Abstract. Accurate identification of non-precipitation echoes (NPEs) in weather radar observations requires effective use of polarimetric signatures together with spatiotemporal structure. Here we present a unified deep-learning framework to quantify the independent and synergistic contributions of model architecture, dual-polarization variables, and short-term temporal evolution to NPE identification. Using data from the Guangzhou S-band dual-polarization radar, we conduct controlled comparative experiments with two representative architectures: a pointwise multilayer perceptron (MLP) and a Transformer-based Swin U-Net that explicitly learns spatial context. We further perform ablation experiments across single- versus dual-polarization inputs and single-volume versus two-volume inputs. Results show that architecture-driven spatial-context learning is the dominant factor: Swin U-Net consistently outperforms the pointwise MLP under all input settings. On a high-confidence test subset, for example, the Critical Success Index (CSI) increases from 0.887 for the dual-polarization MLP to 0.950 for the dual-polarization Swin U-Net. Dual-polarization variables provide essential microphysical constraints and substantially improve class separability, particularly for pointwise classifiers. Incorporating two consecutive volumes further improves performance by capturing short-term echo evolution, with larger gains for the MLP than for Swin U-Net. The best-performing configuration, combining Swin U-Net with dual-polarization and two-volume inputs, achieves a CSI of 0.953 on the high-confidence test subset. Notably, the Swin U-Net using only the reflectivity factor (ZH) as input retains strong skill (CSI = 0.927), indicating that spatial-context learning can partially compensate for missing polarimetry and thus providing a practical pathway for quality control of legacy single-polarization archives.

Rapid urbanization significantly modulates extreme rainfall over the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) along the South China coast. Based on 10 years of surface rainfall observations and radar mosaics, this study identifies the urban-mountain interface in the northeastern GBA as the primary hotspot for extreme short-term heavy rainfall events (76 cases) and tracks the associated organized convective systems (OCSs). This urban hotspot is governed by two dominant synoptic patterns: the southwesterly monsoon pattern (P1) and the low-level vortex pattern (P2). While both primarily involve propagating and local OCSs, the local type predominates (more than 55%). These OCSs exhibit distinct organizational characteristics. Compared to P1, local OCSs in P2 appear as smaller, slower, and more locally confined features. Notably, the urban modulation is synoptic-pattern-dependent, yet it consistently enhances local OCSs as they pass the urban-mountain interface. Local OCSs in P1 display a coherent single-peak expansion, resulting from the temporal alignment between the intensification of inbound convection and the development of urban-initiated cells. Conversely, those in P2 exhibit a double-peak structure, reflecting a separation between the rapid intensification of incoming convection near the urban boundary and the later incubation of new cells within the city. Despite these evolutionary divergences, the peak rainfall consistently occurs approximately 3 hr after urban entry or initiation, indicating a pronounced urban amplification of rainfall.

Spaceborne precipitation radars provide essential three-dimensional observations of cloud and precipitation systems, but their quantitative applications are limited by the suppression of weak echoes and contamination from ground clutter. In this study, we construct improved labels for the GPM Dual-frequency Precipitation Radar (DPR) by combining a three-dimensional noise and sidelobe detection algorithm with a joint Ku–Ka clutter-free bottom (CFB) estimation. These procedures enhance the detection of weak precipitation echoes and improve the separation of precipitation from clutter near the surface, thereby reducing blind zones while preserving as much valid signal as possible. Based on these labels, we develop a deep learning framework using U-Net and a wavelet-enhanced variant (WTU-Net). The WTU-Net incorporates wavelet convolution to better capture multi-scale features and suppress clutter while retaining weak precipitation echoes. Training and validation are conducted entirely on GPM DPR observations, and results show that WTU-Net achieves robust performance under imperfect labels, with notable improvements in weak echo detection and the retention of near-surface precipitation echoes compared with conventional approaches.

The microphysical variability of stratiform precipitation over North China is investigated using in situ aircraft observations from eight vertical spirals in three stratiform events sampled during the 'Demonstration Project for Precipitation Enhancement and Hail Suppression on the Eastern Side of Taihang Mountain' (PPEHS) field campaign (22 May 2017; 21 May 2018; 20 April 2019). A King Air 350 equipped with cloud microphysical probes documented particle size distributions, habits, and bulk properties from the ice region down through the melting layer, together with the ambient thermodynamic and vertical airflow conditions. All events exhibit a robust mean vertical structure in which the volume-weighted diameter D m increases and the total number concentration N t decreases toward the melting layer, indicating efficient aggregation and riming in the lower ice region. Superimposed on this common pattern, however, are pronounced differences in D m, N t, particle habits, and particle size distribution (PSD) shape among spirals and among cases, even under similar values of convective available potential energy (CAPE), relative humidity, and vertical wind shear. In the 22 May 2017 event, layered peaks and valleys in D m indicate alternating dominance of aggregation, riming, breakup, and secondary ice production, while the three spirals sample distinctly different melting-layer structures. The 2018 and 2019 events confirm strong horizontal contrasts between small-particle-rich and large-particle-rich sectors and show that large aggregates and rimed particles carry a substantial fraction of the condensate mass. Compared with midlatitude field campaigns such as BAMEX and PECAN, the stratiform events sampled over North China exhibit weaker fractional decreases in N t but stronger growth in D m and total water content, implying more efficient production of large ice particles. These findings highlight significant subkilometer to mesoscale microphysical heterogeneity in relatively uniform stratiform precipitation events and provide observational constraints for improving microphysical parameterizations in numerical models.

