Reliability of X-band Dual-polarization Phased Array Radars Through Comparison with an S-band Dual-polarization Doppler Radar

Due to China's vast and complex territory, many types of meteorological disasters take place frequently within its borders, including mesoscale or microscale severe convective weather systems such as tornadoes, hailstorms, and downbursts. At present, it is difficult to accurately forecast and provide warnings regarding mesoscale weather systems, the main reason for which is that they usually take place locally and evolve rapidly, and commonly employed detection methods lack the ability to observe their fine-scale structures and rapid evolution. As a result, current understanding with respect to the development and evolution mechanisms of mesoscale weather systems is limited (Wilson et al. ^[1]). Weather radar is currently the most important means of observing mesoscale weather systems. Weather radars used for observation and research, such as the WSR-88D weather radars in the United States, and the new-generation Doppler weather radars in China, mostly use mechanical scanning methods. In daily operations of the new-generation Doppler weather radars in China, the VCP21 mode is commonly used, in which nine elevation angles are scanned in six minutes, with a radial resolution of 1 km. This kind of radar can meet the requirements of weather detection, forecasting, and near-warning of large-scale weather systems such as typhoons and rainstorms (Simmons and Sutter ^[2]; Bai et al. ^[3-4]; Meng et al. ^[5]; Zhou et al. ^[6]; Luo et al. ^[7]). However, its long volume scan time and low spatial resolution make it difficult to observe mesoscale weather systems, especially to obtain detailed and complete information of the development and evolution of meso-γ-scale and microscale weather systems (Liu et al. ^[8]; Zhao et al. ^[9]; Wu et al. ^[10]). Hence, further in-depth studies and improved understanding of mesoscale weather systems are hindered. Therefore, it is essential to improve the ability to monitor and obtain observational data with higher temporal and spatial resolution for more in-depth and detailed research on the development and evolution mechanisms of mesoscale weather systems.

Phased array weather radars have the ability to quickly and accurately steer the beam, so that the radar can obtain observational data with high temporal resolution. In 2000, the United States began to build a one-dimensional X-band phased array weather radar (MWR-05XP), which combines vertical electronic scanning with horizontal mechanical scanning. The data quality of MWR-05XP is equivalent to that of the WSR-88D weather radar, but its scanning speed is much higher (Wurman and Randall ^[11]; Bluestein et al. ^[12]; Weber et al. ^[13]). In 2002, the National Research Committee on Weather Radar Technology recommended phased array technology as the development direction of next-generation weather radars in the United States (Weber et al. ^[14]). In 2006, a decommissioned AN/SPY-1A two-dimensional phased array radar was repurposed to establish a phased array weather radar experimental platform to observe severe weather such as hail and tornadoes (Weadon et al. ^[15]). In addition, the implementation of dual-polarization technology on phased array antennas was also discussed in the United States, the feasibility of Multifunction Phased Array Radar was studied, and its performance specifications were formulated (Zhang et al. ^[16]). At the same time, the Collaborative Adaptive Sensing of the Atmosphere program in the United States also adopted phased array radars to improve low-altitude detection capability. It plans to complete the replacement of the current mechanical scanning Doppler weather radars with phased array radars by 2025 (Heberling et al. ^[17]). Japan began research and development of its next-generation weather radar in 2000, and developed its first single-polarization X-band one-dimensional phased array weather radar (PAWR) in 2012 (Wu et al. ^[18]). In 2016, observation experiments with four networked radars were carried out, and the results indicated that the fast scanning and wide elevation coverage capabilities of PAWR provide earlier detection and a higher detection probability, which will enable earlier and more accurate warnings of severe weather (Mizutani et al. ^[19]; Kurdzo et al. ^[20]).

Taking advantage of the high temporal and spatial resolution of the X-band phased array radar, many scholars have conducted research on mesoscale and microscale severe convective weather systems, such as tornadoes, hailstorms, and downbursts, promoting understanding of the relationship between convective weather systems and disasters near the ground (Adachi et al. ^[21-22]; Kikuchi et al. ^[23]; Wurman et al. ^[24]). Newman et al. ^[25] used a mobile X-band phased array weather radar, a WSR-88D radar, and a TDWR-OKC narrow-beam radar to simultaneously observe the evolution of a tornado in a convective system. It was concluded that the phased array weather radar more clearly captured the evolution of the tornado vortex signature (TVS) and downburst, as well as the evolution of mesoscale cyclones, mid-level convergence, and outflow. Moreover, its deployment could lead to better understanding of the formation process of tornadoes. Meanwhile, analysis of the tornadoes generated by supercells has shown that X-band phased array radar can clearly observe the disappearance, rebirth, and merging of the TVS (French et al. ^[26]), which is useful for further studying and understanding storm processes, including rapid organization, sudden enhancement of mesocyclones and inflow, and the position and movement of the tornado vortex (Kuster et al. ^[27]; Wilson et al. ^[28]). Observational results of the two-dimensional phased array weather radar show that, compared with WSR-88D, it can better and more accurately detect rapidly changing weather systems (Zrnic et al. ^[29]; Heinselman et al. ^[30]; Supinie et al. ^[31]) such as low-altitude convergences within supercells, upper-level rotations in mesocyclones, rapid enhancement of hailstorms and the formation and evolution of the outflow, and the life history and largest outflow values of downbursts. The detected data, with their high temporal and spatial resolution, also help to improve the quality of pattern analysis and recognition. Yussouf and Stensrud ^[32] compared assimilated 15-min observations of a conventional WSR-88D radar with the 6-min volume scan time and a phased array weather radar with a 1-min volume scan time, and made a 50-min weather forecast. The phased array weather radar showed significantly better performance over the conventional Doppler radar in describing and forecasting the processes of supercell evolution.

