The Longmen Cloud Physics Field Experiment Base, China Meteorological Administration

South China, close to the South China Sea (SCS), is in the transport channel of the atmospheric energy and water vapor of the SCS summer monsoon, and closely related to the large-scale sustained drought and flood events in China (Zhang et al. ^[1]; Fu et al. ^[2]; Luo et al. ^[3]; Du and Chen ^[4]). Affected by the monsoon flow, heavy rainfall in south China occurs frequently and causes great damage with urban waterlogging, mountain floods, landslides, and debris flows, etc. Accurate forecasts of heavy rainfall events in south China not only help the rapid development of the society, but also greatly benefit the government to reduce economic and overall social losses due to severe weather.

The heavy rainfall in south China is generally from γ - scale convective cells and well-organized β - scale convection systems, always characterized by rapid development, strong nonlinearity and complex forcing mechanisms (Meng et al. ^[5]; Wang et al. ^[6]; Wu and Luo ^[7]; Wu et al. ^[8]). Due to the limited understanding of the developing and evolving mechanisms of heavy precipitation in south China, it is very difficult for the operational numerical model to accurately predict when and where the heavy rainfall will take place, which leads to poor emergency decision-making and social services.

Improving the accuracy of heavy rainfall forecasts in south China is the key and difficult issue at this stage (Luo et al. ^[9]). The numerical model has been the primary tool of modern weather forecasting. The ability of numerical weather models to predict heavy rainfall depends on the detailed description of the structure of meso - and small-scale systems and associated environment (Xiao et al. ^[10]; Yin et al. ^[11]). This requires improving the convective-scale data assimilation technology that helps to provide better initial state of the hydrometeor and environmental fields (Bao et al. ^[12]; Li et al. ^[13]). Meanwhile, it is necessary to provide the model with precise physical processes, especially the cloud microphysical process.

Cloud and precipitation microphysics and relative parameterization are one of the main factors affecting the forecast accuracy (Igel et al. ^[14]; Tapiador et al. ^[15]). Cloud microphysical processes vary in different regions, and the adaptability of the corresponding parameterization also has obvious regional limitations (Khain et al. ^[16]; Feng et al. ^[17]; Lai et al. ^[18]; Zhou et al. ^[19]). Moreover, a reasonable microphysical scheme can improve the capability of rainstorm range and intensity simulation, and effectively improve the precipitation simulation capability. The World Meteorological Organization (WMO) and many countries have long attached great importance to it and a series of scientific experiments have been carried out. For example, long-term observations were conducted by the Atmospheric Radiation Measurement program (ARM) at several fixed and mobile stations around the world since 1990 (Stokes and Schwartz ^[20]); the North American Monsoon Observation Experiment (NAME) was launched in the North American monsoon region in 2004 (Higgins and Gochis ^[21]); the Tropical Warm Pool International Cloud Experiment (TWP-ICE) was carried out in Darwin, Australia, in 2006 (May et al. ^[22]); and the Southwest Monsoon Experiment and the Terrain-Induced Mesoscale Rainfall Experiment (SoWMEX/TiMREX) was conducted in 2008 (Xu et al. ^[23]). These large-scale observation experiments took the macro- and micro-physical characteristics of clouds, precipitation, aerosols, and environment fields as their study subjects. However, cloud microphysical schemes largely depend on empirical or semi-empirical parameterized formulas in numerical weather models (Tapiador et al. ^[15]; Khain et al. ^[16]). Many key parameters are acquired from observations in middle and high latitudes. Therefore, it is necessary to establish an observation station in south China to obtain relevant data and then improve the cloud microphysical scheme.

Although there are vast operational observation networks in Guangdong province, they are not able to observe the cloud-scale dynamical and thermal conditions on the fine scale, with lacks of directly observed hydrometeors. These deficiencies make it difficult for cloud microphysics and hydrometeors assimilation. The lack of observational instruments and the poor understanding of cloud microphysical processes restrict the improvement of numerical weather forecast in south China. In addition, there are three major rainstorm centers in south China, namely, Longmen-Fogang-Qingyuan, Yangjiang-Yangchun-Enping, and Shanwei-Haifeng-Lufeng, with an average annual precipitation of more than 2000 mm, (Liu et al. ^[24]). In response to the above situation, the Guangzhou Institute of Tropical Marine Meteorology of the China Meteorological Administration (ITMM/CMA), cooperating with the Chinese Academy of Meteorological Sciences (CAMS), built the Longmen Cloud Physics Field Experiment Base (LCPEB), CMA, which covers two rainstorm centers in south China and is beneficial to carrying out in-depth research on the developmental mechanisms of heavy precipitation and improvement of the prediction ability of regional numerical models in south China.