Spaceborne precipitation radars provide vertically resolved observations of precipitating systems with near-global coverage, but their coarse horizontal footprint limits the depiction of finescale convective structures and weak echoes. This study develops a deep-learning super-resolution (SR) framework for the dual-frequency precipitation radar (DPR) aboard the global precipitation measurement (GPM) core observatory satellite based on a latent diffusion model (LDM). A physically consistent training dataset is constructed by coupling convection-permitting weather research and forecasting (WRF) simulations with multifrequency radar forward modeling, where 1-km S-band reflectivity is used as the high-resolution (HR) target, and Ku/Ka reflectivity aggregated to a DPR-like similar to 5-km footprint at multiple heights is used as the conditional input. The proposed LDM performs conditional diffusion in a compact latent space learned by a vector-quantized variational autoencoder (VQ-VAE) and incorporates multibranch fusion of dual-frequency and multiheight features to enhance the effective horizontal resolution while producing S-band-consistent reflectivity factor and partially recovering weak-echo regions below DPR sensitivity. On the synthetic testing set, the LDM achieves substantially improved perceptual and distributional similarity [e.g., markedly lower Frechet inception distance (FID)/learned perceptual image patch similarity (LPIPS)] compared with representative deterministic and generative baselines, while maintaining competitive structure-focused skill for intense echoes. Evaluations on multiple real cases-including landfalling tropical cyclones (TCs) and midlatitude/inland convective systems-demonstrate improved echo continuity and sharper convective/rainband morphology that is broadly consistent with independent ground-based S-band radar observations, indicating promising transferability beyond the simulation domain. The proposed framework offers a practical pathway to enhance the utility of current and future spaceborne precipitation radars for finescale precipitation analysis, while providing an effective bridge between spaceborne radar observations and ground-based radar applications through the S-band-consistent HR reflectivity fields.

Based on 3 years of summertime radar observations in East China, this study quantifies the relationship between polarimetric radar signatures (PRSs) and retrieved raindrop size distributions (RSDs) in heavy-rainfall-producing convection. Multiple PRSs, including the 30-dBZ and 40-dBZ echo-tops, the integrated intensities of and columns, the maximum graupel and hail height, as well as the changes in and within the warm-cloud layer, can to some extent indicate the mean raindrop size, number concentration, and rain rate at the low level (1-km level). A Multi-Layer Perceptron model is designed and trained to preliminarily predict these RSD parameters using the PRSs as inputs. These RSD parameters are generally reproduced and the correlations between predicted and observed values are above 0.8. The study clearly demonstrates the quantitative constraints of PRSs on low-level RSDs in heavy-rainfall-producing convection. Such microphysical constraints have potential applications for polarimetric radar in the prediction of RSDs and rainfall intensity.

On 7 September 2023, a century-record-breaking precipitation occurred in the world's largest urban agglomeration. Notably, the precipitating system of Typhoon Haikui was significantly enhanced rapidly as it approached the metropolitan areas. This study investigates the underlying mechanisms and relative contributions of urban-anthropogenic impacts on metropolitan precipitation using the three-dimensional variational radar data assimilation and the Weather Research and Forecasting model. The results suggest that urban land use (URBAN), urban canopy (UC), anthropogenic sensible heat (ASH), anthropogenic latent heat (ALH), and building heights (BH) enhance precipitation by 40.6%, 18.0%, 17.5%, 12.7%, and 6.5%, respectively. Specifically, ASH and ALH increase temperature and humidity, enhancing the precipitation downstream of the urban core. In contrast, increased surface roughness associated with BH reduces wind speed, promotes low-level convergence and modulates rainfall distribution upstream of the urban core. These findings highlight the critical role of megacities in shaping extreme typhoon-induced precipitation, and underscore the importance of accurately representating UC processes in numerical simulations.

Accurate snowfall measurements are vital for disaster mitigation, climate studies, and many other applications. Widely deployed surveillance cameras offer novel possibilities for fine-scale snowfall observation. This paper proposes HySnowNet, a hybrid framework that first extracts snowflakes using physical frequency domain techniques, then applies deep learning network to estimate snowfall intensity. Extensive experiments demonstrate that the snowflakes extraction module significantly enhances performance across varying snowfall intensities and remains stable under winds up to 5 m/s, enabling HySnowNet to achieve Root Mean Square Deviation values of 2.42 and 1.23 mm/hr on the self-constructed data set and real-world observations, respectively. Moreover, cumulative snowfall estimation accuracy improved by 20.3% over S-band radar, demonstrating its value in supporting radar networks. However, its performance declines under extreme conditions (>5 mm/hr). The findings can build on existing surveillance resources to open a new avenue for high spatiotemporal-resolution ground snowfall measurements, supporting remote sensing observation networks and enabling effective blowing snow monitoring.

