• Albers, S. C., 1995: The LAPS wind analysis. Wea. Forecasting, 10 , 342352.

  • Bargen, D. W., , and Brown R. C. , 1980: Interactive radar velocity unfolding. Proc. 19th Conf. on Radar Meteor., Miami, FL, Amer. Meteor. Soc., 278–283.

    • Search Google Scholar
    • Export Citation
  • Benjamin, S. G., and Coauthors, 2004: An hourly assimilation–forecast cycle: The RUC. Mon. Wea. Rev., 132 , 495518.

  • Bluestein, H. B., , Unruh W. P. , , Dowell D. C. , , Hutchinson T. A. , , Crawford T. M. , , Wood A. C. , , and Stein H. , 1997: Doppler radar analysis of the Northfield, Texas, tornado of 25 May 1994. Mon. Wea. Rev., 125 , 212230.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bousquet, O., , and Chong M. , 1998: A multiple-Doppler synthesis and continuity adjustment technique (MUSCAT) to recover wind components from Doppler radar measurements. J. Atmos. Oceanic Technol., 15 , 343359.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Browning, K. A., , and Wexler R. , 1968: The determination of kinematic properties of a wind fields using Doppler radar. J. Appl. Meteor., 7 , 105113.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chang, P-L., , Wang B-S. , , Shieh S-L. , , and Jou B. J-D. , 2003: Clear-air echo statistics and beam blockage patterns of Taiwan Doppler radar network. Preprints, 31st Conf. on Radar Meteorology, Seattle, WA, Amer. Meteor. Soc., 7B.8. [Available online at http://ams.confex.com/ams/pdfpapers/64485.pdf.].

    • Search Google Scholar
    • Export Citation
  • Collier, C. G., Ed. 1992: International Weather Radar Networking. Kluwer Academic, 344 pp.

  • Crook, N. A., , and Sun J. , 2002: Assimilating radar, surface, and profiler data for the Sydney 2000 Forecast Demonstration Project. J. Atmos. Oceanic Technol., 19 , 888898.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dazhang, T., , Geotis S. G. , , Passarelli R. E. Jr., , Hansen A. L. , , and Frush C. L. , 1984: Evaluation of an alternating-PRF method for extending the range of unambiguous Doppler velocity. Preprints, 22nd Conf. on Radar Meteorology, Zurich, Switzerland, Amer. Meteor. Soc., 523–527.

    • Search Google Scholar
    • Export Citation
  • Eilts, M. D., , and Smith S. D. , 1990: Efficient dealiasing of Doppler velocities using local environment constraints. J. Atmos. Oceanic Technol., 7 , 118128.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gao, J., , Xue M. , , Shapiro A. , , and Droegemeier K. K. , 1999: A variational method for the analysis of three-dimensional wind fields from two Doppler radars. Mon. Wea. Rev., 127 , 21282142.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Holleman, I., , and Beekhuis H. , 2003: Analysis and correction of dual PRF velocity data. J. Atmos. Oceanic Technol., 20 , 443453.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • James, C. N., , and Houze R. A. Jr., 2001: A real-time four-dimensional Doppler dealiasing scheme. J. Atmos. Oceanic Technol., 18 , 16741683.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Joe, P., , and May P. T. , 2003: Correction of dual PRF velocity errors for operational Doppler weather radars. J. Atmos. Oceanic Technol., 20 , 429442.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kim, H-W., , and Lee D-K. , 2006: An observational study of mesoscale convective systems with heavy rainfall over the Korean Peninsula. Wea. Forecasting, 21 , 125148.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kim, Y-S., , Nam K-Y. , , Park O-R. , , and Cho C-H. , 2004: Retrieval of horizontal wind field from Doppler radar radial velocity using a successive correction method (in Korean). J. Kor. Meteor. Soc., 40 , 229239.

    • Search Google Scholar
    • Export Citation
  • Klazura, G. E., , and Imy D. A. , 1993: A description of the initial set of analysis products available from the NEXRAD WSR-88D system. Bull. Amer. Meteor. Soc., 74 , 12931311.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Köck, K., , Leitner T. , , Randeu W. L. , , Divjak M. , , and Schreiber K. J. , 2000: OPERA: Operational Programme for the Exchange of Weather Radar Information: First results and outlook for the future. Phys. Chem. Earth, 25B , 11471151.

    • Search Google Scholar
    • Export Citation
  • Lemon, L., , and Doswell C. A. III, 1979: Severe thunderstorm evolution and mesocyclone structure as related to tornadogenesis. Mon. Wea. Rev., 107 , 11841197.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Makihara, Y., 1996: A method for improving radar estimates of precipitation by comparing data from radars and raingauges. J. Meteor. Soc. Japan, 74 , 459480.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Markowski, P. M., 2002: Hook echoes and rear-flank downdrafts: A review. Mon. Wea. Rev., 130 , 852876.

  • May, P. T., 2001: Mesocyclone and microburst signature distortion with dual PRT radars. J. Atmos. Oceanic Technol., 18 , 12291233.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Meischner, P., , Collier C. , , Illingworth A. , , Joss J. , , and Randeu W. , 1997: Advanced weather radar systems in Europe: The COST 75 action. Bull. Amer. Meteor. Soc., 78 , 14111430.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mohr, C. G., , Miller L. J. , , Vaughan R. L. , , and Frank H. W. , 1986: The merger of mesoscale datasets into a common Cartesian format for efficient and systematic analyses. J. Atmos. Oceanic Technol., 3 , 143161.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nam, K-Y., , Kim Y-H. , , Kim K-E. , , and Nam J-C. , 2005: Study on the multi-cell storm structure using dual Doppler radars observations (in Korean). J. Kor. Meteor. Soc., 41 , 967981.

    • Search Google Scholar
    • Export Citation
  • Ray, P. S., , Wagner K. K. , , Johnson K. W. , , Stephens J. J. , , Bumgarner W. C. , , and Mueller E. A. , 1978: Triple-Doppler observations of a convective storm. J. Appl. Meteor., 17 , 12011212.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ray, P. S., , Ziegler C. L. , , and Bumgarner W. , 1980: Single- and multiple-Doppler radar observations of tornadic storms. Mon. Wea. Rev., 108 , 16071625.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rosenfeld, D., , Wolff B. D. , , and Atlas D. , 1993: General probability-matched relations between radar reflectivity and rain rate. J. Appl. Meteor., 32 , 5072.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rossa, A. M., 2000: COST-717: Use of radar observations in hydrological and NWP models. Phys. Chem. Earth, 25B , 12211224.

  • Rotunno, R., 1986: Tornadoes and tornadogenesis. Mesoscale Meteorology and Forecasting, P. S. Ray, Ed., Amer. Meteor. Soc., 414–436.

  • Steiner, M., , and Smith J. A. , 2002: Use of three-dimensional reflectivity structure for automated detection and removal of nonprecipitating echoes in radar data. J. Atmos. Oceanic Technol., 19 , 673686.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Suk, M-K., , Nam K-Y. , , Kim Y-H. , , and Oh S-N. , 2005: Estimation of quantitative rain intensity from radar reflectivities using a window probability matching method (in Korean). J. Kor. Meteor. Soc., 41 , 123138.