The field base follows the strategy of developing from "scientific observation" to "improving mechanism understanding and forecasting technology". The scientific objectives of the base construction are mainly in the following five aspects of 1) making up for the drawbacks of the operational detection network, and obtaining three-dimensional macro- and micro- physical characteristics of clouds and precipitation; 2) improving the understanding of the mechanism of heavy rainfall formation in south China; 3) solving the problem of assimilating dense and unconventional observation data; 4) improving the applicability of cloud microphysics schemes in south China; and 5) solving the problem of the comparison and application of new observation data.

The following sections are intended to provide an overview of the LCPEB observations and some preliminary research results.

Focusing on the heavy rainfall centers in the central and western coastal areas of the Guangdong province (Liu et al. ^[24]), Longmen is set as the main site of LCPEB, together with some sub-sites (e. g., Yangjiang, Jiangmen, Guangzhou). Several instruments were installed at the LCPEB to improve the ability of observing detailed information on wind, temperature and humidity, as well as kinematic structure and microphysical properties of aerosol-cloud-precipitation interactions. The distribution of the observation stations and the main instruments are shown in Fig. 1. These instruments include a vertical-pointing radar with C-band frequency modulation continuous wave (VPR-CFMCW), a Ka/Ku band polarimetric radar, a Ka/X band polarimetric radar, several two-dimensional video disdrometer (2DVD) instruments, micro rain radars (MRR), several ceilometers, several wind profilers, a cloud condensation nuclei counter (CCN), a hydrometeor videosonde sounding system (HYVIS), a microwave radiometer, and solar radiation observation systems (Fig. 2). Details of these instruments are given in Table 1.

Figure 1.  The geographical locations of main instruments at the Longmen Cloud Physics Field Experiment Base, CMA. Terrain is shaded (units: m). The Longmen (LM) is the main site of LCPEB together with some sub-sites, e. g., Yangjiang (YJ), Shangchuandao, Jiangmen (SCD), Huangpu, Guangzhou (HP), Bohe, Maoming (BH).

Figure 2.  Major instruments installed at the Longmen Cloud Physics Field Experiment Base, CMA: (a) Vertical pointing radar with C-band frequency modulation continuous wave (VPR-CFMCW); (b) Ka/Ku band polarimetric radar; (c) Ka/X band polarimetric radar; (d) two-dimensional video disdrometer (2DVD); (e) micro rain radar (MMR); (f) Vaisala ceilometer (CL51); (g) C-band dual-polarization weather radar; (h) wind profile radar; (i) microwave radiometer; (j) cloud condensation nuclei counter; (k) hydrometeor videosonde sounding system (HYVIS); and (l) solar radiation observation system.

Relying on the LCPEB, the South China Monsoon Rainfall Experiment (SCMREX) (Luo et al. ^[9]), the Cooperative Observation Experiment of South China Monsoon/Typhoon Heavy Rainfall, and the Cooperative Experiment of Observation and Forecast in Large Cities of Pearl River Delta have been carried out since 2016. A large number of multi-source observations of the hydrometeors and environmental fields of heavy rainfall systems were obtained.

According to the statistical shape-slope ($\mu-\mathit{\Lambda} $) relationship observed by the 2DVD at the LCPEB, a constrained gamma (C-G) model was proposed to retrieve rain drop size distributions (RSDs) from S-band polarimetric radar observations in south China. The μ-Λ relationship, Λ=0.0241 μ² + 0.867 μ + 2.453, was obtained based on the 2DVD observations at the Longmen station in Huizhou City, which has a very good representation in this area (Liu et al. ^[25]). Based on the S-band polarimetric radar measurements of radar reflectivity (Z_HH) and differential reflectivity (Z_DR), the gamma (Γ) size distribution parameters (N₀, μ and Λ) can be retrieved by the C-G model of retrieval scheme (Fig. 3). Results showed that the radar retrievals were consistent with 2DVD observations. Meanwhile, with the C-band radar data quality controlled (Xia et al. ^[26]), the C-G model of retrieval scheme for the C-band polarimetric radar was also proposed (Ding et al. ^[27]).

