Surface precipitation phase transition is conducive to devastating snowstorms and avalanches yet remains a global challenge due to the scarcity of surface observations. Here, we present the Real-time Precipitation Phase-Intensity Collaborative Retrieval Network (RePPIC-Net), a hybrid AI framework that quantifies surface precipitation phase from satellite observations. By integrating real-time 3D atmospheric physics fields from the AI-driven FuXi model with operational geostationary satellite observations through a hierarchical architecture, our system enables real-time monitoring of surface precipitation phase, as opposed to at least 4-hour latency of current operational systems. Validated against ground stations in China, RePPIC-Net achieves a Critical Success Index for Phase and Detection of 0.1574 (snowfall) and 0.3147 (rainfall) for 0.1-5 mm/h precipitation, outperforming 4-hour latency operational products' respective scores of 0.1001 and 0.3064. The real-time precipitation phase discrimination capability of RePPIC-Net allows the development of a satellite-based surface precipitation phase nowcasting system, meeting the need for 1-3 hour global surface precipitation phase transition warnings. RePPIC-Net provides a replicable blueprint for AI-powered real-time weather monitoring, filling a gap in wintertime weather disaster warnings.

To get more insights into the charge structure inside thunderstorms over high plateau regions, this poses significant challenges. Based on the data from the high-precision lightning VHF interferometer, C-band Doppler radar and multi-band electromagnetic field observation, this study presented the complete evolutionary characteristics of charge structure in an atypical bottom-heavy thunderstorm producing solely intra-cloud (IC) lightning over the Tibetan Plateau (TP). A persistent inverted dipole and a tripolar charge structure with a larger-than-usual lower positive charge center (LPCC) coexisted within this thunderstorm. At the initial and the sustained weak convection stages, the thunderstorm demonstrated a negative dipolar charge structure. As the convection further developed into a mature stage, an additional positive charge region appeared in the upper level, forming a tripolar charge structure with a larger LPCC at the bottom of an intense convective region. In contrast, a neighboring convective region preferentially developed the upper-level dipole charge structure rather than forming an LPCC. The IC flashes, with a maximum of 4 fl/min, predominantly occurred as negative IC flashes originating from the lower negative dipole, while few positive IC flashes occurred between the upper dipole charge structure. The LPCC was primarily associated with positive charged graupel particles. The middle negative charge region involved substantial negative charge carried by ice crystals and snow aggregates with graupel particles acting as the charge carrier in its upper part. The upper positive charge region was mainly contributed by ice crystals and snow aggregates.

Abstract. Current spaceborne precipitation radars, including the Global Precipitation Measurement Dual-frequency Precipitation Radar and the FengYun-3G Precipitation Measurement Radar (FY-3G PMR), provide unique three-dimensional (3D) observations of global precipitation systems. However, limited spatiotemporal sampling and relatively long revisit intervals constrain the monitoring of rapidly evolving storms. This study proposes a discrete multi-beam precipitation radar concept for small-satellite constellations and develops a radar-constrained active–passive inpainting framework for reconstructing unsampled inter-beam observations. Collocated FY-3G PMR reflectivity and Microwave Radiation Imager for the Rainfall Mission (MWRI-RM) brightness temperatures observed from the same platform are used to emulate sparse radar sampling. The proposed model integrates sparse radar reflectivity and 26 passive microwave channels within a squeeze-and-excitation U-Net. Independent experiments are conducted for the PMR Ku- and Ka-band observations under different beam-spacing configurations. Reconstruction performance is evaluated exclusively within the masked inter-beam regions, and a boundary-smoothness constraint promotes continuity between reconstructed and observed beams. Channel-ablation experiments are further used to diagnose model sensitivity to individual passive microwave channels. The results demonstrate that sparse radar profiles and passive microwave observations can jointly recover coherent 3D reflectivity structures over a range of sampling densities. Across both radar frequencies, the overall mean absolute error and error standard deviation remained below approximately 2.0 and 3.0 dB, respectively, across configurations in which one to five beams were skipped between adjacent retained beams. The dependence of reconstruction accuracy on beam spacing provides quantitative guidance for balancing beam number, cross-track sampling extent, and constellation cost, while the channel-sensitivity results inform the design of future active–passive payloads. The proposed framework also provides a transferable basis for retrieval and reconstruction algorithms for future sparse-sampling radar missions.