    • Search Google Scholar
    • Export Citation
  • Wakimoto, R. M., , Liu C. , , and Cai H. , 1998: The Garden City, Kansas, storm during VORTEX 95. Part I: Overview of the storm’s life cycle and mesocyclogenesis. Mon. Wea. Rev., 126 , 372392.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xiao, Q., , Kuo Y-H. , , Sun J. , , Lee W-C. , , Lim E. , , Guo Y-R. , , and Barker D. M. , 2005: Assimilation of Doppler radar observations with a regional 3DVAR system: Impact of Doppler velocities on forecasts of a heavy rainfall case. J. Appl. Meteor., 44 , 768788.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xiao, Q., , Kuo Y-H. , , Sun J. , , Lee W-C. , , Barker D. M. , , and Lim E. , 2007: An approach of radar reflectivity data assimilation and its assessment with the inland QPF of Typhoon Rusa (2002) at landfall. J. Appl. Meteor. Climatol., 46 , 1422.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    Topography and locations of the operational Doppler radars over the southern Korean Peninsula. The plus sign, filled circle, and open triangle denote the radars operated by USAF, KMA, and KAF, respectively, and the boldface and italic letters denote 10- and 5.5-cm wavelength radars, respectively. The open circles with a cross and open circles denote the four UHF wind profiler and seven rawinsonde sites, respectively.

  • View in gallery

    Wind-retrievable area (gray shaded area) at heights of (a) 2, (b) 4, (c) 7, and (d) 13 km. The plus sign, filled circle, and open triangle denote the radars used as defined in Fig. 1.

  • View in gallery

    PPI images of the radial velocity of the RMYN radar at an elevation angle of 1.5° at 0305:45 UTC 10 Jul 2006: (a) the observed velocity and (b) the corrected velocity.

  • View in gallery

    PPI images of the (a) observed radial velocity and (b) filtered radial velocity of the RKWK radar at an elevation angle of 3.1° at 0303:12 UTC 10 Jul 2006.

  • View in gallery

    Composite images of the synthesized radar reflectivity and horizontal wind vectors retrieved from the radar network at (a) 1800 UTC 9 Jul, (b) 0000 UTC 10 Jul, (c) 0600 UTC 10 Jul, and (d) 0900 UTC 10 Jul 2006. The open circles with a cross denote the typhoon center. The half and full wind barbs denote 2.5 and 5 m s−1, respectively. The pennant denotes 25 m s−1.

  • View in gallery

    Comparisons of horizontal wind vectors retrieved from the radars and observed by the rawinsondes: (a) 1800 UTC 9, (b) 0000 UTC 10, and (c) 0600 UTC 10 Jul 2006.

  • View in gallery

    Comparisons of the horizontal wind vectors retrieved from the radars and measured by the wind profilers during 1800 UTC 9 Jul–1100 UTC 10 Jul 2006 with an interval of 1 h at the (a) WPMUS, (b) WPKNG, (c) WPKSN, and (d) WPMAS wind profilers.

  • View in gallery

    Scatterplots of the horizontal winds obtained from radars and wind profilers: (a) u and (b) υ components.

  • View in gallery

    Composite images of radar reflectivity and horizontal wind vectors at heights of (a) 2 and (b) 6 km at 1440 UTC 30 Jun 2005. The half and full wind barbs denote 2.5 and 5 m s−1, respectively. The pennant denotes 25 m s−1.

  • View in gallery

    Vertical cross sections of the (a) radar reflectivity and cross-sectional wind vector, (b) vertical air velocity, (c) horizontal divergence, and (d) vertical vorticity along the A–B line in Fig. 9. The overlapping contours denote the radar reflectivity.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 50 50 2
PDF Downloads 31 31 1

Retrieval of High-Resolution Wind Fields over the Southern Korean Peninsula Using the Doppler Weather Radar Network

View More View Less
  • 1 Atmospheric Science Program, School of Earth and Environmental Sciences, Seoul National University, Seoul, Korea
© Get Permissions
Full access

Abstract

The performance of a radar network for retrieving high-resolution wind fields over South Korea is examined. The network consists of a total of 18 operational radars. All of the radars possess the Doppler capability and carry out plan position indicator (PPI) volume scans comprising 6–15 elevation steps at every 6 or 10 min. An examination of the coverage of the radar network reveals that the radar network allows the retrieval of three-dimensional high-resolution wind fields over the entire area of the southern Korean Peninsula as well as nearby oceans above a height of approximately 3 km. After the quality control procedures of the radar measurements, the high-resolution wind fields (a few kilometers) are extracted using multiple-Doppler wind synthesis in the Custom Editing and Display of Reduced Information in Cartesian Space (CEDRIC) package developed by NCAR. The radar-retrieved winds are evaluated using the following two rain events: 1) Typhoon Ewiniar in 2006, which resulted in strong winds and heavy rainfall over the entire southern Korean Peninsula, and 2) a well-developed hook echo with a relatively small-scale diameter of about 30 km. The wind fields retrieved from the radar network exhibit counterclockwise rotation around the typhoon center and a general structure around a hook echo such as a cyclonically rotating updraft (i.e., mesocyclone). Comparisons with the wind measurements from four UHF wind profilers for the typhoon case reveal that the u- and υ-wind components retrieved from the radar network deviate by standard deviations of 3.6 and 4.5 m s−1 over ranges from −30 to 20 m s−1 and from 0 to 40 m s−1, respectively. Therefore, it is concluded that the operational radar network has the potential to provide three-dimensional high-resolution wind fields within the mesoscale precipitation systems over almost the entire area of the southern Korean Peninsula.

Corresponding author address: Prof. Dong-Kyou Lee, Atmospheric Science Program, School of Earth and Environmental Sciences, Seoul National University, Seoul 151-747, Korea. Email: dklee@snu.ac.kr

Abstract

The performance of a radar network for retrieving high-resolution wind fields over South Korea is examined. The network consists of a total of 18 operational radars. All of the radars possess the Doppler capability and carry out plan position indicator (PPI) volume scans comprising 6–15 elevation steps at every 6 or 10 min. An examination of the coverage of the radar network reveals that the radar network allows the retrieval of three-dimensional high-resolution wind fields over the entire area of the southern Korean Peninsula as well as nearby oceans above a height of approximately 3 km. After the quality control procedures of the radar measurements, the high-resolution wind fields (a few kilometers) are extracted using multiple-Doppler wind synthesis in the Custom Editing and Display of Reduced Information in Cartesian Space (CEDRIC) package developed by NCAR. The radar-retrieved winds are evaluated using the following two rain events: 1) Typhoon Ewiniar in 2006, which resulted in strong winds and heavy rainfall over the entire southern Korean Peninsula, and 2) a well-developed hook echo with a relatively small-scale diameter of about 30 km. The wind fields retrieved from the radar network exhibit counterclockwise rotation around the typhoon center and a general structure around a hook echo such as a cyclonically rotating updraft (i.e., mesocyclone). Comparisons with the wind measurements from four UHF wind profilers for the typhoon case reveal that the u- and υ-wind components retrieved from the radar network deviate by standard deviations of 3.6 and 4.5 m s−1 over ranges from −30 to 20 m s−1 and from 0 to 40 m s−1, respectively. Therefore, it is concluded that the operational radar network has the potential to provide three-dimensional high-resolution wind fields within the mesoscale precipitation systems over almost the entire area of the southern Korean Peninsula.

Corresponding author address: Prof. Dong-Kyou Lee, Atmospheric Science Program, School of Earth and Environmental Sciences, Seoul National University, Seoul 151-747, Korea. Email: dklee@snu.ac.kr

1. Introduction

Many countries have constructed an operational radar network to watch and predict meteorological phenomena over all of its regions. The Next Generation Radar (NEXRAD) network was established in the United States in order to predict flash floods, warn about aviation hazards, estimate rainfall for water resource management, and protect military operations and installations (Klazura and Imy 1993). The Japan Meteorological Agency (JMA) has been operating a network consisting of 19 operational radars, including conventional and Doppler radars (Makihara 1996). In Taiwan, the radar network composed of four Doppler radars widely covering over the island has been operated since 2001 (Chang et al. 2003). In Europe, most countries also have their own national operational radar network consisting of several conventional and Doppler radars and sometimes a few polarimetric radars. For the past 30 yr, countries in the European Union have made efforts to combine the national networks of each member country and utilize the radar information over the entire European region, as considered in the series of Cooperation in Science and Technology (COST) actions (Collier 1992; Meischner et al. 1997; Rossa 2000) and the Operational Programme for the Exchange of Weather Radar Information (OPERA; Köck et al. 2000). A main purpose of these radar networks is to collect high-resolution meteorological information such as wind fields and rain amounts over all regions of a country; to improve the detection and forecasting of hazardous weather phenomena; and finally to prevent human and socioeconomic losses.

