-99: Observations of tropospheric scattering layers using multiple-frequency range imaging, Radio Sci., 36, Skolnik, M. I. (2001), Introduction to Radar System, McGraw- Hill, New York. SmaÏni, L., H. Luce, M. Crochet, and S. Fukao (2002), An improved high-resolution processing method for a frequency domain interferometric imaging (FII) technique, J. Atmos. Oceanic Technol., 19, Stitt, G. R., and E. Kudeki (1991), Interferometric cross-spectral studies of mesospheric scattering layers, Radio Sci., 26, Woodman, R. F. (1985), Spectral moment estimation in MST radars, Radio Sci., 20, Yu, T.-Y. (2000), Radar study of the atmosphere using spatial and frequency diversity, Ph.D. thesis, Univ. of Nebr., Lincoln. Yu, T.-Y., and R. D. Palmer (2001), Atmospheric radar imaging using multiple-receiver and multiple-frequency techniques, Radio Sci., 36, Yu, T.-Y., R. D. Palmer, and D. L. Hysell (2000), A simulation study of coherent radar imaging, Radio Sci., 35, Yu, T.-Y., W. O. J. Brown, S. A. Cohn, D. B. Parsons, M. B. Parlange, and M. Pahlow (2002), High-resolution observations of the boundary layer using multiple-frequency range imaging, in Proceedings of 82nd Annual Meeting, Sixth Symposium on Integrated Observing Systems, pp , Am. Meteorol. Soc., Boston, Mass. Zhang, G., R. J. Doviak, and J. Viekanandan (2003), Crosscorrelation ratio method to estimate cross beam wind and comparison with the full correlation analysis, Radio Sci., 38(3), 8052, doi: /2002rs W. O. J. Brown, Atmospheric Technology Division, National Center for Atmospheric Research, Boulder, CO , USA. (brown@atd.ucar.edu) T.-Y. Yu, School of Electrical and Computer Engineering, University of Oklahoma, Norman, OK , USA. (tyu@ou.edu) 13 of 13