Research and development with respect to phased array weather radar technology in China is happening relatively later than in other parts of the world, but considerable progress has nevertheless already been made. In 2007, in cooperation with several other research institutes, the China Meteorological Research Institute developed an S-band phased array weather radar, and the feasibility of using phased array radar for weather observation was proved through various observation experiments. Subsequently, Anhui Sun-Create Electronics Co., Ltd. developed an X-band phased array Doppler weather radar, which improved the observation mode and accuracy (Wu et al. ^[33]). In 2019, the Meteorological Observation Center of the China Meteorological Administration used three X-band phased array transceiver sub-arrays to form a networked weather radar, and carried out observation experiments at Changsha Airport. The results showed that more detailed and complete features of small-scale weather systems can be obtained (Ma et al. ^[34]). In 2015, Zhuhai Naruida Technology Ltd. developed an X-band dual-polarization phased array weather radar (DP-PAWR), which further realized the dual-polarization function based on the X-band phased array weather radar. Combining phased array radar with dual-polarization detection technology, DP-PAWR not only can observe the precipitation and dynamic structures inside convective storms, but also can obtain the microphysical structures.

At present, five Naruida DP-PAWRs have been deployed in the city of Guangzhou (radar sites: Panyu, Nanhai, Huadu, Maofengshan and Nansha) and a collaborative radar observation network has been formed with these DP-PAWRs. The DP-PAWRs have been subjected to quality control such as ground clutter suppression, noise elimination, and attenuation correction after their deployment (Cheng et al. ^[35]). In order to test the reliability of radar data after quality control and their ability to detect fine-scale structures of mesoscale weather systems, this paper analyzes the observational results of a squall line case on 11 May 2020 and a stratiform precipitation case on 6 June 2020 using two Naruida DP-PAWRs and an S-band dualpolarization Doppler weather radar (CINRAD/SA-D). After location matching, the observation results of the two kinds of radars are compared, including reflectivity (Z_H), radial velocity (V), differential reflectivity (Z_DR), and specific differential phase (K_DP). The advantages of high temporal and spatial resolution data in studying the development and evolution of mesoscale systems are also preliminarily analyzed.

On the basis of a squall line case that occurred on 11 May 2020 and a stratiform precipitation case that occurred on 6 June 2020, this paper compares the simultaneous observational data of Panyu DP-PAWR [64 m above mean sea level (MSL)] and Guangzhou CINRAD/SA-D (179 m MSL), which is located 3.6 km southeast of Panyu DP-PAWR, to verify the reliability of DP-PAWR data. Previous studies have suggested that CINRAD/SA-D data are reliable (Hu et al. ^[36]).

During the movement of the squall line on 11 May 2020, Nanhai DP-PAWR (38 m MSL) observed the triggering and development of a cell in front of the squall line and the merging of this cell with the squall line. The observational data were preliminarily analyzed to evaluate the role of data with high temporal and spatial resolution in studying the development and evolution of mesoscale weather systems.

The DP-PAWRs use a one-dimensional phase scan mode (electronic scanning in the direction of elevation and mechanical scanning in the azimuthal direction) and can complete a volume scan (12 continuous elevation angles, 0.9° - 20.7° in 1.8° steps; 400 azimuthal angles, 0°-360°, in 0.9° steps) within 90 seconds. CINRAD/SA-D uses the VCP21 scanning mode, which can complete a volume scan (nine non-continuous elevation angles, 0.5°-19.5°) in six minutes. The main operating parameters of the two kinds of radars are shown in Table 1.

Due to the obvious differences in the geographic locations and scanning methods of the Panyu DP-PAWR and Guangzhou CINRAD/SA-D, the observational data cannot be directly compared. Therefore, it was necessary to match the radar data in a certain way before performing comparative analysis. Using the method proposed by Zhang et al.^[37] and Wu et al.^[38], we first converted the DP-PAWR observational data from the polar coordinate system (azimuthal angle, elevation angle, and slant distance) to the geodetic coordinate system (latitude, longitude, and altitude), and then used the location information of the CINRAD/SA-D site to represent the DP-PAWR observational data using the CINRAD/SA-D polar coordinate system. This enabled the converted DP-PAWR data and interpolated CINRAD/SA-D data to be at the same location so they could be directly compared.