Snow formation is a complex interplay of multiple microphysical growth processes, and the prevailing snow characteristics are inherently linked to local climate. However, the persistent shortage of observations for characterizing snow microphysics at a global scale continues to constrain our understanding of snow growth processes. Here, we investigate snow riming and aggregation signatures in stratiform precipitation through triple-frequency radar observations collected during coordinated field campaigns across Southern China, the Eastern United States, Western Europe, Northern Europe and Antarctica. The results suggest that the velocity-based riming estimates are generally consistent with triple-frequency observations, and the riming frequency increases with temperature. Our analysis of dual-frequency observations in these field campaigns qualitatively indicate the dendritic growth zone around -15 degrees C playing a key role in initiating enhanced snow size growth, and reveals a generally temperature-dependent snowflake growth characteristics. The snow over Eastern US is characterized by the most prominent riming growth, corresponding to moderate to heavy riming. Triple-frequency signatures of snowflakes over west Europe are consistent with Southern China, while the latter shows a higher degree of riming. The weakest snow growth signatures were found over west Antarctica, potentially owing to the scarcity of ice nucleating particles and available water vapor for deposition. In addition, our statistics reveal a latitudinal dependence for snowfall detection limitations with current spaceborne Ku- and Ka-band radars, and shed novel insights into future triple-frequency satellite missions as well as joint application of weather and spaceborne radars.

Highlights What are the main findings? What are the implications of the main findings?Highlights What are the main findings? What are the implications of the main findings?Abstract Accurate retrieval of cloud liquid water content (LWC) and ice water content (IWC) vertical profiles remains limited by strong vertical variability and nonlinear dependencies among observed variables. Ground-based cloud radar reflectivity and microwave radiometer-derived thermodynamic profiles provide complementary constraints, but their joint use requires consistent time-height matching and bias-controlled predictors. This study develops a vertically structured machine-learning framework that explicitly represents profile-level dependencies by constructing vertical-structure-enhanced features to encode local gradients and contextual information, integrating multiple tree-based learners with heterogeneous configurations through a profile-aware stacking strategy, and introducing a profile-level refinement step to suppress layer-to-layer inconsistencies. The framework is evaluated using year-round Cloudnet observations from the Lindenberg site, where IWC RMSE decreases from 0.0152 g m-3 to 0.0092 g m-3 with R2 increasing from 0.412 to 0.784, and LWC RMSE decreases from 0.0786 g m-3 to 0.0591 g m-3 with R2 increasing from 0.303 to 0.606. Additional boundary-region evaluation shows that the improvement is particularly evident near radar-derived cloud boundaries, where cloud structure and hydrometeor content may vary rapidly with height. These results indicate that treating cloud retrieval as a vertically structured learning problem reduces inconsistencies inherent in pointwise models and establishes a data-driven baseline for incorporating vertical constraints into atmospheric profile retrieval.

Multi-instrument synergistic observation is vital for studying cloud and precipitation physics. However, using the nearest scan time for matching inevitably introduces temporal mismatches. Here we employ three advection correction methods for temporal matching in weather radar and spaceborne radar observations: Lucas-Kanade (LK), Variational Echo Tracking (VET), and Anisotropic Diffusion (AD). These methods calculate the movement speed of the storms using optical flow methods, and then determine their positions based on the elapsed time between instruments. Next, we conducted a quantitative assessment of the performance of these three methods based on the consistency of storm morphology and rainfall rates. Our results demonstrate that all three advection correction methods effectively reduce the discrepancies in morphology and rainfall rate among multi-source data. Without correction, the Coincidence Rate (CR) and Structural Similarity (SSIM) were 30.96% and 0.689 in the US and 29.44% and 0.670 in China, respectively. In comparison, applying the LK, VET, and AD methods increased those indices to 32.94%, 32.72%, 32.85% and 0.718, 0.715, 0.716 in the US, and 31.34%, 31.17%, 31.24% and 0.696, 0.694, 0.693 in China, respectively. The rainfall rate inconsistencies were also effectively reduced after advection correction. The performances among the three methods were similar. Overall, the LK method performed slightly better than AD, followed by VET.