Figure 3.  Scattering of RSD parameters Λ, Z_DR, and combined parameter log₁₀ (Z_HH/N₀), with heavy black lines indicating the fitted curves of filtered samples (Liu et al. ^[25]).

To estimate the rain and ice water contents from the S-band polarimetric radar observations, the relationship between Z^rain (the horizontal polarized reflectivity for rain, dBZ) and Z_DP (the difference reflectivity, dB) is derived from statistical analysis by using 2-year 2DVD measurements at the LCPEB (Li et al. ^[28]), as shown in Fig. 4. Based on the Z^rain-Z_DP relationship, the rain and ice masses are then estimated using the Z-M method provided by previous studies (Wang et al. ^[29]; Cifelli et al. ^[30]; Chang et al. ^[31]; Wu et al. ^[32]; Wu et al. ^[33]).

Figure 4.  Scattering of Z_DP and Z^rain, with the heavy red line indicating the fitted curve of RSD samples in Guangdong province, the dash blue line indicating the fitted curve in the Taiwan region (Chang et al. ^[31]), and the dash green line indicating the fitted curve in the Nanjing area (Wang et al. ^[29]).

The solid-state transmitter Ka-band vertical pointing radar (Ka-VPR) operates by three different operational modes with different pulse widths and coherent integration and non-coherent integration numbers to meet the requirement for long-term cloud measurements. To resolve the data quality problems caused by coherent integration and pulse compression, which are used to detect weak clouds in the cloud radar, studies focus on, firstly, analyzing the consistencies of reflectivity spectra using the above three modes and the influence of coherent integration and pulse compression, secondly, developing an algorithm for doppler spectral density data quality control, and thirdly, merging based on multiple-mode observation data (Liu et al. ^[34]; Liu and Zheng ^[35]).

Based on the Γ RSD assumption, the effects of temperature and turbulence on the ratio of Ka-band and Ku-band reflectivity density spectra and their relationships with RSD parameters were analyzed combined with the Ka/Ku-VPR system parameters (Zheng and Liu ^[36]). The effects of the sensitivity of the cloud radar on retrieved air vertical velocity, RSD and attenuation correction were simulated. The retrieving accuracy of the micro-precipitation dynamics and microphysical parameters from KaKu-VPR was better than that of the single frequency VPR, such as Ka-VPR or Ku-VPR from this evaluation.

Based on the study of the dynamic and microphysical parameters of precipitation using the KaKu-VPR, a retrieval algorithm of Double-Wavelengths Spectral Density (DWSZ) for the V_air, RSD, liquid water content, rain rate, and radar wave attenuation correction in rain areas with the reflectivity spectral density (SZ) data of the KaKu-VPR were presented in Fig. 5 (Liu et al. ^[37]). The V_air was studied using the KaKu-VPR and single Ka-VPR and Ku-VPR in a precipitation case in Longmen. The V_air and microphysical parameters retrieved from the KaKu-VPR were examined by using the RSD data from a disdrometer.

Figure 5.  Time-height profiles of (a) N_w (normalized drop number concentration), (b) D_m (mass-weighted mean diameter), and (c) liquid water content retrieved by DWSZ during 1223-1554 BJT on 8 May 2019 (Liu et al. ^[37]).

The retrieval method for estimating vertical air motion and three RSD parameters from spectrum moments estimated from two collocated 5550-MHz (C-VPR) and 35-GHz (Ka-VPR) was implemented (C/Ka-VPRs). The C-VPR operated within the raindrop Rayleigh scattering regime, and the Ka-VPR primarily operated within the Mie scattering regime (a non-Rayleigh scattering regime). The different scattering regimes result in differences in the estimated radar moments due to reflectivity weighting, which is a function of the raindrop size. The retrieval method uses mean RSD and the C-VPR reflectivity and the Ka-VPR spectrum variance. The RSD retrieval method uses Look-Up-Tables (LUTs) and simple calculations to guarantee efficiency. Finally, reflectivity is decomposed into two terms by using retrieved RSD parameters (Fig. 6), one representing number concentration and the other representing the RSD shape (Pang et al. ^[38]). These two terms are effective tools to characterize rain microphysics with quantities related to number-controlled and size-controlled processes.