Assuming the coordinates of the DP-PAWR site are (α_x, β_x, h_x), where α_x is the latitude, β_x is the longitude, and h x is the height, the azimuth, elevation and slant distance of any range bin in its polar coordinate system are (a_x, e_x, r_x), and the latitude, longitude and altitude of the range bin in the geodetic coordinate system are recorded as (α_g, β_g, h_g). Then, (α_g, β_g, h_g) can be calculated using the spherical triangle formula:

Here, R_e is Earth's radius, R_m is Earth's equivalent radius (and $R_{m}=\frac{4}{3} R_{e} $), and s is the distance between two points on the Earth's surface. Then, combined with the location of the CINRAD/SA-D site (α_s, β_s, h_s), the grid point (α_g, β_g, h_g) can be represented using the CINRAD/SA-D polar coordinate system, and the corresponding azimuthal angle, elevation angle and slant distance can be written as a_s, e_s and r_s, respectively. Then, the DP-PAWR data are matched with the CINRAD/SA-D data as follows:

where

in which the azimuthal angle a_s is

the elevation angle e_s is

and the slant distance r_s is

In this way, each location (a_x, e_x, r_x) of the DP-PAWR range bin can be mapped to the corresponding location (a_s, e_s, r_s) in the CINRAD/SA-D polar coordinate system. Then, the CINRAD/SA-D observational data are interpolated to calculate the values at the same position location (a_s, e_s, r_s), and then the data of the two radars are matched and can be compared.

For the interpolation, in order to retain the echo structure features in the original volume scan data, the nearest neighbors in the radial and azimuthal directions and vertical linear interpolation methods were used. Interpolation data were applied to location (a_s, e_s, r_s), where a_s and r_s use the nearest-neighbor azimuthal angle and slope distance. When the es angle was within the detection range of CINRAD/SA-D, the two adjacent elevations e_s1 and e_s2 were used for interpolation. (a_s, e_s1, r_s) and (a_s, e_s2, r_s) are the intersection points of the vertical line passing through the points (a_s, e_s, r_s) and the CINRAD/SA-D up and down elevation beam axis, and the data value f^a(a_s, e_s, r_s) at point (a_s, e_s, r_s) can be obtained using vertical linear interpolation of data values of f^a(a_s, e_s1, r_s) and f^a(a_s, e_s2, r_s):

Here, ω_e1 and ω_e2 are interpolation weights for f^a(a_s, e_s1, r_s) and f^a(a_s, e_s2, r_s):

In this way, for the observational data of any range bin of DP-PAWR, the observational data of CINRAD/SA-D matching its location could be found, enabling direct comparison and analysis of the observational data from the two radars. It should be noted that the V data of the two radars could not be directly compared via this matching method. They could only be compared with the data on the contour line 1-2 km above the two radars (Liu et al.^[39]).

At 20: 36 Beijing Time (BT, UTC + 8, the same below) 11 May 2020, Panyu DP-PAWR and Guangzhou CINRAD/SA-D were behind the squall line. The observational data of the two radars were processed using the matching method formulated in Eqs. (1)-(12), and then the PPI, RHI structures, and the data changes with distance and azimuthal angle were compared and analyzed.

Figure 1 compares the PPI images of Z_H at various elevation angles. It can be seen that the PPI images of Z_H observed by the two radars show consistent structural features, except for the images below the elevation angle of 2.7°, which are affected by the obstruction of groundlevel objects, and the images above the elevation angle of 15.3° where the overlapping area of the detection range of the two radars is decreased. However, the PPI images of Z_H obtained by DP-PAWR are finer. Taking the PPI images of Z_H at the elevation angle of 4.5° as an example, it can be seen that the convective precipitation echoes observed by the two radars are mainly located in the area where the azimuthal angle is between 60° and 240° and a range of 20-42 km, and where the intensity is between 35 and 65 dBZ. The other areas are dominated by stratiform rain echoes with intensity below 35 dBZ, and the characteristics of the radar echo structures are basically consistent. However, due to the attenuation of strong echoes, the DP-PAWR data are missing in two areas: one with 170° - 225° azimuthal angles and range of 30-42 km, and the other with 90° - 120° azimuthal angles and range of 40-42 km. Therefore, it is necessary to network DP-PAWRs to obtain more detailed information of severe convective storms. In addition, CINRAD/SA-D was able to observe the weak echoes below 10 dBZ in the area with a 315° azimuthal angle and range beyond 30 km. However, it was difficult for DP-PAWR to observe weak echoes at long distances owing to its low transmission power and low sensitivity. This is more obvious in the high elevation angles above 11.7°. Fig. 2a shows the variation in minimum Z_H with range at the current time-namely, the minimum value of the Z_H data of the two radars that was not interpolated for each elevation angle in 1-km steps. It can be seen that the minimum measurable Z_H of the DP-PAWR increases as the distance increases, which is as high as 14.34 dBZ at 30 km. However, the minimum measurable Z_H of CINRAD/SA-D is much smaller than that of DP-PAWR, which is - 8.7 dBZ at 30 km. The difference between the minimum measurable Z_H of the two radars is 23 dBZ, which will have a substantial impact on the echo structures of DP-PAWR.