The melting layer (ML) in stratiform precipitation can have attenuation effects on microwave signals, introducing major uncertainties in remote sensing and telecommunications. In this study, we derived parameterizations for ML attenuation at the Ku- and Ka-bands, which are frequencies widely used in ground-based and spaceborne meteorological radars. To achieve this, a Doppler-spectra-based dual-frequency differential approach was applied to a unique, long-term multi-frequency radar data set. Our results suggest that total ML attenuation strongly depends on the ML thickness, while the specific ML attenuation aligns well with previous case studies and simulations. Furthermore, we quantified the impact of riming snow on ML attenuation, and found that the Ka-band specific ML attenuation for rimed snow is nearly half that of unrimed snow in light rainfall, whereas this distinction becomes obscured as the rain rate increases. Based on the Ku- and Ka-band ML parameterizations (derived as functions of rain rate and reflectivity, respectively), we extended these parameterizations to the W-band to facilitate broader applications across diverse radar missions. By presenting the first long-term observational analysis of Ku- and Ka-band ML attenuation, this study facilitates the accurate quantification of ML attenuation, serving as a unique reference for ground-based and spaceborne radar observations.

Cloud detection is a fundamental task in atmospheric science and satellite remote sensing. While numerous algorithms utilizing multiple visible and infrared channels have been developed, the absence of visible light at night forces most current methods to rely on multi-channel thermal infrared (TIR) observations. Consequently, detection accuracy is significantly reduced due to the minimal thermal contrast between low clouds and the ground. Furthermore, distinguishing clouds under strictly moonless conditions remains a critical challenge. Leveraging the low-light observation capability of the Visible Infrared Imaging Radiometer Suite Day/Night Band (VIIRS/DNB), this study proposes a single-channel cloud detection algorithm. Based on the physical scattering of ground-based artificial lights by clouds, the algorithm integrates a feature-engineering layer with a Random Forest machine learning model. This moonlight-independent approach can rapidly determine cloudy conditions, offering a novel method for high-precision nighttime cloud detection. Validation experiments using a single fixed radar site in Longmen, China, with 97 rigorously synchronized satellite-radar sample pairs, demonstrate that the proposed algorithm achieves an overall accuracy of 86.6% (95% CI: 78.4–92.0%) against millimeter-wave cloud radar observations. While strictly reliant on stable artificial ground lights—making it primarily applicable to urban and artificially lit regions—this method provides a valuable supplementary tool for nighttime monitoring.

Abstract Our understanding of raindrop size distributions (DSDs) aloft is largely limited, owing to a scarcity of effective measurements. In this study, a two‐step optimal estimation technique is developed for retrieving binned DSDs based on a newly developed vertically‐pointing triple‐frequency radar system. In the first step, we employ the Tridon and Battaglia (2015, https://doi.org/10.1002/2014jd023023) approach for retrieving the DSD shape using Ka‐ and W‐band radar Doppler spectra. In the second step, we propose the synergetic use of C‐band radar (negligible attenuation) and W‐band radar (significant attenuation) for the retrieval of DSD number concentrations, making it a fully radar‐based technique. We applied this approach to a Mei‐yu frontal rainfall event in Eastern China. Validation against ground‐based OTT Parsivel2 disdrometer observations quantitatively shows good agreement. Beyond traditional disdrometer observations, our radar retrieval captures the first view of binned DSD variations in Meiyu front rainfall.

Abstract In this study, we characterized cold air outbreak (CAO) snowfall over the Shandong Peninsula, East China, using gauge and weather radar observations from 2020 to 2024. We developed a gauge‐radar matched data set and determined the coefficients in Z = aS b every 30 min using a differential evolution algorithm. The results show that the exponent b is relatively stable at 1.72 whereas the prefactor a shows large variations. Then, we fixed b at 1.72 and classified a values into three groups. Our analysis suggests that a larger prefactor a is characterized by smaller peak reflectivity, larger differential reflectivity, larger specific differential phase, larger mean volume diameter, smaller number concentration and smaller snowfall rates and vice versa. We present two cases with contrasting values of a (109.0 vs. 567.8). Surface disdrometer observations suggest that the first event is characterized by significant riming, whereas the latter exhibits signatures of unrimed snow aggregates. Further spectral analysis of the radar mosaic for the two events reveals wavelengths of 29.5 and 141.1 km, with corresponding periods of 0.6 and 3.1 hr, respectively. Multi‐event 2DFFT wavelength analyses suggest that the microphysics of CAO snowfall are associated with the spatial structure of the snowstorm system.