In South Korea, a total of 18 operational radars have been operated since the summer of 2006: 2 radars are owned by the U.S. Air Force (USAF), 11 radars by the Korea Meteorological Administration (KMA), and 5 radars by the Korean Air Force (KAF). The observable ranges of these radars entirely cover the southern Korean Peninsula. Despite having 18 radars in operation in South Korea, the utilization of these radars in hydrometeorological studies is still restricted. Suk et al. (2005) used reflectivity measurements of the KMA radars to estimate the rainfall amounts over South Korea using the window probability matching method algorithm (Rosenfeld et al. 1993). With regard to studies on wind estimation, Kim et al. (2004) applied a few KMA radars to the wind retrieval algorithm proposed by Albers (1995). Nam et al. (2005) analyzed the structure and evolution of a severe convective storm that was detected by two radars southwest of the Korean Peninsula. Kim and Lee (2006) found the characteristics of mesoscale convective systems (MCSs) accompanied by heavy rainfall that developed over the central Korean Peninsula by analyzing a single Doppler velocity measurement.

In contrast to the networks of the European countries and Japan, all of the operational radars throughout South Korea have Doppler capabilities. The radars are separated by a distance of approximately 120 km on average and their unambiguous ranges are greater than 100 km; hence, most areas over the southern Korean Peninsula are overlapped by the coverage of two or more radars. These characteristics of the Korean operational radar network allow the retrieval of high-resolution winds by dual- or multiple-Doppler wind synthesis over almost the entire southern Korean Peninsula region. High-resolution wind fields are essential for improving the understanding and forecasting of hazardous precipitation systems and hence they have been highly requested by many researchers and forecasters in South Korea.

The purpose of this study is to examine the performance of the operational radar network for retrieving high-resolution wind fields over the entire area of South Korea. Under the present operational radar network and their scanning strategies, areas where winds can be retrieved are first examined. High-resolution winds are then retrieved for two rain events using a multiple-Doppler wind synthesis and they are compared with those measured by rawinsondes and wind profilers, in order to evaluate whether the operational radar network can be employed to obtain reliable winds over South Korea. In section 2, the operational radars in South Korea are introduced. An algorithm for retrieving wind fields from the radar network, including the preprocessing of radar measurements, is outlined in section 3. The radar wind retrievals are evaluated in section 4 for two rain events: a typhoon and a well-developed hook echo. In section 5, the results are summarized and future works for operational use in real time are described.

2. Radar network

a. Operational Doppler radars

Figure 1 shows the locations of the operational radars used in this study and Table 1 lists the operational volume scanning strategies of those radars. The RKSG and RKJK radars are operated by the USAF and are the Weather Surveillance Radars-1988 Doppler (WSR-88Ds). The KAF radars are RKWJ, RSCN, RTAG, RYCN, and RWNJ, which are operated near 5.5-cm wavelength. The others—12 radars operated near 5.5- or 10-cm wavelengths—are operated by KMA. All of the KMA and KAF radars are equipped with klystron tubes as transmitters and employ the pulse pair processing (PPP) algorithm for processing the Doppler velocity signals. The radar network shown in Fig. 1 and outlined in Table 1 is the one in use by the summer of 2006. KMA continues to upgrade the 5.5-cm wavelength radars to 10-cm wavelength radars, mainly in order to mitigate the attenuation effect due to precipitation. The RKSN radar was upgraded to a 10-cm wavelength Doppler radar during 2006–07. The existing RDNH radar at 5.5-cm wavelength will be upgraded to a 10-cm wavelength Doppler radar by the summer of 2009. The site of RDNH will also be altered, because its present location is severely affected by occultation due to the Taebaek Mountain Range, which runs from the north to south with elevations of over 1000 m to the west of RDNH, as shown in Fig. 1. Another 5.5-cm wavelength radar, RCJU, was retired in the summer of 2006 in accordance with the start of the observations of two new 10-cm wavelength radars of RGSN and RSSP on Cheju Island.

As shown in Fig. 1, 18 radars have been impartially distributed over the South Korea region, spanning an area of approximately 100 000 km2. This distribution indicates that the radar network of South Korea is very dense compared with those of other countries. The JMA operates 19 radars over an area of 370 000 km2 (Makihara 1996) and the NEXRAD network over the contiguous United States consists of a total of 136 radars (Klazura and Imy 1993) over an area of 8 296 000 km2. Thus, the radar network in South Korea has about 3.5 and 11 times denser coverage than those of the JMA and NEXRAD networks, respectively.

Another notable advantage is that all of the operational radars possess Doppler capabilities and thus they can measure the radar reflectivity factor (Z), radial velocity (Vr), and spectrum width (SWD), as shown in Table 1. With regard to the velocity measurements, the radars have an unambiguous range from 100 to 250 km, with a range resolution from 0.125 to 1.0 km, and an azimuth resolution of approximately 1°. The USAF radars are mostly operated on the schedule of the volume coverage pattern 21 (VCP 21) for severe precipitation among the several volume scanning strategies of NEXRAD. That is, a volume scan is framed by full 360° scans at nine elevation angles (0.5°, 1.45°, 2.4°, 3.35°, 4.3°, 6.0°, 9.9°, 14.6°, and 19.5°). This volume scan repeats every 6 min. The KAF radars frame a volume scan with six to nine elevation angles from 0.5° to 35.0°. One volume scan of the KAF radar needs approximately 4 min and this is repeated every 10 min. The KMA radars, except for RBRI, RCJU, RDNH, and RKSN, carry out one volume scan with 10–15 elevation angles from 0° to 45° every 10 min.

The four KMA radars at 5.5-cm wavelength (RBRI, RCJU, RDNH, and RKSN) scan at relatively low elevation angles (below 7.0°). This is because the main purpose of these radars is to measure the radar reflectivity at a long distance for observing precipitation rather than for making velocity measurements. Therefore, these four radars are operated at relatively low pulse repetition frequencies (PRFs) of between 250 and 499 Hz; accordingly, their unambiguous ranges are relatively long—from 240 to 256 km—as compared with the velocity measurements of general 5.5-cm wavelength Doppler radars. Further, the low PRFs result in low unambiguous (or Nyquist) velocities. These four radars have unambiguous velocities of 3.5 or 6.7 m s−1, while the other radars have the velocities greater than 13 m s−1. In particular, the unambiguous velocities of the 10-cm wavelength radars of KMA (e.g., RGDK, RKWK, RSSP, etc.) are greater than 30 m s−1 due to the extension of unambiguous velocity by the dual-PRF capability using approximately 400 and 600 Hz.