Figure 1.  PPI images of Z_H observed by (a) DP-PAWR and (b) CINRAD/SA-D at 20:36 BT 11 May 2020. The subscript numbers denote different elevation angles as follows: (1) 0.9°; (2) 2.7°; (3) 4.5°; (4) 6.3°; (5) 8.1°; (6) 9.9°; (7) 11.7°; (8) 13.5°; (9) 15.3°; (10) 17.1°; (11) 18.9°; (12) 20.7°. The circles are in intervals of 14 km (the same in subsequent figures below).

Figure 2.  (a) Variation in minimum Z_H with range at 20:36 BT 11 May 2020 (red line, DP-PAWR data; black line, CINRAD/SA-D data). (b, c) RHI images of Z_H at the azimuthal angle of 110°: (b) DP-PAWR; (c) CINRAD/SA-D.

Figures 2b and 2c show the RHI images of Z_H of the two radars along the azimuthal angle of 110°. It can be seen that the echo structures observed by the two radars are also relatively consistent. Within a 10-30 km range, echoes with intensity greater than 35 dBZ are mostly below the height of 5 km, and their intensity gradually increases with range. At the range beyond 30 km, the echo with intensity exceeding 35 dBZ rapidly expands to the height above 10 km, and the strong echo center is located in the area with a range of 40 km and height near 3 km, with the intensity exceeding 60 dBZ. Compared with CINRAD/SA-D, DP-PAWR can obtain finer structural features of the echoes. It also has good coverage for the low-altitude blind area of CINRAD/SA-D, and hence the starting height of the 35 dBZ echo can be clearly observed. This is favorable for observing the key areas of triggering initiation at mid and low altitudes in mesoscale weather systems. In addition, since the highest elevation angle of DP-PAWR reaches 20.7°, compared with CINRAD/SA-D, its maximum detection altitude at the same distance is higher.

In order to further analyze the differences in the detection capabilities of the two radars with range, Fig. 3 shows the radial distribution of the elevation angle Z_H of 4.5° to 13.5°. In order to reduce the data difference caused by different horizontal beam widths, the azimuthal angle ranges from 90° to 99°. DP-PAWR uses pulse compression technology, which will cause blind areas at close distances. Although a combination of long and short pulses has been used to fill the blind areas, there will still be certain detection errors (Zhang et al. ^[40]). In addition, ground clutter will also affect the quality of the two radars'close-range observational data. Therefore, this paper mainly compares and analyzes the data quality of the two radars beyond 5 km. It can be seen from Fig. 3 that the changes in Z_H with range observed by the two radars are relatively consistent. When the range exceeds 30 km, the Z_H values of CINRAD/SA-D gradually become smaller, especially in the mid - and low-level convection development heights. This might be caused by the mechanical scanning and time limitation of CINRAD/SA-D, the vertical beam spacing increase at long distances, the interpolation smoothing effect, or the attenuation correction error of DP-PAWR.

Figure 3.  Variation in averaged Z_H with range at 20:36 BT 11 May 2020 (azimuthal angle: 90°-99°; red line, DP-PAWR; black line, CINRAD/SA-D) at elevation angles of (a) 4.5°, (b) 6.3°, (c) 8.1°, (d) 9.9°, (e) 11.7°, and (f) 13.5°.

In order to analyze the pointing accuracy of DP-PAWR and its detection capability along the azimuthal angle, Fig. 4a shows the distribution of Z_H of the two radars at the elevation angle of 4.5° and along azimuthal angles of 90°-270°. It can be seen that the variations in averaged Z_H with the azimuthal angles of the two radars are very consistent, which shows that the positioning of DP-PAWR is relatively accurate. Fig. 4b shows the probability distribution of Z_H of the two radars at the elevation angle of 4.5°. It can be seen that the trend fluctuations of the probability distribution profiles of the two radars are basically the same, but that the DP-PAWR data are slightly smaller than those of CINRAD/SA-D by about 1 dBZ.

Figure 4.  (a) Variation in averaged Z_H with azimuth (range: 14-28 km) and (b) probability distribution of Z_H at the elevation angle of 4.5° at 20:36 BT 11 May 2020 (red line, DP-PAWR; black line, CINRAD/SA-D).

Since the two radars are not at the same location, the PPI image of V could not be directly compared, and the comparison could only be made on the contour line of 1-2 km above the connecting line between the two radars. Fig. 5 shows the V at a height of 2 km above the line between the two radars with the location of DP-PAWR as the origin. The CINRAD/SA-D V data have been inverted. As can be seen, the V values and trends of change are relatively consistent. The V away from DP-PAWR is 12-26 m s^-1, while that towards DP-PAWR is 11-13 m s^-1.

Figure 5.  Comparison of V at 2 km above the line between DP-PAWR (red) and CINRAD/SA-D (black) at 20:36 BT 11 May 2020.