Cold-air outbreak (CAO) clouds significantly impact regional weather, precipitation, and aviation safety. This impact is especially pronounced in confluence zones, where cold continental air masses interact with warm, moisture-laden air currents emanating from oceans or lakes. In this article, an objective classification algorithm of cloud phase was suggested using the K-means clustering algorithm based on Polly(XT) lidar method and a gradient-based cloud identification algorithm, in order to determine the depolarization ratio threshold for supercooled water, mixed-phase, and ice clouds. The analyses of cloud phase characteristics and vertical structure were conducted for CAO events over the Shandong Peninsula, a region frequently affected by sea-effect snow disasters, using the winter period of 2024-2025 continuous observation data from the Eastern China Cold-Air Outbreak Snowfall (ECHOES) campaign. The results show the following. The supercooled water clouds were predominantly located above the ice clouds in the mid-to-upper troposphere, implying heterogeneous ice nucleation processes. The analysis of diurnal variation reveals that the probability of occurrence of supercooled water clouds remains constant during the whole day because of the stable atmospheric conditions. Synergy of lidar and radar observations clearly demonstrates their complementary detection abilities: radar outperforming lidar in lower troposphere and daytime conditions. The high coverage of the supercooled water in the lower troposphere poses a significant safety threat for aircraft icing. These results provide an observational basis for improving the cloud microphysics parametric schemes in numerical weather prediction and aviation safety.

The paper outlines research achievements of Chinese Academy of Meteorological Sciences (CAMS) in the development of new multi-band weather radar detection technologies, field experiments, radar data quality control, generation of secondary products, studies of cloud-precipitation processes and structures. In response to scientific frontiers and national requirements, CAMS has successively developed a C-band transportable dual-polarization radar, an X-band phased-array weather radar, a multi-band cloud radar observation system, and a C-band continuous-wave vertically pointing radar. Cloud-precipitation observation bases and severe convective weather observation bases have been established at Longmen, Shenzhen, and Foshan in Guangdong, as well as cloud-precipitation observation bases at Naqu and Mêdog on the Tibetan Plateau. Long-term field experiments are carried out focusing on rainstorms, typhoons, microphysical structures of clouds and precipitation at these sites. Utilizing these field experiment data together with observations from the national weather radar network, studies have been conducted on data quality control methods for W/Ka/Ku multi-band cloud radars, C-band vertically pointing continuous-wave radar, wind profilers, dual-polarization weather radars, and phased-array dual-polarization weather radars, aiming to remove non-meteorological echoes, reduce attenuation effects, and mitigate systematic biases in radar measurements. Mosaic and integration methods for S-, C-, and X-band weather radars have been developed to extend the spatial coverage of radar data, reduce biases in X-band radar data, and produce high-quality gridded radar data with high spatiotemporal resolution. Employing dual-band Doppler power spectrum analysis technique, methods have been investigated for retrieving vertical air velocity, raindrop size distributions, and drop size distributions of solid precipitation particles of different shapes, as well as vertical profiles of water content and rainfall intensity from multi-band cloud radars. Quantitative precipitation estimation, hydrometeor classification, tornado and mesocyclone identification, and networking approaches for multi-band weather radars have also been studied. Furthermore, nowcasting research based on artificial intelligence has been conducted, thereby enhancing the capability of radar systems to detect cloud and precipitation microphysical and dynamical parameters. Using field experiment data, the fine microphysical structure and evolution of clouds and precipitation in regions such as South China and the Tibetan Plateau have been investigated, providing more detailed data and products for cloud physics and severe weather monitoring and early warning. Many of these research results have been operationally applied to severe weather monitoring and warning services in China. Enhancing the detection capabilities, optimizing the accuracy of observation modes, and expanding the application of phased array weather radar technology, as well as advancing detection techniques using shorter wavelengths and their application in cloud process observation, will remain key research directions for CAMS in the future.