Figure 6.  Contoured frequency-by-altitude diagrams (CFADs) of (a) - (d) vertical air motions, rain rate, liquid water content, and mean diameter of stratiform rain on 12 Jun 2016, (e)-(h) vertical air motion, rain rate, liquid water content, and mean diameter of convective rain on 9 Jun 2016, and (i)-(l) vertical air motion, rain rate, liquid water content, and mean diameter of convective rain on 14 Jun 2016. The black lines stand for the frequency of occurrence (Pang et al. ^[38]).

Using long-term observations of several 2DVD at the LCPEB, Lai et al.^[18] and Feng et al.^{[17, 39-41]} reveal the raindrop size distribution characteristics of heavy rainfall in typhoon, squall line, dry seasons, and wet seasons in south China. Fig. 7 shows the mean raindrop number concentration (N(D)) and raindrop diameter (D) in the typhoons, dry seasons, and wet seasons of south China. The N(D) in typhoons, dry seasons, and wet seasons all had a raindrop diameter peak of 0.3 mm. For midsize raindrops (raindrops diameter of 1-3 mm), the N (D) of typhoons is higher than that of dry seasons and wet seasons. For large raindrops (raindrops diameter > 3 mm), the N(D) of wet seasons is obviously higher than those of typhoons and dry seasons. Furthermore, wet seasons also have the maximum diameter.

Figure 7.  The mean raindrop size distributions of typhoons (red solid line), dry seasons (green solid line), and wet seasons (blue solid line) observed by several 2DVDs in south China.

According to the high spatiotemporal resolution measurements obtained by weather radars, the detail evolution characteristics of mesoscale convective systems can be analyzed in detail (Wang et al. ^[29]; Cifelli et al. ^[30]; Zhang et al. ^[42]; Wen et al. ^[43]; Chen et al. ^[44]; Guo et al. ^[45]; Han et al. ^[46]; Zhang et al. ^[47]). Based on the vertical structure characteristics of the radar reflectivity observed by VPR-CFMCW, the precipitation systems that occur during the rainy season in south China are classified into four types (i. e., stratiform, convective, mixed, and shallow systems) (Huo et al. ^[48]). Fig. 8 shows the radar reflectivity vertical structure of typical stratiform, convective, mixed, and shallow precipitation systems in the form of contoured frequency by altitude diagram (CFAD). Statistical results show that the convective precipitation contributes 62.7% of the total rainfall, and the occurrence probability of stratiform precipitation reaches 43.1%. The convective precipitation has the heaviest rain rate, largest mass-weighted mean diameter (D_m), and liquid water content, followed by the mixed precipitation, and the shallow precipitation is the smallest.

Figure 8.  Radar reflectivity CFAD of (a) typical stratiform precipitation, (b) convective precipitation, (c) mixed precipitation and (d) shallow precipitation observed by VPR-CFMCW at the Longmen station (Huo et al. ^[48]).

The convective and microphysical characteristics of extreme precipitation features (EPFs) over the Pearl River Delta are revealed by the Guangzhou S-band polarimetric radar measurements and 2DVD observations at the LCPEB (Yu et al. ^[49]). The convection intensity of EPFs are categorized into "intense", "moderate", and"weak"groups according to the maximum height of 40 dBZ echo-top, and they account for 17.3%, 48.6% and 34.1% of the total rainfall, respectively. The C-G model for retrieving Γ RSDs provided by Liu and Zheng ^[35] and the Z^rain-Z_DP relationship for retrieval of LWC and IWC provided by Wang et al. ^[29] are used to investigate the difference of the microphysical characteristics for these three types of EPFs. The vertical profiles of LWC, IWC, D_m and log₁₀N_w are presented in Fig. 9. It is noted that the vertically integrated ice water path (IWP) in the"intense convection"EPFs is the largest, and it is about half as large in the"moderate convection"EPFs and only a quarter as large in the"week convection"EPFs. There is a dependence of the RSD and warm-rain processes on the convective intensity, i.e., stronger convection shows a smaller LWP-to-IWP ratio, a larger raindrop size, increased raindrop mass, and stronger size sorting and more breakup of large raindrops.