Z_DR is the ratio of the horizontal reflectivity to the vertical reflectivity within the beam volume, and the unit is dB. Similar to Z_H, the Z_DR positions and structures of the two radars after location matching also show relatively consistent characteristics in the PPI images of each elevation angle. Taking the elevation angle of 4.5° as an example (Figs. 6a and 6b), it can be seen that the echo positions of the two radars exceeding 0.5 dB are basically the same, and Z_DR correlates closely with Z_H, which increases as Z_H increases (Fig. 1a₃). However, the DP-PAWR data are slightly smaller than those of CINRAD/SA-D. In addition, there are two obvious low-value areas near the azimuthal angles of 40° and 200°, which are caused by the lightning rod installed near the Panyu DP-PAWR site. From Figs. 6c and 6d, it can be seen that the area with Z_DR values greater than 0.5 dB is mainly below the height of 5 km, and the Z_DR values above 5 km are lower. Combined with Z_H, it can be preliminarily concluded that the height of the melting layer is 5 km, and the precipitation particles below the melting layer are mainly liquid, while above the melting layer ice particles are dominant. The weather balloon data of Qingyuan weather station (55 km away from the Panyu DP-PAWR site) showed that the height of the melting layer was 5.14 km at 20:00 11 May 2020, which is consistent with the radar observation results.

Figure 6.  (a, b) PPI images of Z_DR at the elevation angle of 4.5° and (c, d) RHI images of Z_DR at the azimuthal angle of 110° obtained by (a, c) DP-PAWR and (b, d) CINRAD/SA-D at 20:36 BT 11 May 2020.

Figure 7a shows the radial distriution of Z_DR at the elevation angle of 4.5° and azimuthal angles from 90° to 99°. It can be seen that the two radars'trends of change are generally consistent. It can also be seen from Fig. 7b that the Z_DR values of the two radars have good consistency with azimuthal angle. The Z_DR probability distribution curves of the two radars are close to a normal distribution (Fig. 7c), but the distribution of DP-PAWR is narrow and concentrated, with a peak value at -0.2 dB, accounting for 17.34%. At the same time, the data below - 2 dB are relatively insufficient, which is 5.96% lower compared to CINRAD/SA-D. The peak value of CINRAD/SA-D is 0.2 dB, accounting for 11.04%, and the mean difference between the two radars is 0.07 dB.

Figure 7.  Variation in averaged Z_DR with (a) range (azimuthal angle: 90°-99°) and (b) azimuth (distance: 14-28 km), and (c) the probability distribution of Z_DR at the elevation angle of 4.5° at 20:36 BT 11 May 2020 (red line, DP-PAWR; black line, CINRAD/SA-D).

K_DP represents the rate of change in the phase difference between horizontally and vertically polarized pulses with range when electromagnetic waves pass through precipitation particles. The unit of K_DP is ° km^-1 and its value is related to the wavelength of electromagnetic waves. Theoretically, the K_DP of liquid precipitation is generally positive. From Figs. 8a and 1a₃, it can be seen that when Z_H is less than 35 dBZ, the K_DP of DP-PAWR is less than 2° km^-1. When Z_H increases to more than 35 dBZ, the K_DP begins to increase as Z_H increases. Comparison between Figs. 8b and 1a₄ shows that the K_DP of CINRAD/SA-D has lower sensitivity than that of DP-PAWR; it begins to increase until Z_H increases to more than 45 dBZ and the intensity is lower than that of DP-PAWR.

Figure 8.  PPI images of the K_DP obtained by (a) DP-PAWR and (b) CINRAD/SA-D at an elevation angle of 4.5° at 20:36 BT 11 May 2020.

Figure 9 shows scatterplots of K_DP and Z_H obtained by DP-PAWR and CINRAD/SA-D. As can be seen, the mean K_DP obtained by the two radars increases as the Z_H increases. However, there are also obvious differences between them. The K_DP of DP-PAWR is generally greater than 0. When Z_H is less than 35 dBZ, the K_DP of DP-PAWR is mostly at 0-2° km^-1, and the average value is almost 0 with no significant increase. When Z_H increases to more than 35 dBZ, the K_DP begins to increase as the Z_H increases, and the dispersion gradually increases. It is also noticeable that when Z_H exceeds 50 dBZ, the average value of K_DP first falls and then rises. This is because a window smoothing algorithm is adopted when calculating K_DP, which will have a certain impact on the data quality when there are data missing. For CINRAD/SA-D, when the Z_H is less than 45 dBZ, the K_DP is mostly within - 1 to 2 ° km^-1, but the data dispersion is larger and the minimum value is close to -4° km^-1. When the Z_H exceeds 45 dBZ, the K_DP begins to gradually increase, but the increase is significantly smaller than that of DP-PAWR. Based on the above analysis, the K_DP of DP-PAWR has higher sensitivity than that of CINRAD/SA-D, meaning it provides a better indication of precipitation.

Figure 9.  Scatterplots of the Z_H and K_DP of (a) DP-PAWR and (b) CINRAD/SA-D at an elevation angle of 4.5° at 20:36 BT 11 May 2020. The red curve is the mean K_DP curve obtained by dividing Z_H into 1-dB intervals.

Figure 10 compares the PPI images of Z_H, Z_DR and K_DP at the elevation angle of 4.5° at 18: 00 BT 6 June 2020. It can be seen that the positions and structures of Z_H observed by the two radars are relatively consistent, the convective echoes are mainly located in the area where the azimuthal angle is between 350° and 360° and the range is 30-42 km, and the intensity is between 35 and 55 dBZ. The K_DP observed by the two radars reflects the strong echo areas above, while similar to the previous case, the value observed by DP-PAWR is greater than that of CINRAD/SA-D. The areas with ZDR values greater than 0.5 dB observed by the two radars are similar, but the details are different.