The locations of the rawinsondes and wind profilers that are used in comparison with winds from the radars are also plotted in Fig. 1. A total of seven rawinsondes are presented and they are operated either by KMA or KAF: Baeckreong Island (BRI), Osan (OSN), Sockcho (SCH), Hucksan Island (HSI), Kwangju (KWJ), Pohang (POH), and Gosan (GSN). They are launched every 6 h. Wind profilers are used at four sites—Munsan (WPMUS), Kangnung (WPKNG), Kunsan (WPKSN), and Masan (WPMAS), which are operated by KMA. The wind profilers are operated at a frequency of 1290 MHz and measure the wind components using five beams: one in the vertical and four in oblique directions slanted by 17° from the vertical toward the north, east, south, and west. The wind is measured every 10 min in two different modes: one is a high-altitude mode in which winds up to a height of 12 km are measured with an interval of 165 m and the other is a low-altitude mode whose maximum observable altitude is 5 km with an interval of 72 m. In this study, the winds from the high-altitude mode are used to compare with those retrieved from the radar network.

b. Wind-retrievable area

The highly dense radar sites and volume scans imply that the radar network can collect the radar measurements with high spatial and temporal resolutions over the southern Korean Peninsula. However, since the observations of the individual radars suffer from occultation by the complex orography shown in Fig. 1, the radar network should comprise as many radars as possible, including the KMA, KAF, and USAF radars. Figure 2 shows the areas where winds can be retrieved from the operational radars on the basis of their present scanning strategies by using the same methodology as that described in section 3b. To take into consideration the occultation of radar beams by the complex topography, a digital elevation model (DEM) with an interval of 3 s (horizontal distance of approximately 100 m) was used. The effect of the earth’s curvature was also considered. This examination of the wind-retrievable area was conducted in Cartesian coordinates with a resolution of 5 km × 5 km. The observations from the RBRI, RCJU, RDNH, and RKSN radars were not included in this examination because their unambiguous velocities are too low to resolve the aliasing effect of the radial velocity measurements.

At a height of 2 km (Fig. 2a), it is not possible to retrieve high-resolution winds over the entire southern Korean Peninsula region. The eastern and western regions of Cheju Island are blind areas for wind retrievals because the observations from the RGSN and RSSP radars are occulted by Mount Hanra, whose altitude is 1950 m. In addition, the RJNI radar cannot scan the regions due to the earth’s curvature. The eastern costal region (around the RDNH radar) of the Korean Peninsula is a blind area due to occultation by the Taebaek Mountain Range. The radar observations in the inland area also suffer from occultation by mountains. In addition, winds cannot be retrieved at areas near the baselines connecting the radars (e.g., the areas between RKSG and RKJK, RGDK and RKWK, etc.), because of the high geometric errors of the radars with respect to the grid points. Thus, the wind-retrievable area at the low height of 2 km is irregular and relatively small. On the other hand, at a height of 4 km (Fig. 2b), the wind-retrievable area covers almost the entire southern Korean Peninsula region and nearby oceans. However, the eastern costal region around the RDNH radar is still a no-wind-retrieval area, because the RDNH radar observations could not be used and also because of the occultation by the Taebaek Mountains. In the case of higher altitudes such as 7 (Fig. 2c) and 13 km (Fig. 2d), the wind-retrievable area covers almost the entire southern Korean Peninsula region. Thus, in cases where radar echoes are detected, the present operational Doppler radar network can be employed for retrieving high-resolution wind fields over the entire southern Korean Peninsula at heights above approximately 3 km (all the heights are AGL).

The wind-retrievable area will be widened after upgrading the existing RBRI, RDNH, and RKSN radars to a 10-cm wavelength Doppler radar system similar to the RGSN and RSSP radars, particularly over the northwestern and northeastern seas of South Korea. In addition, the present radar network shown above may be organized better for retrieving wind fields by means of optimizations such as the rearrangement of the radar locations, reconfiguration of their volume scanning strategies (e.g., altering their elevation steps and resolutions), and improvements in the accuracy of the radar measurements by calibration and reconfiguration of the radar systems. This effort, however, is beyond the scope of this study, which is to utilize the present operational radar network for retrieving wind fields. This is a challenging research area for better use of the radar network over South Korea.

3. Wind retrieval

a. Preprocessing of the radar measurements

Noises and errors embedded in the radar measurements are first processed by the on-site signal processors of the individual radars. The remaining ground clutter and the echoes affected by the partial blockage of the radar beams were eliminated by mapping the radar measurements onto the DEM data used in section 2b. The anomalous echoes were removed manually using images of the radar measurements. Further, the radar measurements were processed in terms of the aliasing effect and the dual-PRF velocity error.

1) Correction of aliasing effect

The aliasing effect of the radial velocity measurements is corrected for under a spatial continuity constraint along the radial and azimuth directions. The radial velocities observed by radar are first processed along the radial direction at each azimuth ray, starting from the gate adjacent to the radar at which the observed velocity is assumed to be a true value unaffected by the aliasing effect. In this study, the correction procedure is initiated from the 10th range gate, where a mean velocity over the previous four consecutive gates is used. This first step is based on a method proposed by Bargen and Brown (1980). After terminating at all the azimuth rays, the correction procedure is then performed along the azimuthal direction. In this second step, winds derived from a simplified velocity–azimuth display (VAD; Browning and Wexler 1968) method are used as the reference values for determining whether the radial velocities are aliased or not. The simplified VAD method fits the radial velocities (Vr) to a form of Fourier expansion with two harmonic coefficients (a1 and a2) defined as
i1520-0434-24-1-87-e1
where θ is the azimuth angle of the antenna. The coefficients a1 and a2 are related to u0cosϕ and υ0cosϕ, respectively, where u0 and υ0 denote the mean horizontal wind components, and ϕ is the elevation angle of the antenna.
In cases where the radial velocity at the starting gate in the first step is already aliased, the aliased velocities at further gates are not corrected properly. To overcome this, in the first step, the wind measurements at the surface observatories around the radar are compared. A 10-min mean wind of the surface observatories within a radius of 30 km from the radar was used. A total of about 130 surface observatories are impartially distributed over the southern Korean Peninsula region, and they measure the fundamental meteorological parameters (wind, pressure, temperature, humidity, and rain amount) every minute. The wind from the surface observatory is related to the radial velocities at each azimuth as
i1520-0434-24-1-87-e2
where uenv and υenv denote the horizontal wind components measured by the surface observatories and Vf denotes the fall velocity of a particle. The radial velocities derived from (2) were used as the reference values in a manner similar to that done for the VAD curve in the second step. Here, the mean wind from the surface observatories is employed only for the starting gate at the lowest elevation scan, whose elevation angle is below 1°, and hence the third term of (2) was neglected because it is relatively very small compared to other terms. The correction at higher-elevation scans above the lowest elevation is performed by using the radial velocities corrected previously at an elevation step immediately below, instead of the winds from the surface observatories.

Figures 3 shows an example for supporting the reliability of the correction algorithm described above as taken from the RMYN radar, which has the lowest unambiguous velocities (13.5 m s−1; see Table 1) among the radars used in the wind retrieval. Figure 3a shows the PPI image of the radial velocity observed by the RMYN radar at 0305:45 UTC 10 July 2006 during the passage of Typhoon Ewiniar (TY0603). At this time, the center of the typhoon was located southwest of the radar, as will be shown in section 4. Two surface observatories within a radius of 30 km from the RMYN radar measured the 10-min-mean wind speed and direction to be 7.8 m s−1 and 100.8°, respectively. Therefore, it can be expected that pattern of the radial velocites observed by RMYN should be south-southeasterly winds, which corresponds to movement toward the rader (negative velocities) at points south of the radar, and vice versa at points north of the radar. However, the pattern of the observed radial velocities shown in Fig. 3a does not depict the expected winds, but the velocities within a range of the unambiguous velocity (±13.5 m s−1) are repeated along radial directions. In contrast to the pattern of the observed radial velocities, the corrected velocities effectively reveal the expected south-southeasterly winds (Fig. 3b). In addition, the corrected radial velocities do not show a local discontinuity, but they gradually vary along the radial and azimuth directions over the entire region. Although the correction algorithm above produced good results for the two rain cases analyzed in this study, for operational usage this algorithm needs to be examined and improved upon in various situations such as echo tops characterized by high shear and no echoes at consecutive wide gates, conditions that are frequently encountered in real radar observations. A more complicated and robust method, such as the methodology of Eilts and Smith (1990) or James and Houze (2001), is required.