Figure 9.  Vertical profiles of (a, b, c, and d) averaged liquid water content and ice water content (unit: g m⁻³) with bars representing the range from the 25th to 75th percentile and (e) averaged D_m and (f) log₁₀N_w retrieved from the Guangzhou S-band dual-polarization radar observations for all extreme precipitation features (EPFs) and the EPFs with intense, moderate, and weak convection, respectively (Yu et al. ^[49]).

In recent years, several extreme rainfall events occurred over south China. The multiscale characteristics and formation mechanisms of these rainfall events have been further investigated by using multisource data from the LCPEB (Xiao et al. ^[10]; Li et al. ^[13]; Wang et al. ^[29]; Li et al. ^[50]; Pu et al. ^[51]; Rao et al. ^[52]; Zhang et al. ^[53]; Ye et al. ^[54]; Chen et al. ^[55]). These studies suggest that taking full advantage of multisource observations is important to reveal new findings related to extreme rainfall.

Based on high-resolution reanalysis from the Variational Doppler Radar Assimilation System (VDRAS) that assimilated observations by the S-band Doppler radars and densely distributed automated weather stations, Rao et al.^[52] demonstrated the triggering mechanisms of a nocturnal convection event that occurred over the northern south China on August 14, 2014. Results showed that the enhanced large-scale prevailing monsoonal winds combined with local circulations (e. g., the northern downslope winds and deflected easterly winds) caused by multiscale orography, creating a strong near-surface convergence zone and updrafts, and thereby triggering nocturnal convection (Fig. 10).

Figure 10.  Conceptual diagrams for the nocturnal convection initiation mechanism over the northern Pearl River Delta in south China (Rao et al. ^[52]).

By utilizing data from the wind profiler, microwave radiometer, Guangzhou S-band dual-polarization radar, and the Guangdong-Hong Kong-Macao lightning location system as well as other traditional observations, Li et al.^[50] performed a detailed analysis of the record-breaking rainfall event (with the maximum hourly rainfall at 219 mm h^-1) over Guangzhou on 7 May 2017. They found that a shallow meso-γ-scale vortex helped organize convective updrafts, which was significant in the development of the extreme-rainfall storm. As shown in Fig. 11, dual-polarization radars and lighting observations illustrated that warm-rain processes were active in that storm and produced substantial large raindrops.

Figure 11.  Schematic diagram of the extreme-rainfall-producing storm over Guangzhou on May 7, 2017 (Li et al. ^[50]).

Combining detailed observational analysis with high-resolution simulations, Wang et al. ^[29], Pu et al. ^[51] and Ye et al. ^[56] showed that the warm-sector heavy rainfall over the western coast of Guangdong on 21-22 June 2017 was related to the quasi-stationary convective system. The convective system was sustained by the mesoscale convergence line, which was formed by the convergence of cold northerly winds (related to land breeze and cold pool outflows) and warm southerly winds. The heavy rainfall was dominated by warm-rain processes. As shown in Fig. 12, the humid environment tended to reduce the rain evaporation process, and the concentration of raindrops was high in heavy rainfall processes.

Figure 12.  (a) Time series of the averaged absolute humidity below 4 km; (b) time-height plots of relative humidity observed from the microwave radiometer at Bohe on 21-22 June 2017; and (c) frequency distributions of D_m (mm) and logarithmic N_w (mm^-1 m^-3) retrieved from the Yangjiang polarimetric radar over Jinjiang during the mature stage of the rainfall event (Pu et al. ^[51]).

New observations can be used to improve the numerical prediction through data assimilation. In recent years, dual-polarization radars have been considered in data assimilation. In order to improve the analysis and forecast of Tropical Storm Ewiniar (2018), Li et al.^[13] directly assimilated polarimetric observations, including reflectivity and differential reflectivity, with an ensemble Kalman filter based on the Weather Research and Forecasting model. As shown in Fig. 13, assimilating differential reflectivity and reflectivity can improve the forecasting accuracy of the rainbands location and intensity, especially for heavy rainfall exceeding 250 mm.