Figure 10.  PPI images of (a, b) Z_H, (c, d) Z_DR and (e, f) K_DP, at an elevation angle of 4.5°, observed by (a, c, e) DP-PAWR and (b, d, f) CINRAD/SA-D at 18:00 BT 06 June 2020.

It can be seen from Fig. 11a that the trend fluctuations of the Z_H probability distribution profiles of the two radars are basically the same, but the DP-PAWR data are smaller than those of CINRAD/SA-D by about 2 dBZ. Fig. 11b shows that the V values and trends of change of the two radars are relatively consistent. The V away from DP-PAWR is 5-10 m s^-1, while that towards DP-PAWR is 8-12 m s^-1. The Z_DR probability distribution curves of the two radars are close to a normal distribution (Fig. 11c), the DP-PAWR data are slightly smaller than those of CINRAD/SA-D, and the mean difference between the two radars is 0.01 dB. From Figs. 11d and 11e, it is noticeable that when Z_H is less than 35 dBZ, the K_DP of DP-PAWR is mostly within 0-2° km^-1, with no significant increase. When Z_H increases to more than 35 dBZ, the K_DP begins to increase as the Z_H increases, and the dispersion gradually increases. By contrast, the K_DP of CINRAD/SA-D begins to increase when Z_H increases to more than 45 dBZ, and the intensity is lower than that of DP-PAWR.

Figure 11.  Probability distribution of (a) Z_H and (c) Z_DR at an elevation angle of 4.5°, and (b) a comparison of V at 2 km above the line between the two radars (red line, DP-PAWR data; black line, CINRAD/SA-D data). (d, e) Scatterplots at of Z_H and K_DP observed by (d) DP-PAWR and (e) CINRAD/SA-D at an elevation angle of 4.5° (the red curve is the mean K_DP curve obtained by dividing Z_H into 1-dB regions) at 18:00 BT 6 June 2020.

Due to the influence of ground clutter, the correlation coefficient (ρ_hv) obtained by Guangzhou CINRAD/SA-D within 50 km from the radar site was poor (not shown), so the ρ_hv of the two radars were not compared in this study.

As the squall line moved southeast on 11 May 2020, several convective cells in front of it were triggered, and then developed before merging with it. The development and evolution of one of these convective cells is taken as an example here to discuss the application of the high temporal and spatial resolution data of DP-PAWR in the analysis of mesoscale convective systems.

Figure 12 shows the composite reflectivity (CR) obtained by Nanhai DP-PAWR from 18:25 to 18:47 11 May 2020. As can be seen, during the movement of the squall line, Nanhai DP-PAWR observed the triggering, development and merging of meso-γ-scale convective cells with the main body of the squall line at multiple locations (point A with an azimuthal angle of 310° and range of 35 km; point B with an azimuthal angle of 332° and range of 31 km; and point C with an azimuthal angle of 342° and range of 20 km). Using methods described in the research of Gauthier et al. ^[41] and Yi et al. ^[42], taking the initial enclosing of the 35-dBZ CR outer boundary as the start point, and merging of the strong echo core as the end point, the process of triggering, development and merging of the cells at points A, B and C lasted for 10 minutes (18:25-18:35), 10 minutes (18: 28-18: 38) and 15 minutes (18: 32-18: 47) respectively, within which the merging process only took 6 minutes, 4 minutes and 7 minutes, respectively. We take the development and evolution of the convective cell at point C as an example to analyze in detail the echo intensity, V, and minute-level data evolution characteristics of various dual-polarization parameters of Nanhai DP-PAWR, and discuss the application of high temporal and spatial resolution data in observing and analyzing mesoscale convective weather systems.

Figure 12.  Composite reflectivity sequence obtained by Nanhai DP-PAWR between 18:25 BT and 18:47 BT 11 May 2020 (temporal interval: 90 s).

Figure 12 shows that the cell at point C gradually weakened and separated into two cells while moving eastward. From 18:32, the cell at the back strengthened significantly, and the strong-echo area kept increasing. At 18: 35, a new cell was triggered at the rear, which then quickly merged with the original cell, and the echo intensity of the newly merged cell significantly increased. As the main body of the squall line continued to move south, an echo bridge appeared between the convective cell and the squall line at 18:40, connecting the two, and the merging began. Thereafter, the merging process continued to develop, and the strength of the original cell gradually increased to more than 65 dBZ. At 18:47, the center of the cell and the main body of the squall line were completely merged, and the merging process was over. The position of the convective cell remained almost unchanged during the whole process.