2) Filtering of dual-PRF velocity error

In contrast to the 5.5-cm wavelength radars of KMA and KAF, the aliasing effect is insignificant in the measurements of the 10-cm wavelength radars of KMA due to the extended unambiguous velocity over 30 m s−1, which is a result of the dual-PRF capability (Dazhang et al. 1984). However, the use of the dual-PRF capability results in errors with a type of speckling in the images of the radial velocity measurements (May 2001; Joe and May 2003; Holleman and Beekhuis 2003). As an example of the dual-PRF velocity error, Fig. 4a shows the speckling that occurred in the velocity measurements of the RKWK radar at 0303:12 UTC 10 July 2006. The RKWK radar is operated at PRFs of 600 and 400 Hz (3:2 ratio), as shown in Table 1. A region around the coordinates (100, 10) displays the speckling with velocities over +20 m s−1. However, the velocities from −10 to +10 m s−1 are prevalent. Speckling is also seen around the coordinates (150, 50), (30, 90), (30, −30), and (70, −40).

In this study, the speckling due to the dual-PRF velocity error was removed by a simple technique where the radial velocity at a given gate was compared with the mean value within an area surrounded by 15 gates along the radial direction and seven rays along the azimuth direction centered at the given gate. That is, when the difference from the mean value was greater than 10 m s−1, the radial velocity at the given gate was considered to be noise and therefore removed. Figure 4b shows the radial velocities filtered out by the elimination technique. As compared to the observed velocities, it is shown that the dual-PRF velocity errors with a particular type of speckling are completely eliminated, although a few neighboring gates that were unaffected by the dual-PRF error were also eliminated. The threshold (10 m s−1) and the area (seven azimuth rays and 15 range gates) used in the elimination technique were determined empirically based on experiments using various values. There are several techniques for correcting the velocities affected by dual-PRF errors, such as those proposed by Joe and May (2003) and Holleman and Beekhuis (2003). Using these techniques in the KMA radars at 10-cm wavelength may improve analyses at finer scales by recovering the velocities affected by the dual-PRF error and by not deleting them was as done in this study.

After the correction of the aliasing effect and the elimination of the dual-PRF velocity error, the radial velocities were finally filtered and smoothed by means of a median filtering scheme over seven consecutive gates in order to eliminate the high-frequency fluctuations from gate to gate.

b. Wind retrieval

Once the measurements of the individual radars were processed by the procedures mentioned above, the wind fields were then retrieved using the Sorted Position Radar Interpolation (SPRINT) and Custom Editing and Display of Reduced Information in Cartesian Space (CEDRIC) packages of the National Center for Atmospheric Research (NCAR) (Mohr et al. 1986). With regard to the interpolation onto Cartesian coordinates by SPRINT, the site of the RKSG radar was taken to be the center of the analysis region as shown in Fig. 1, which corresponds to an area from −300 to 300 km in the east–west direction and from −500 to 300 km in the south–north direction. The bilinear scheme was employed to interpolate the radar data onto Cartesian coordinates from the radar coordinates.

The CEDRIC codes were modified slightly so as to synthesize up to 20 radars in order to include all the radars over South Korea. The u and υ components at each grid point were retrieved from the three-equation solution using three or more radars (Ray et al. 1978, 1980). When the velocity measurements from more than three radars are available, the wind components are mathematically overdetermined and they are calculated in the least squares sense from the velocity measurements obtained from all of the radars. This leads to an improvement in the accuracy and stability of the wind retrievals in a mathematical sense. Since the accuracy of the wind components depends on the geometry of the radars with respect to the points at which the wind is measured, in this study the winds were extracted only from regions where the geometric errors were less than 2 m s−1. In the wind retrieval process using CEDRIC, note again that the velocity measurements from the four radars (RBRI, RCJU, RDNH, and RKSN) were not used because of their low unambiguous velocities as mentioned in section 2, but their reflectivity measurements were used in the derivation of a synthesized reflectivity field.

The w component was derived from the integration of the anelastic continuity equation using a variational scheme with the boundary condition of w = 0 at the surface and echo top (the height is assumed to be 17 km) with an assumption that the air density decreases exponentially with height. Although the w component can be directly extracted from the three-equation solution, it is known that the w component extracted directly is more erroneous than that derived from vertical integration (Ray et al. 1978). The study of Ray et al. showed that the w component extracted directly had an extremely large error near the surface due to a singularity and that its error decreased with increasing height (h), which corresponded to the 1/h2 dependency. Finally, the horizontal wind components were variationally adjusted to minimize the mean difference between the horizontal and vertical divergences, and then the horizontal divergence and vertical vorticity were derived.

4. Results

To examine the performance of the operational radar network for retrieving wind fields, two rain events were examined: 1) Typhoon Ewiniar (TY0603), which was accompanied by strong winds over the entire area of the southern Korean Peninsula, and 2) a distinct hook echo surrounding a weak echo region (WER) with a diameter of about 30 km.

a. Case I: Typhoon Ewiniar

The radar observations during 1800 UTC 9 July to 1100 UTC 10 July 2006, when the typhoon had passed through the southern Korean Peninsula, were used for the performance examination of the radar network. During this period, the RSCN radar was not operated. The RKSN radar was not operated because it was being upgraded to a new system, as mentioned earlier. The RSSP radar was also not operated during the analysis period. The RGSN radar was operated starting at 2200 UTC 9 July. The other radars routinely scanned according to the volume scanning strategies described in section 2. Based on these radar observations, high-resolution winds with a horizontal resolution of 5 km × 5 km and a vertical resolution of 0.5 km were retrieved over the entire area of the southern Korean Peninsula. In the wind retrieval procedure using the CEDRIC package, the moving velocity of the typhoon system was determined from the movement of the typhoon center as reported by the Regional Specialized Meteorological Center (RSMC) Tokyo-Typhoon Center and was applied to displace the points of the radar measurements to new locations.

Figure 5 shows the retrieved winds and the synthesized reflectivity fields from the radar network at a height of 3 km. The retrieved winds effectively depict the cyclonic circulations around the typhoon center, and their movement also agrees with the movement of the typhoon center with time. It is worthwhile to note in Fig. 5 that these high-resolution winds are retrievable over the entire area of the southern Korean Peninsula using the operational radar network. No wind retrievals, however, were obtained around Cheju Island at 1800 UTC 9 July (Fig. 5a) because the RGSN and RSSP radars were not operating at this time.

To evaluate the radar-retrieved winds shown in Fig. 5, they were compared with the winds observed from rawinsondes and wind profilers shown in Fig. 1. To negate the differences in the sampling of the radar observations and the rawinsonde and wind profiler observations, the radar-retrieved winds were first averaged in time and space. For temporal consistency between the radar volume and the rawinsonde observations, the radar-retrieved winds every 10 min were averaged for 1 h. For spatial consistency, the radar-retrieved winds were averaged over nine grids around the grid point closest to each rawinsonde and wind profiler sites at each height.