Under climate warming, frequent short-duration extreme precipitation events in coastal megacities exacerbate urban waterlogging, whereas the associated convective mechanisms over complex underlying surfaces remain poorly understood. On 21 July 2023, an extreme short-duration rainfall event (14:00–19:00 LST, peak intensity 127.3 mm h−1) struck Shanghai under weak vertical wind shear (VWS) conditions that cannot be fully explained by classic storm dynamics. Based on multi-source observations, this study shows that the middle and lower troposphere was controlled by warm, moist southwesterly flows, with highly favorable thermodynamic conditions (CAPE ~3300 J kg−1, CIN near zero) that only required weak local lifting to trigger convection. Both 0–1 km and 0–6 km VWS were below 7 m s−1, maintaining stable, upright updrafts that favored high precipitation efficiency. The formation and maintenance of the quasi-linear convective system and the resultant extreme precipitation depended critically on the southerly sea breeze, local mesoscale convergence, and cold pool feedback. Convergence induced by the complex underlying surface (urban friction, high-rise building blocking) played important roles in initiating convective cells, while the interaction between cold pool outflows and the sea breeze from the East China Sea and Hangzhou Bay sustained the system, which evolved into a unique “fish-shaped” rainstorm. Driven by dominant convective propagation toward unstable inland areas, the system moved west–southwestward across the coastal zone into central urban Shanghai. This mechanism differs from both the cold pool–VWS balance under strong shear and the urban convective relay propagation mode under weak VWS documented in previous studies. These findings provide new observational insights into the formation and maintenance of weak-shear, short-duration extreme rainfall in coastal megacities, and carry important implications for identifying convectively prone zones, optimizing spatial development patterns, and improving climate-resilient land management and urban planning practices.

A 10 h needle-shaped snow process occurred in Weihai,east of Shandong Province,on 21 February 2024.The snowfall amount reached the blizzard level,which is a rare occurrence.In this paper,the synoptic background and microphysical characteristics of the needle-shaped snow process are analyzed based on comprehensive observations of dual polarization radars,precipitation weather instruments,ground automatic stations,soundings,ERA5 reanalysis data and Quasi-Vertical Profiles(QVP)method.The causes of the needle-shaped snow are discussed.The results are as follows:(1)The needle-shaped snow process occurred under the background of large-scale rain and snow in China.During the needle-shaped snow period,freezing rain turned into ice particles in southern Shandong Province,and ice particles transformed into sheet or branch snow in central and northern Shandong Province.The influencing system was a return-flow situation,with strong northeasterly winds below 925 hPa and strong southwesterly winds above 700 hPa.(2)The cloud top height of the needle-shaped snow event was about 500 hPa,and temperature below 600 hPa maintained at-6—-3℃when the needle-shaped snow occurred.This is also the main characteristic that distinguishes needle-shaped snow from other types of snowfall such as ice pellets,freezing rain and plate crystals.(3)The diameter of needle-shaped crystal particles was 3-4 mm,the maximum was 8 mm,the final falling velocity was largely below 2 m/s,and the particle number concentration was two orders of magnitude higher than that of sleet.The snowfall intensity had a certain relationship with the size and particle number concentration of snowfall particles.The diameter of heavy snowfall particles with hourly snowfall greater than 1 mm was larger and the particle number concentration was higher.(4)Reflectance factor(Ze)was generally within 20-30 dBz,differential reflectivity(ZDR)reached up to 0.8-1.0 dB,and the high value area of differential propagation phase shift(KDP)was concentrated below 1 km during heavy snowfall period.(5)Supercooled water was abundant during the needle-shaped snow process,and there existed secondary production of ice,which led to a high ice crystal particle number concentration.