Figure 13 shows the RHI images of Z_H and V passing through the centroid of point C (along the dotted line in Fig. 12). It can be seen from the RHI images of V that Nanhai DP-PAWR clearly observed the flow-field structure and evolutionary characteristics of the main body of the squall line and its front. The near-surface layer was a wedge-shaped cold airflow from the northwest, and the mid - and low-altitude layers were composed of warm and humid air flowing from the southeast along the cold airflow to the main body of the squall line, and gradually rose. As the squall line continued to move closer, the intensity and thickness of the cold outflow from the main body and front of the squall line continuously enhanced and thickened. The maximum strength exceeded 20 m s^-1 and the thickness exceeded 3 km. At the same time, the warm inflow of the main body and front of the squall line also continued to increase, the maximum inflow height kept increasing, and the strength of the updraft also kept increasing. The maximum intensity exceeded 10 m s^-1, and the maximum inflow height reached 6 km. Combined with the RHI image of Z_H, we can clearly see the detailed evolutionary characteristics of the triggering, development, and merging of the convective cell with the squall line. In the early stage, due to the convergence of the southeast airflow and the northwest airflow below 3 km, at 18: 34 there was an enhancement of the triggering of the convective cell at a distance of 21 km and height of 4 km. Subsequently, the southeast airflow was maintained at the middle and low levels, the convergence gradually strengthened, and the high-altitude divergence gradually increased, making the convective cell gradually strengthen and stretch downward, and the maximum echo intensity reached 40 dBZ. At 18:35, the convergence caused a new cell to be generated at the rear of the original cell and in front of the squall line (distance, 24 km; height, 4 km). Subsequently, the new and old cells quickly merged, and the height of the merging was about 5 km. During the merging process, the strength of both cells increased significantly, and the height of the maximum echo gradually increased. By the end of the merging at 18:40, the maximum echo intensity had reached 60 dBZ, the height had reached 6 km, and the obvious weak echo area and echo overhang structure had appeared, indicating that there was very strong ascending motion in the lower altitudes. During this process, the low-altitude southeast airflow continued to stretch upwards and its intensity increased, and the convergence of wind speeds also continued to increase and gradually developed towards the rear side of the cell. The upper-altitude divergence reached its strongest level. At this stage, the main body of the squall line also continued to move southeastward, gradually approaching the convective cell. After 18: 41, the maximum strength of the merged cell exceeded 65 dBZ, its strong center stretched downward, and the strong echo of more than 50 dBZ gradually stretched below 2 km. The weak echo area and the echo overhang structure were more obvious, indicating that the ascending motion had been further enhanced. On the V map, it is also apparent that there was a continuous enhancement and upward development of the positive velocity zone. Affected by the convergence of airflow, there was a new convection system developing behind the new cell (distance, 26 km; height, 3-6 km). As the squall line moved south, an echo bridge appeared between the convective cell and the main body of the squall line at 18: 43 near the ground and at a height of 5 km. After that, the main body of the squall line continued to move south, the merging continued, and the height of the strong echo center gradually decreased. At this stage, the airflow at 2-4 km always maintained clear convergence. When the merging completed at 18: 47, the airflow convergence was also significantly weakened. In addition, a gale core with intensity greater than 20 m s^-1 at distance of 34-38 km and height below 1 km was observed by DP-PAWR at 18: 40, and the nearest automatic weather station (AWS) also reported strong surface winds of 15.6 m s^-1 (not shown). After that, the height of the gale core kept increasing and reached 2 km at 18: 43. At 18: 45, the surface wind observed by the nearest AWS also increased to 22.6 m s^-1, indicating that DP-PAWR is capable of revealing in detail the characteristics of the variation in winds in the surface layer.

Figure 13.  RHI sequence of the (a) Z_H and (b) V obtained by Nanhai DP-PAWR between 18:32 BT and 18:47 BT on 11 May 2020 (temporal interval: 90s).

By analyzing the characteristics of the dualpolarization parameters, a preliminary analysis was made on the phase change of the particles during this process. Fig. 14 shows the RHI images of Z_DR and K_DP during the process (along the dotted line in Fig. 12). As can be seen, in the triggering and development stage of the original cell (18:32-18:34), the area where Z_DR was greater than 0.5 dB was mainly located below 5 km, the intensity was 0.5-4 dB, and the Z_DR value above 5 km was about 0 dB, indicating at this time there were ice particles above the melting layer and below the melting layer were mainly small raindrops. With the triggering of the new cell at the back of the original cell and the strengthening of the two cells together, the Z_DR value continued to increase, the positive-value area gradually extended upwards, and the K_DP above the height of 3 km also increased. By the end of the merging of the two cells at 18: 40, the area where Z_DR exceeded 3 dB had increased significantly and stretched to the height of 5 km, and the maximum K_DP above the height of 5 km had exceeded 6° km^-1. Combined with the Z_H, it shows that with the continuous development of convection, heavy raindrops in the middle and low altitudes increased, while liquid particles entered above the melting layer to form wet hail or larger supercooled water droplets. With the merging of the convective cell with the main body of the squall line, the Z_DR value continued to increase. At 18:44, at the height of 1-6 km, the Z_DR was above 3 dB. The maximum Z_DR value was at the height of 1.5-4.5 km, exceeding 4 dB, and the K_DP was greater than 9° km^-1. The Z_DR and K_DP below 1.5 km decreased as the height decreased. This was because the high-altitude ice particles melted into heavy raindrops during the falling process, and then the heavy raindrops gradually broke up. Between 18:40 and 18:45, the nearest AWS observed 0.5 mm of precipitation (not shown). After that, the K_DP intensity increased and the height declined. Between 18: 45 and 18:50, the precipitation observed by the nearest AWS also increased to 6 mm. Additionally, according to the changes in dual-polarization parameters, it can be preliminarily inferred that the high-altitude ice particles gradually melted and shattered during the process of falling, and did not reach the ground. In this process, there was no report of hail observed on the ground. Hence, the dual-polarization parameters of DP-PAWR can better reflect the characteristics of the phase state change of the particles during the triggering, development and merging of the convective cell with the squall line.