Figure 6 shows the comparison of the horizontal wind vectors from the radars and rawinsondes. At 1800 UTC 9 July, the typhoon center was located at a distance of about 80 km southwest of Cheju Island and the radar-retrieved winds were obtained over only the southern edge of the Korean Peninsula, as shown in Fig. 5a. Therefore, only the radar-retrieved winds over the rawinsonde site at KWJ could be compared. As shown in Fig. 6a, the radar-retrieved winds at KWJ agree with the rawinsonde winds, although there are slight differences at a height of around 6 km. At 0000 UTC 10 July when the typhoon center was located north of Cheju Island (Fig. 5b), the winds of the rawinsondes at OSN, KWJ, and POH could be compared, while no radar-retrieved winds were obtained over the other rawinsonde sites because of a lack of radar echoes. As shown in Fig. 6b, the radar-retrieved winds correspond well to the variations in the rawinsonde winds with height. At 0600 UTC 10 July (Fig. 6c), only the OSN could be compared, while the other sites had neither radar-retrieved winds (KWJ, HSI, and GSN) due to the absence of radar echoes nor rawinsonde observations (BRI, SCH, and POH). Thus, the radar-retrieved winds at Osan are in good agreement with those obtained from the rawinsonde. For the pairs of the winds shown in Fig. 6, the standard deviations of the u and υ components are 3.5 and 7.8 m s−1, respectively.

For a more quantitative estimation of the reliability of the radar-retrieved winds, a comparison was performed between the winds obtained by the wind profilers and the radar-retrieved winds, because only a few data pairs could be compared in Fig. 6 and further the samplings of the radar observations are closer to those of the wind profilers than those of the rawinsondes. Figure 7 shows the comparisons between the horizontal wind vectors from the radar network and the four wind profilers from 1800 UTC 9 July to 1100 UTC 10 July with an interval of 1 h. It is shown that the radar-retrieved winds are in good agreement with those from the wind profilers. The variations in the wind speeds and directions with time and height are comparable. The comparisons with WPKNG (Fig. 7b), however, reveal relatively large differences as compared to the other wind profiler sites. At heights above 5 km, the southerly wind is prevalent in the winds obtained by the wind profiler, while in the radar-retrieved winds the southeasterly wind is more likely. In addition, no wind was retrieved below a height of 4 km. These results can be ascribed to the fact that the WPKNG site is located near the edge of the wind-retrievable area (Fig. 2), where the RWNJ and RYCN radar observations were obstructed by the Taebaek Mountains and the RDNH radar measurements could not be used due to too low unambiguous velocity. In addition, near the edge region the radar measurements from the individual radars have relatively low reliability because the radar sampling beams are widened at long distance from each radar and the gate spacing in the azimuthal direction is also widened. The winds retrieved from the small number of radars and the low reliability of the radar measurements are easily affected by noises and errors inherent in the radar measurements and consequently they may tend to be erroneous.

Figure 8 shows the scatterplots of the winds shown in Fig. 7. The u components from the radars agree with those from the wind profilers (Fig. 8a) with a standard deviation (STD) of 3.6 m s−1 over the range from −30 to 20 m s−1. However, it is shown that the u components from the radars are slightly smaller than those obtained from the wind profilers. This underestimation is mainly due to the differences in the wind directions over the WPKNG site, as shown in Fig. 7b. On the other hand, the υ components obtained from the radars correspond well to the one-to-one line of those obtained from the wind profilers (Fig. 8b) with a standard deviation of 4.5 m s−1. This deviation of the υ components is trivial, considering that all of the υ components have values ranging from 0 to 40 m s−1. The standard deviations for wind speed and direction are 4.1 m s−1 and 12.1°, respectively. It should be noted that these differences are mainly caused by the difficulty in performing comparisons due to inconsistencies in the samplings by the radar and the profiler, and by errors in the wind retrieval procedure such as errors in the interpolation into the grid points, in the upper and lower boundary conditions, and in the estimations of the motion and evolution of the precipitating system during the volume scans. The accuracy of the radar-retrieved winds is also affected by the geometry of the radars with respect to the grid points and the accuracy of the measurements of the individual radars.

b. Case II: Hook echo

The performance of the radar network for high-resolution wind retrievals was further examined with a well-developed hook echo with a diameter of about 30 km. Figure 9 shows the synthesized radar reflectivity and horizontal wind vectors at heights of 2 and 6 km at 1440 UTC 30 June 2005 when a hook echo passed near Chuncheon (37.9°N, 127.7°E) in Kangwon Province. At this time, the surface meteorological observatories around Chuncheon reported strong tornadic winds, hail, and lightning. The wind fields shown in Fig. 9 were retrieved with a horizontal resolution of 1 km × 1 km and a vertical resolution of 0.5 km.

At a height of 2 km (Fig. 9a), a hook echo with reflectivities over 50 dBZ is well recognized near 37.75°N, 127.65°E, surrounding a distinct WER. At a height of 6 km (Fig. 9b), a WER or hook echo does not occur, while the radar reflectivity pattern is circular with very high values over 60 dBZ, indicating the existence of hail. The WER could be recognized up to a height of approximately 3.5 km. This three-dimensional structure of the radar reflectivity is similar to that of hook echoes within well-developed supercell thunderstorms in Lemon and Doswell (1979) and Markowski (2002). A WER is an indicator implying the existence of strong updrafts, cyclonic rotation, and hail aloft, which consequently describe the storm severity. In addition, WER is frequently indicative of tornadogenesis (Rotunno 1986). Therefore, it can be expected that the radar-retrieved winds present a cyclonically rotating updraft (mesocyclone) around the hook echo.

The radar-retrieved winds at a height of 2 km (Fig. 9a) reveal that the surroundings of the WER are characterized by a cyclonically rotating vortex and an inflow from the south to the WER. The mean wind speed around the WER is about 9 m s−1. At a height of 6 km (Fig. 9b), where the hook echo does not occur, the cyclonic circulation is weak.

Figure 10 shows the vertical structure penetrating the WER along the A–B line shown in Fig. 9, which is parallel to the direction of the storm movement (290° and 15 m s−1 in the same sense as the definition of the wind). The reflectivity pattern (Fig. 10a) reveals a maximum value of 64 dBZ at a height of 6 km and WER at low heights. Further, it is shown that an overshooting of the echo occurs at a height of 12 km over the storm core, and a plume with weak reflectivity of about 20 dBZ occurs aloft ahead of the storm. The reflectivity pattern is tilted in a direction ahead of the storm. The wind vectors along the cross-sectional line reveal convergence at low heights and upward motion at the storm core. The pattern of the vertical air velocity (Fig. 10b) reveals a strong upward motion above the WER with a maximum of +13 m s−1 at a height of 6 km. Aloft ahead and below the rear end of the storm’s core, weak downward motions (about −3 m s−1) are shown. Corresponding to these reflectivity and vertical velocity patterns, the horizontal divergence (Fig. 10c) and vertical vorticity (Fig. 10d) fields show a coupled pattern according to which convergence at the low heights of the storm core is correlated with the divergence at the echo top. The divergence and vorticity at a height of 3 km above the WER are −2.7 × 10−3 and +4.2 × 10−3 s−1, respectively. At a height of 11 km near the echo top, the divergence and vorticity are +4.5 × 10−3and −1.5 × 10−3 s−1, respectively. The positive vorticity area is extended up to a height of approximately 8 km above the WER, which indicates that the strong updraft motion rotates cyclonically, which is indicative of a mesocyclone. The positive area of vorticity is also tilted in a manner similar to that of the reflectivity pattern. This radar-derived wind structure, as described above, is quite consistent with the fundamental features found in the literature around a well-developed hook echo (Lemon and Doswell 1979; Bluestein et al. 1997; Wakimoto et al. 1998).

5. Summary and conclusions

The performance of the operational radar network over South Korea for retrieving high-resolution wind fields was examined. The wind-retrieving algorithm based on the multiple-Doppler wind synthesis in the CEDRIC package was applied to two rain events, and the radar-retrieved winds were compared with those from the rawinsondes and wind profilers. According to the results, it can be concluded that the present operational Doppler radar network has the potential to provide reliable high-resolution wind fields over almost the entire southern Korean Peninsula region, even within a mesocyclone with a diameter of about 30 km as well as at the relatively large scales of typhoons. Therefore, it is expected that high-resolution wind fields obtained from the radar network play an important role as fundamental information in the studies that require high-resolution winds, such as studies on the development of hazardous mesoscale precipitation systems, the improvement of weather forecasting through data assimilation for NWP models, and nowcasting.