Motorcycle crashes are a major contributor to road traffic fatalities in Cambodia, where motorcycles represent the dominant mode of transportation. Given the spatial dependence and heterogeneity inherent in crash data, this study examines spatial associations between built environment characteristics, climatic factors, and motorcycle crash frequency across 197 districts in Cambodia in 2019. Global Moran's Index was used to assess spatial autocorrelation in crash frequency and explanatory variables. After evaluating the distributional properties of crash counts and multicollinearity among predictors, several regression models were estimated and compared, including Ordinary Least Squares regression (OLS), Poisson regression (PR), Negative Binomial regression (NBR), and Geographically Weighted Negative Binomial Regression (GWNBR). The results indicate that the GWNBR model outperforms global models by more effectively capturing spatial heterogeneity in the relationships between environmental factors and motorcycle crash frequency. Several variables exhibit relatively consistent spatial association patterns across districts: road length, road density, residential land use proportion, and precipitation are positively associated with motorcycle crash frequency in many locations, whereas population density, intersection density, and the number of annual rainy days are predominantly negatively associated. By revealing spatially varying association patterns in motorcycle crashes, this study provides evidence to support geographically differentiated approaches to motorcycle safety analysis and planning in Cambodia and other low- and middle-income countries.

The thermodynamic phase of clouds is fundamental to rain initiation and development. Although it is well established that majority of rainfall from mid- to-high latitudes originates from ice, there is a lack of consensus on the major thermodynamic phase of raining clouds over low latitudes. Previous CloudSat-based studies have yet reached consistent results on the fraction of ice-phase raining clouds Fice $\left({F}_{\mathit{ice}}\right)$ over Southern China. In this work, we found that CloudSat-based approaches are subject to metrics used for flagging rainfall events, which may lead to different Fice ${F}_{\mathit{ice}}$ estimates. Analysis of multi-year surface radar observations suggests that Fice ${F}_{\mathit{ice}}$ has diurnal variation and increases with rain rate, while the surface rainfall originating from warm clouds does not exceed 50 mm/h $\mathrm{m}\mathrm{m}/\mathrm{h}$. Using disdrometer observations to flag surface rainfall events, we found that about 71% $\%$ and 95% $\%$ of surface rainfall events and rainfall accumulation come from cold clouds, respectively. In addition, the observed rainfall accumulations over Longmen are consistent with the fifth generation atmospheric reanalysis (ERA5), whereas the frequency of warm rainfall over Longmen is significantly underestimated in ERA5.

Catastrophic extreme rainfall events pose significant threats on economics and human life, while such events are poorly predicted by current numerical models, partly due to a poor understanding and parameterization of extreme rainfall storm microphysics. However, fine-resolution raindrop size distribution (DSD) observations are usually not available due to the sparse deployment of disdrometers. To fill in this gap, this study applies an optimal estimation (OE) algorithm using the three parameters of the Gamma distribution as state variables to retrieve DSD parameters from S-band dual-polarization radar observations, thereby revealing the high-resolution spatiotemporal evolution of extreme precipitation. Results show that the radar retrievals significantly outperform traditional empirical approaches and demonstrate high consistency with disdrometer observations. The maximum hourly rainfall recorded by rain gauges during Zhengzhou and Qinzhou extreme rainfall events was 201.9 and 189.6 mm, respectively, while the corresponding radar-retrieved values were 193.23 and 203.7 mm. We show that Zhengzhou event with is characterized by a combination of high-particle number concentration ( log(10)N(w) approximate to 4.8 log(10)m(-3).mm(-1) ) and small-to-medium-sized droplets ( Dm approximate to 1.8 mm); in contrast, Qinzhou event exhibits lower concentrations ( log(10)N(w) approximate to 4.5 log(10)m(-3).mm(-1) ) and larger droplet sizes ( D-m approximate to 2.0 mm), suggesting more active warm rain processes. This study offers a robust technical approach for high-resolution characterization of microphysical processes in extreme precipitation events, with potential applications in improving microphysical parameterization schemes in numerical models, enabling the development of tailored forecasting strategies for different types of precipitation, and providing scientific support for early warning systems and disaster risk reduction decision-making.