Figure 13.  ETSs of 24-h predicted accumulative precipitation from 1800 UTC 7 June 2018 at 250-mm thresholds for CNTL (without data assimilation), AZ (assimilating reflectivity), and AD (assimilating both reflectivity and differential reflectivity) experiments (Li et al. ^[13]).

According to the high-accuracy linear shape-slope (μ - Λ) relationship observed by several 2DVDs at the LCPEB, a high-precision and fast-solution method of gamma (Γ) RSD function based on the zeroth order moment (M₀) and the third order moment (M₃) of RSD was proposed by Liu et al.^[57]. Compared with the current usually used exponential RSD in the 2M microphysical scheme, the R and M₆ values obtained by this linear C-G method were obviously in better agreement with the Γ - fit RSD results from the 2DVD observations (as shown in Fig. 14). The zeroth order moment (M₀) and the third order moment (M₃) of RSD can be easily calculated from rain mass mixing ratio (Q_r) and total number concentration (N_tr) simulated by the 2M microphysical scheme, respectively.

Figure 14.  Comparison of the rain rate (a) and M₆ (b) calculated from the linear C-G method solution with the Γ fitted RSD from 2DVD observations in 2019; Comparison of rain rate (c) and M₆ (d) of the exponential function RSD with the Γ fitted RSD from 2DVD observations in 2019 (Liu et al. ^[57]).

The instruments and field experiments of the Longmen Cloud Physics Field Experiment Base, CMA, were introduced in this paper. Based on the dataset acquired by long-term field experiments, some preliminary results have been concluded in this paper, such as, a retrieval algorithm of microphysical parameters, macro - and micro - physical characteristics and formation mechanisms of heavy rainfall in south China as well as their application in numerical weather prediction models.

(1) According to the long-term 2DVD observations, a Γ function RSD parameters retrieval scheme and a rain/ice masses estimate scheme for S/C band polarimetric radars in south China have been established. The retrieval algorithm of V_air, RSD, LWC, R, and radar wave attenuation correction for KaKu-VPR, and the retrieval algorithm of V_air, D_m, and R for C-VPR have also been developed based on the long-term field experiment datasets. These results are useful for obtaining the 3-D microphysical structure of rainfall systems.

(2) The statistical RSD characteristics of rainy season, dry season, and typhoon rainfall in south China were revealed by long-term 2DVD observations. The vertical radar reflectivity structure characteristics of four typical rainfall (i. e., stratiform, convective, mixed, and shallow rainfall) during the rainy season in south China were obtained by VPR-CFMCW. Meanwhile, the differences of convective and microphysical characteristics of different-intensity EPFs over the Pearl River Delta were compared, and the results showed that "intense convection"EPFs have much more LWC and IWC and larger D_m, but slightly less N_w.

(3) By using multisource data from the LCPEB and operational observation networks, the multiscale characteristics and formation mechanisms of extreme rainfall events that occurred in south China have been further investigated. The results show that enhanced large-scale prevailing monsoonal winds combined with local orography circulations to create strong near-surface convergence zone and updrafts, triggering nocturnal convection over the northern Pearl River Delta. The results show that warm-rain processes were active and produced substantial large raindrops in the record-breaking rainfall event over Guangzhou on 7 May 2017. The warm-sector heavy rainfall event over the western coast of Guangdong on 21−22 June 2017 was dominated by warm-rain processes. Meanwhile, the concentration of raindrops was high and evaporation process tended to reduce.

(4) The research results were also applied in numerical weather prediction models, i. e., data simulation and microphysical schemes. The S-band polarimetric observations, including reflectivity and differential reflectivity with an ensemble Kalman filter, have improved the forecast of the location and intensity of Ewiniar's rainbands. A high-precision and fast-solution method of gamma (Γ) RSD function, suitable for the 2M microphysical scheme, has been established according to the high-accuracy linear shape-slope (μ-Λ) relationship observed by several 2DVDs at the LCPEB.

Despite the above-mentioned advances, several aspects remain in need of further improvement in future work: 1) multi-source fine observation capabilities of boundary layer thermodynamic structure and microphysical characteristics in coastal areas and SCS; 2) further in-depth mechanisms for convection initiation and evolution of monsoonal heavy rainfall in south China, especially for warm-sector heavy rainfall; and 3) capability of rainstorm forecast in south China through assimilating new observational data and optimizing parameterization of key physical processes and applying artificial intelligence.