Although it was shown in this study that the radar network could provide high-resolution winds over a nationwide area, additional work on improving the accuracy of wind retrievals and extending this process to operational use in real-time settings should be persued. For improving the accuracy of wind retrievals shown in the present study, it is necessary to improve the accuracy of the radial velocity measurements of the KMA radars. The KMA radars are presently operated at low PRFs (about 400–600 Hz) in order to detect echoes at long distances of up to 250 km and hence the number of pulses is only 30–40. The radial velocity measurements, therefore, yield high fluctuations from gate to gate, severe dual-PRF velocity errors, and widely empty gates where no radial velocity data exist, but where the corresponding radar reflectivity data are valid. One of the simple methods for improving the accuracy is to increase the PRFs and decrease the unambiguous ranges. Second, the number of elevation steps of the KAF radars should be increased so that the finer structures of precipitating systems can be detected. The KAF radars presently scans at only six to nine elevation angles, which is a relatively small number compared with the KMA radars.

In order to improve the accuracy of the wind retrievals, it is also necessary to modify the boundary conditions in the vertical velocity calculation, not the fixed conditions. A consideration on the lower boundary using the DEM data and the determination of an echo top using the structure of the radar reflectivity are required. Another deficiency is that the present radar network can retrieve winds in very limited regions at low altitudes (below approximately 2 km) mainly due to complex topography. A method for obtaining winds at low altitudes should be developed, such as the reconstruction of the present radar scanning strategies or the extrapolation of winds at high altitudes. Further, the use of the variational approach in a multiple-Doppler wind synthesis procedure, such as those described in Bousquet and Chong (1998), Gao et al. (1999), and Crook and Sun (2002), may also improve the wind retrieval accuracy.

The present wind retrieval algorithm is performed in a semioperational mode. The algorithm is organized into a single c-shell script and it takes about 8 min to process the volume data of the individual radars and then to retrieve the wind fields in a PC-based Linux computer (a dual-CPU 2.4-GHz machine with 2 GB of RAM). Although the volume data of the individual radars are presently transferred to the KMA data center in real time via 2-Mbps network links, we are not authorized to freely access the radar data in real time. If we can access the radar measurements in real time, the present wind retrieval algorithm can be also processed in an operational mode. For real-time processing, however, it is necessary to improve the quality control procedures for removing the nonmeteorological echoes. KMA is developing an operational procedure for the quality control of the radar measurements, based on the algorithm of Steiner and Smith (2002). Employment of this procedure may be useful for the real-time processing of the wind retrievals. An operational algorithm for correcting the aliasing effect is also required, as was mentioned earlier. Another point to consider is determining the moving velocities of the individual precipitation systems occurring at the same time over the entire area of South Korea. Since it is common over this wide area that several precipitating systems move with different directions and speeds, determining and using a mean moving velocity over the entire area can degrade the accuracies of the wind retrievals.

For improving forecasting of hazardous precipitating systems, finally, it is highly recommended that the radar-retrieved winds be assimilated into NWP models in real-time settings together with a radar-data receiving mode. In particular, we focus on the assimilation of the retrieved wind components into the Weather Research and Forecasting (WRF) model developed by NCAR, rather than use of bulk or integrated products such as VAD wind profiles (Benjamin et al. 2004), Doppler radial velocity measurements by a single radar (Xiao et al. 2005), or reflectivity-derived products such as the total water mixing ratio (Xiao et al. 2007).

Acknowledgments

The authors thank three anonymous reviewers for their insightful comments and criticisms. We acknowledge Mrs. H.-W. Kim and H.-S. Kim of Korea Air Force (KAF) for providing data and information on the KAF radars, and also Ms. Y.-J. Lee for providing information on the Korea Meteorological Administration (KMA) radars. Dr. B.-H. Heo provided the KMA wind profiler data. The first author especially thanks Mr. Y. Sasaki of the Akita Prefectural University, Akita, Japan, for valuable comments on the CEDRIC package. This work was supported by the KMA Research and Development Program under Grant CATER 2006-2303 and by the Brain Korea 21 Project in 2006.

REFERENCES

  • Albers, S. C., 1995: The LAPS wind analysis. Wea. Forecasting, 10 , 342352.

  • Bargen, D. W., , and Brown R. C. , 1980: Interactive radar velocity unfolding. Proc. 19th Conf. on Radar Meteor., Miami, FL, Amer. Meteor. Soc., 278–283.

    • Search Google Scholar
    • Export Citation
  • Benjamin, S. G., and Coauthors, 2004: An hourly assimilation–forecast cycle: The RUC. Mon. Wea. Rev., 132 , 495518.

  • Bluestein, H. B., , Unruh W. P. , , Dowell D. C. , , Hutchinson T. A. , , Crawford T. M. , , Wood A. C. , , and Stein H. , 1997: Doppler radar analysis of the Northfield, Texas, tornado of 25 May 1994. Mon. Wea. Rev., 125 , 212230.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bousquet, O., , and Chong M. , 1998: A multiple-Doppler synthesis and continuity adjustment technique (MUSCAT) to recover wind components from Doppler radar measurements. J. Atmos. Oceanic Technol., 15 , 343359.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Browning, K. A., , and Wexler R. , 1968: The determination of kinematic properties of a wind fields using Doppler radar. J. Appl. Meteor., 7 , 105113.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Chang, P-L., , Wang B-S. , , Shieh S-L. , , and Jou B. J-D. , 2003: Clear-air echo statistics and beam blockage patterns of Taiwan Doppler radar network. Preprints, 31st Conf. on Radar Meteorology, Seattle, WA, Amer. Meteor. Soc., 7B.8. [Available online at http://ams.confex.com/ams/pdfpapers/64485.pdf.].

    • Search Google Scholar
    • Export Citation
  • Collier, C. G., Ed. 1992: International Weather Radar Networking. Kluwer Academic, 344 pp.

  • Crook, N. A., , and Sun J. , 2002: Assimilating radar, surface, and profiler data for the Sydney 2000 Forecast Demonstration Project. J. Atmos. Oceanic Technol., 19 , 888898.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dazhang, T., , Geotis S. G. , , Passarelli R. E. Jr., , Hansen A. L. , , and Frush C. L. , 1984: Evaluation of an alternating-PRF method for extending the range of unambiguous Doppler velocity. Preprints, 22nd Conf. on Radar Meteorology, Zurich, Switzerland, Amer. Meteor. Soc., 523–527.

    • Search Google Scholar
    • Export Citation
  • Eilts, M. D., , and Smith S. D. , 1990: Efficient dealiasing of Doppler velocities using local environment constraints. J. Atmos. Oceanic Technol., 7 , 118128.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gao, J., , Xue M. , , Shapiro A. , , and Droegemeier K. K. , 1999: A variational method for the analysis of three-dimensional wind fields from two Doppler radars. Mon. Wea. Rev., 127 , 21282142.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Holleman, I., , and Beekhuis H. , 2003: Analysis and correction of dual PRF velocity data. J. Atmos. Oceanic Technol., 20 , 443453.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • James, C. N., , and Houze R. A. Jr., 2001: A real-time four-dimensional Doppler dealiasing scheme. J. Atmos. Oceanic Technol., 18 , 16741683.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Joe, P., , and May P. T. , 2003: Correction of dual PRF velocity errors for operational Doppler weather radars. J. Atmos. Oceanic Technol., 20 , 429442.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kim, H-W., , and Lee D-K. , 2006: An observational study of mesoscale convective systems with heavy rainfall over the Korean Peninsula. Wea. Forecasting, 21 , 125148.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kim, Y-S., , Nam K-Y. , , Park O-R. , , and Cho C-H. , 2004: Retrieval of horizontal wind field from Doppler radar radial velocity using a successive correction method (in Korean). J. Kor. Meteor. Soc., 40 , 229239.