This study provides the first evaluation of the detection capabilities of the FY-3G dual-frequency Precipitation Measurement Radar (PMR) using storm top observations. Additionally, the impact of the GPM-CO orbit boost in November 2023 on Dual-frequency Precipitation Radar (DPR) observations is assessed from the perspective of the PMR. The minimum detectable radar reflectivities for the FY-3G Ku- and Ka-band radars were determined to be 12.03 dBZ and 8.60 dBZ, respectively. Notably, the better sensitivity of the FY-3G Ka-band radar enabled the detection of more and higher storm tops and significantly reduced the misidentification of the melting layer as the storm top compared to the GPM Ka-band Precipitation Radar (KaPR). At the same time, the PMR is more significantly affected by ground clutter, partly as a trade-off for its enhanced sensitivity, compared to the DPR. This increased susceptibility necessitates more rigorous ground validation of its near-surface precipitation retrieval. The sensitivity degradation of the KaPR resulting from the GPM-CO orbit boost further intensifies the underestimation of snow dual-frequency ratio (DFR) measurements, resulting in an increased classification of precipitation as convective by the DFRm method. Our results are expected to illuminate future algorithm development and data comparisons of FY-3G and GPM-CO radars.

In this study, a regional Parsivel OTT disdrometer network covering urban Zhengzhou and adjacent areas is employed to investigate the temporal–spatial variability of raindrop size distributions (DSDs) in the Zhengzhou extreme rainfall event on 20 July 2021. The rain rates observed by disdrometers and rain gauges from six operational sites are in good agreement, despite significant site-to-site variations of 24-h accumulated rainfall ranging from 198.3 to 624.1 mm. The Parsivel OTT observations show prominent temporal–spatial variations of DSDs, and the most drastic change was registered at Zhengzhou Station where the record-breaking hourly rainfall of 201.9 mm over 1600–1700 LST (local standard time) was reported. This hourly rainfall is characterized by fairly high concentrations of large raindrops, and the mass-weighted raindrop diameter generally increases with the rain rate before reaching the equilibrium state of DSDs with the rain rate of about 50 mm h−1. Besides, polarimetric radar observations show the highest differential phase shift (Kdp) and differential reflectivity (Zdr) near surface over Zhengzhou Station from 1600 to 1700 LST. In light of the remarkable temporal–spatial variability of DSDs, a reflectivity-grouped fitting approach is proposed to optimize the reflectivity–rain rate (Z–R) parameterization for radar quantitative precipitation estimation (QPE), and the rain gauge measurements are used for validation. The results show an increase of mean bias ratio from 0.57 to 0.79 and a decrease of root-mean-square error from 23.69 to 18.36 for the rainfall intensity above 20.0 mm h−1, as compared with the fixed Z–R parameterization. This study reveals the drastic temporal–spatial variations of rain microphysics during the Zhengzhou extreme rainfall event and warrants the promise of using reflectivity-grouped fitting Z–R relationships for radar QPE of such events.

Within a meso‐ γ ‐scale convective storm, dynamic processes play a pivotal role in extreme rainfall production. However, there are still large unexplained gaps in understanding the effects of dynamic processes on the generation of extreme short‐term rainfalls. In this study, a nocturnal rainfall event with an extreme hourly rainfall (EHR) of 184 mm on 7 May 2017 over the coastal city of Guangzhou is examined based on cloud‐permitting simulations, focusing on the generation of the EHR. Results reveal that the EHR is featured by obvious horizontally delivered rainwater ( q r ) from the front to the rear within a meso‐ γ ‐scale convective storm. The horizontally delivered q r from the front of the storm overlayed on the q r produced by cloud microphysical processes locally overhead in the rear of the storm, leading to a deep q r layer with values over 4 g·kg −1 at the lowest 0–4 km levels above the ground. Thus, huge q r poured down in a short time, resulting in the EHR. According to statistical results, at least 80 mm q r was provided by horizontal delivery for the majority of grid points with hourly rainfall over 120 mm. This dynamic delivery mechanism is further confirmed by a trajectory analysis of raindrops. We argue that this mechanism may play a decisive role in EHR formation in particular scenarios while admitting that EHR can also be produced sometimes mainly via cloud microphysical processes. The formation mechanism of EHR proposed herein may help further understand and forecast localized extreme short‐term rainfall.