Figure 14.  RHI sequence of the (a) Z_DR and (b) K_DP observed by Nanhai DP-PAWR between 18:32 BT and 18:47 BT on 11 May 2020 (temporal interval: 90 s).

For comparison, Fig. 15 shows the RHI images of the CR, Z_H and V of CINRAD/SA-D from 18:30 to 18: 48 on 11 May 2020 at 6-min intervals. As can be seen, CINRAD/SA-D was able to observe the development and evolution of the convective cell, but it had difficulty in capturing a detailed and complete picture of the evolutionary characteristics of the triggering, merging and strengthening of the new and old cells, as well as the merging process with the main body of the squall line-specifically, for instance, the start time and location of the merging process. In addition, due to the limitation of the observational blind zone, the low-altitude echo characteristics were not obtained during the process. It is evident that CINRAD/SA-D lost some of the features of the storm that are of key significance in furthering our understanding of the development and evolution mechanisms of meso-γ-scale and microscale weather systems and improving the ability of nowcasting.

Figure 15.  RHI sequence of the (a) CR, (b) Z_H and (c) V obtained by CINRAD/SA-D between 18:30 BT and 18:48 BT on 11 May 2020 (temporal interval: 6 min).

However, it should also be noted that the attenuation and limited range made it difficult for DP-PAWR to obtain a complete picture of the squall line. Therefore, DP-PAWRs can be used as a supplement to the CINRAD/SA-D network in daily operations, and their advantages in terms of a wide detection range in the vertical direction and high temporal and spatial resolution can be exploited to improve our ability to monitor mesoscale and microscale weather systems.

Based on observational data of the DP-PAWR in Panyu and CINRAD/SA-D in Guangzhou, the rationality of the Z_H, V, Z_DR and K_DP obtained by DP-PAWR was preliminary analyzed by using the location matching method. In addition, Nanhai DP-PAWR was used to observe the process of a convective cell in front of the squall line that was triggered and then developed and merged with the squall line. The advantages of the high temporal and spatial resolution data of DP-PAWR in studying the development and evolution of mesoscale convective weather systems were preliminarily examined. The main findings of this study can be summarized as follows:

(1) Compared with CINRAD/SA-D, DP-PAWR can obtain more detailed detection data, and shows better capability in detecting mesoscale convective weather systems, which is helpful in further studying and understanding the development and evolution mechanisms of meso-γ-scale and microscale weather systems. Additionally, the multiple elevation angles of RHI scanning enables a wider detection range in the vertical direction, covering the blind spots of CINRAD/SA-D in the low- and high-altitude areas.

(2) DP-PAWR's Z_H and V structures are acceptable, but its sensitivity is worse than that of CINRAD/SA-D. The Z_H of DP-PAWR suffers from attenuation and the Z_H area distribution is distorted around strong rainfall regions.

(3) DP-PAWR's Z_DR is close to a normal distribution but slightly smaller than that of CINRAD/SA-D. When the Z_H is greater than 35 dBZ (45 dBZ), DP-PAWR's (CINRAD/SA-D's), K_DP begins to increase as Z_H increases. The shorter wavelength of the X-band DP-PAWR makes the K_DP more sensitive and provides a better indication of precipitation.

(4) The minute-level data of DP-PAWR clearly revealed the triggering and development of local convective cells and the merging process of the two cells, showing a detailed and complete picture of the evolutionary characteristics of the cell merging with the main body of the squall line. It can better reflect the characteristics of particle phase change in the development and evolution of mesoscale weather systems. However, the attenuation and limited range made it difficult for DP-PAWR to obtain a complete picture of the squall line. Therefore, DP-PAWRs can be used as a supplement to the CINRAD/SA-D network in daily operations to improve our ability to monitor mesoscale and microscale weather systems.

It is worth noting that, although quality-control procedures such as ground-clutter suppression, noise elimination, and attenuation correction have been carried out for DP-PAWR prior to deployment, there are still certain differences in the Z_H and Z_DR intensities between DP-PAWR and CINRAD/SA-D. We need to analyze more cases to find the causes of the differences and to improve the quality control of the DP-PAWR measurements. Meanwhile, information such as signalto-noise ratio (SNR) is suggested to add, which is very helpful for better using of DP-PAWR. In addition, there is still room for improvement in the scanning strategy of DP-PAWR. The scanning mode with 12 layers of elevation angles between 0.9° and 20.7° with a resolution of 1.8° in the vertical direction limits its ability to detect low - and high-altitude areas. The scanning mode in the vertical direction needs to be optimized according to the actual situation of weather processes to improve the observational precision and detection ranges of DP-PAWR.