    • Search Google Scholar
    • Export Citation
  • Klazura, G. E., , and Imy D. A. , 1993: A description of the initial set of analysis products available from the NEXRAD WSR-88D system. Bull. Amer. Meteor. Soc., 74 , 12931311.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Köck, K., , Leitner T. , , Randeu W. L. , , Divjak M. , , and Schreiber K. J. , 2000: OPERA: Operational Programme for the Exchange of Weather Radar Information: First results and outlook for the future. Phys. Chem. Earth, 25B , 11471151.

    • Search Google Scholar
    • Export Citation
  • Lemon, L., , and Doswell C. A. III, 1979: Severe thunderstorm evolution and mesocyclone structure as related to tornadogenesis. Mon. Wea. Rev., 107 , 11841197.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Makihara, Y., 1996: A method for improving radar estimates of precipitation by comparing data from radars and raingauges. J. Meteor. Soc. Japan, 74 , 459480.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Markowski, P. M., 2002: Hook echoes and rear-flank downdrafts: A review. Mon. Wea. Rev., 130 , 852876.

  • May, P. T., 2001: Mesocyclone and microburst signature distortion with dual PRT radars. J. Atmos. Oceanic Technol., 18 , 12291233.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Meischner, P., , Collier C. , , Illingworth A. , , Joss J. , , and Randeu W. , 1997: Advanced weather radar systems in Europe: The COST 75 action. Bull. Amer. Meteor. Soc., 78 , 14111430.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mohr, C. G., , Miller L. J. , , Vaughan R. L. , , and Frank H. W. , 1986: The merger of mesoscale datasets into a common Cartesian format for efficient and systematic analyses. J. Atmos. Oceanic Technol., 3 , 143161.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nam, K-Y., , Kim Y-H. , , Kim K-E. , , and Nam J-C. , 2005: Study on the multi-cell storm structure using dual Doppler radars observations (in Korean). J. Kor. Meteor. Soc., 41 , 967981.

    • Search Google Scholar
    • Export Citation
  • Ray, P. S., , Wagner K. K. , , Johnson K. W. , , Stephens J. J. , , Bumgarner W. C. , , and Mueller E. A. , 1978: Triple-Doppler observations of a convective storm. J. Appl. Meteor., 17 , 12011212.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ray, P. S., , Ziegler C. L. , , and Bumgarner W. , 1980: Single- and multiple-Doppler radar observations of tornadic storms. Mon. Wea. Rev., 108 , 16071625.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rosenfeld, D., , Wolff B. D. , , and Atlas D. , 1993: General probability-matched relations between radar reflectivity and rain rate. J. Appl. Meteor., 32 , 5072.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Rossa, A. M., 2000: COST-717: Use of radar observations in hydrological and NWP models. Phys. Chem. Earth, 25B , 12211224.

  • Rotunno, R., 1986: Tornadoes and tornadogenesis. Mesoscale Meteorology and Forecasting, P. S. Ray, Ed., Amer. Meteor. Soc., 414–436.

  • Steiner, M., , and Smith J. A. , 2002: Use of three-dimensional reflectivity structure for automated detection and removal of nonprecipitating echoes in radar data. J. Atmos. Oceanic Technol., 19 , 673686.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Suk, M-K., , Nam K-Y. , , Kim Y-H. , , and Oh S-N. , 2005: Estimation of quantitative rain intensity from radar reflectivities using a window probability matching method (in Korean). J. Kor. Meteor. Soc., 41 , 123138.

    • Search Google Scholar
    • Export Citation
  • Wakimoto, R. M., , Liu C. , , and Cai H. , 1998: The Garden City, Kansas, storm during VORTEX 95. Part I: Overview of the storm’s life cycle and mesocyclogenesis. Mon. Wea. Rev., 126 , 372392.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xiao, Q., , Kuo Y-H. , , Sun J. , , Lee W-C. , , Lim E. , , Guo Y-R. , , and Barker D. M. , 2005: Assimilation of Doppler radar observations with a regional 3DVAR system: Impact of Doppler velocities on forecasts of a heavy rainfall case. J. Appl. Meteor., 44 , 768788.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xiao, Q., , Kuo Y-H. , , Sun J. , , Lee W-C. , , Barker D. M. , , and Lim E. , 2007: An approach of radar reflectivity data assimilation and its assessment with the inland QPF of Typhoon Rusa (2002) at landfall. J. Appl. Meteor. Climatol., 46 , 1422.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Topography and locations of the operational Doppler radars over the southern Korean Peninsula. The plus sign, filled circle, and open triangle denote the radars operated by USAF, KMA, and KAF, respectively, and the boldface and italic letters denote 10- and 5.5-cm wavelength radars, respectively. The open circles with a cross and open circles denote the four UHF wind profiler and seven rawinsonde sites, respectively.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 2.
Fig. 2.

Wind-retrievable area (gray shaded area) at heights of (a) 2, (b) 4, (c) 7, and (d) 13 km. The plus sign, filled circle, and open triangle denote the radars used as defined in Fig. 1.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 3.
Fig. 3.

PPI images of the radial velocity of the RMYN radar at an elevation angle of 1.5° at 0305:45 UTC 10 Jul 2006: (a) the observed velocity and (b) the corrected velocity.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 4.
Fig. 4.

PPI images of the (a) observed radial velocity and (b) filtered radial velocity of the RKWK radar at an elevation angle of 3.1° at 0303:12 UTC 10 Jul 2006.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 5.
Fig. 5.

Composite images of the synthesized radar reflectivity and horizontal wind vectors retrieved from the radar network at (a) 1800 UTC 9 Jul, (b) 0000 UTC 10 Jul, (c) 0600 UTC 10 Jul, and (d) 0900 UTC 10 Jul 2006. The open circles with a cross denote the typhoon center. The half and full wind barbs denote 2.5 and 5 m s−1, respectively. The pennant denotes 25 m s−1.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 6.
Fig. 6.

Comparisons of horizontal wind vectors retrieved from the radars and observed by the rawinsondes: (a) 1800 UTC 9, (b) 0000 UTC 10, and (c) 0600 UTC 10 Jul 2006.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 7.
Fig. 7.

Comparisons of the horizontal wind vectors retrieved from the radars and measured by the wind profilers during 1800 UTC 9 Jul–1100 UTC 10 Jul 2006 with an interval of 1 h at the (a) WPMUS, (b) WPKNG, (c) WPKSN, and (d) WPMAS wind profilers.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 8.
Fig. 8.

Scatterplots of the horizontal winds obtained from radars and wind profilers: (a) u and (b) υ components.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 9.
Fig. 9.

Composite images of radar reflectivity and horizontal wind vectors at heights of (a) 2 and (b) 6 km at 1440 UTC 30 Jun 2005. The half and full wind barbs denote 2.5 and 5 m s−1, respectively. The pennant denotes 25 m s−1.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Fig. 10.
Fig. 10.

Vertical cross sections of the (a) radar reflectivity and cross-sectional wind vector, (b) vertical air velocity, (c) horizontal divergence, and (d) vertical vorticity along the A–B line in Fig. 9. The overlapping contours denote the radar reflectivity.

Citation: Weather and Forecasting 24, 1; 10.1175/2008WAF2007084.1

Table 1.

Operational scan modes of the radars over the Korean Peninsula for the lowest elevation level; Rmax and Vmax denote the unambiguous range and velocity, respectively.

Table 1.
Save