China New Generation Doppler Weather Radar (CINRAD) plans to upgrade its hardware and software to achieve polarimetric function. However, the small-magnitude polarimetric measurements were negatively affected by the scattering characteristics of ground clutter and the filter’s response to the ground clutter. This polarimetric contamination was characterized by decreased differential reflectivity (Z_DR) and cross-correlation coefficient (ρ_hv), as well as an increased standard deviation of the differential phase (Φ_DP), generating a large-area and long-term observational anomaly for eight polarimetric radars in South China. Considering that outliers simultaneously appeared in the radar mainlobe and sidelobe, the variations in the reflectivity before and after clutter mitigation (ΔZ_H) and ρ_hv were used for quantitatively describing the random dispersion caused by mainlobe and sidelobe clutters. The performance of polarimetric algorithms was also reduced by clutter contamination. The deteriorated membership functions in the hydrometeor classification algorithm changed the proportion of classified echoes. The empirical relations of R(Z_H, Z_DR) and R(K_DP) were broken in the quantitative precipitation estimation algorithm and the extra error considerably exceeded the uncertainty caused by the drop-size distribution (DSD) variability of R(Z_H). The above results highlighted the negative impact of clutter contamination on polarimetric applications that need to be further investigated.

Raindrop size distribution (DSD) over the complex terrain of Guangdong Province, southern China, was studied using six disdrometers operated by the Guangdong Meteorology Service during the period 1 March 2018 to 30 August 2022 (~5 years). To analyze the long-term DSD characteristics over complex topography in southern China, three stations on the windward side, Haifeng, Enping and Qingyuan, and three stations on the leeward side, Meixian, Luoding and Xuwen, were utilized. The median mass-weighted diameter (D_m) value was higher on the windward than on the leeward side, and the windward-side stations also showed greater D_m variability. With regard to the median generalized intercept (log₁₀N_w) value, the log₁₀N_w values decreased from coastal to mountainous areas. Although there were some differences in D_m, log₁₀N_w and liquid water content (LWC) frequency between the six stations, there were still some similarities, with the D_m, log₁₀N_w and LWC frequency all showing a single-peak curve. In addition, the diurnal variation of the mean log₁₀N_w had a negative relationship with D_m diurnal variation although the inverse relationship was not particularly evident at the Haifeng site. The diurnal mean rainfall rate also peaked in the afternoon and exceeded the maximum at night which indicated that strong land heating in the daytime significantly influenced the local DSD variation. What is more, the number concentration of drops, N(D), showed an exponential shape which decreased monotonically for all rainfall rate types at the six observation sites, and an increase in diameter caused by increases in the rainfall rate was also noticeable. As the rainfall rate increased, the N(D) for sites on the windward side (i.e., Haifeng, Enping and Qingyuan) were higher than for the sites on the leeward side (i.e., Meixian, Luoding and Xuwen), and the difference between them also became distinct. The abovementioned DSD characteristic differences also showed appreciable variability in convective precipitation between stations on the leeward side (i.e., Meixian, Luoding and Xuwen) and those on the windward side (Haifeng and Enping, but not Qingyuan). This study enhances the precision of numerical weather forecast models in predicting precipitation and verifies the accuracy of measuring precipitation through remote sensing instruments, including weather radars located on the ground.

Radar quantitative precipitation estimation (QPE) is a crucial product for nowcasting and disaster warning. However, its accuracy is constrained by factors such as radar band, attenuation effects, and variations in the phase and microphysical properties of precipitation particles. Based on X-band phased-array radar data from Zhongshan City, Guangdong Province, this study compares and evaluates the QPE correction performance of three deep learning models: stacking ensemble learning, gated recurrent unit (GRU), and three-dimensional convolutional neural network (3D CNN). The aim is to explore the applicability of different model types under complex precipitation conditions. Data from August 2023 to August 2024 were used to construct the samples, with records from May 2024 held out as an independent test set and excluded from model training and hyperparameter tuning. Model performance was assessed under different radar combinations (three-radar, dual-radar, and single-radar configurations), temporal scales (minute and hourly), and precipitation intensities. The results show that: (1) at the minute scale, all three models improved the original QPE, reducing average relative error (RE) by approximately 24.6–29.5%, mean absolute error (MAE) by 23.2–27.7%, and root-mean-square error (RMSE) by 19.7–22.8%, while increasing correlation coefficient (CC) by approximately 20.4–20.9%. Specifically, GRU achieved the largest reduction in RE, stacking showed slight advantages in controlling MAE and RMSE, and 3D CNN and GRU showed similar improvements in CC. (2) At the hourly scale, the correction effect varied with precipitation intensity. In the light-to-moderate rainfall range ($0.1\le R<8.0\mathrm{mm}{\mathrm{h}}^{-1}$, where R denotes hourly rainfall), 3D CNN generally showed better error-control performance, whereas the advantage of GRU was less consistent among radar combinations. In the heavy-rainfall range ($R\ge 16.0\mathrm{mm}{\mathrm{h}}^{-1}$), stacking and GRU provided complementary value in some radar configurations, although model performance remained configuration dependent. (3) Case analysis shows that stacking can improve the original QPE at some extreme-precipitation stations, but correction performance in the extreme high-value range remains unstable, and GRU and 3D CNN are more prone to underestimation. Oriented toward operational applications, this study systematically evaluates the applicability and limitations of three model types under different scenarios while considering computational-resource constraints and timeliness requirements, thereby providing a reference for model selection and operational application in radar QPE correction.