This study presents a novel method for measuring ground wind speed (WS) using audio data collected from surveillance cameras. The continuous wavelet transform is employed to model wind sounds and capture the dynamic variations over time. A deep-learning model integrating attention-enhanced Convolutional Neural Network and Bidirectional Gated Recurrent Unit architectures is developed to extract WS features from the time-frequency domain of the surveillance audio. For model training, a surveillance audio-based WS data set is constructed. Extensive experiments demonstrate that the proposed model achieves a WS level prediction accuracy of 84.56% for a self-constructed data set and 82.25% in real-world tests. Additionally, the model yielded root mean square error values of 1.84 m/s and 1.49 m/s for two typhoon events. Although challenges remain in improving low-speed wind measurement accuracy, this approach highlights the potential of a high-resolution, low-cost, urban wind observation network using surveillance cameras, significantly enhancing the granularity of urban ground wind observations.

Hailstorms rank among the most destructive extreme weather events globally, causing substantial property damage. While limited case studies suggest that cities may exacerbate hailstorms, the underlying mechanisms remain uncertain because of the complex physical processes. Here, we examine a hailstorm formation pathway associated with convective merging process using long-term observational data and high-resolution numerical simulations. This pathway helps explain the rising frequency of hailstorms across two distinct climate regimes, North America and East Asia. We find that merger hailstorms (MHs) occur approximately twice as often and tend to be more intense than non-merging normal hailstorms (NHs), which have been traditionally considered as the primary hailstorm formation mode. Favorable environmental conditions support the initiation of multiple convective cells and their subsequent merging, a tendency that may be enhanced by anthropogenic heat in large cities. Projections from a machine-learning model indicate an increase in the MH frequency and a decrease in NH frequency in North America. Together, these findings highlight an underexplored hailstorm formation pathway and suggest that climate change and human activities may play a role in shaping future hailstorm characteristics and the associated risks.

Abstract. Single-transmit, dual-receive millimeter-wave radars can be attractive for compact and scanning precipitation observations, but they do not necessarily provide the conventional horizontal/vertical dual-polarization variables used to constrain raindrop size distributions. This study evaluates whether a fixed slant-linear basis can turn the weak backscattering anisotropy of oblate raindrops into a useful liquid-rain observable. In the proposed geometry, a +45° transmitted polarization and co-/cross-slant receive channels project the difference between horizontal and vertical co-polar backscattering amplitudes into an orthogonal-slant return. We use explicit T-matrix scattering calculations, normalized-gamma and observed drop size distributions, and controlled neural-regression diagnostics to quantify the information content and observability of the resulting slant-linear depolarization ratio (SLDR). For the baseline Ka-band gamma ensemble, the reflectivity (Z+)–SLDR space separates characteristic drop size and concentration that are ambiguous under reflectivity alone. With 1 dB SLDR uncertainty, adding SLDR reduces the independent-test RMSE of mass-weighted mean diameter (Dm) from 0.672 to 0.243 mm and reduces the RMSE of log-transformed rain rate (log10R) from 0.163 to 0.055. The same sign of improvement persists for Nanjing 2DVD, EPFL HyMeX, NASA IFLOODS, and variable-shape gamma checks. Sensitivity tests show that random SLDR uncertainty degrades retrievals gradually, whereas relative co-/cross-slant gain bias and polarization leakage define stronger practical constraints. Through the controlled forward-simulation information-content tests, SLDR is identified as a physically distinct measurement coordinate for liquid-rain microphysics. The calibration and detectability conditions are also defined for a field test with suitable single-transmit, dual-receive radars.

Persistent knowledge gaps in precipitation microphysics, particularly the nonlinear coupling between microphysical process hierarchies and raindrop size distribution (DSD) variability, keep introducing systemic uncertainties into precipitation retrievals and model simulations. Here, we address this challenge through a unified framework that integrates observations from China's national-scale disdrometer network (1,031 sites) and 10-year global dual-frequency precipitation satellite dataset. First, a region-independent DSD continuum characterized by a universal linear relationship between raindrop diameter and concentration across diverse climatic zones is identified, extending and refining the conventional maritime-like and continental-like category. Then, we quantify the vertical stratification of microphysical processes in shaping and shifting this continuum. Implementations of our findings to reduce biases in current microphysics parameterizations are proposed and discussed. This study advances our fundamental understanding of the apparent heterogeneity yet inherent homogeneity in the microphysics of heavy precipitation, providing mechanistic insights to improve the performance of weather and climate models.

Abstract Accurate nowcasting of severe convective precipitation is critical for early warning yet still remains challenging. Although deep learning methods show promise, most models lack physical constraints, limiting their consistency with atmospheric processes. We introduce FURECast, a deep learning model that not only leverages three‐dimensional structure of polarimetric radar variables (ZH, ZDR, KDP) but also embeds their intrinsic self‐consistency relation as a physical constraint. The model employs an encoder‐translator‐decoder architecture, integrating multi‐level inputs via late fusion and evolving features through cascaded multiscale blocks. Moreover, a novel physical loss term is introduced to enforce microphysical consistency during training. Evaluated on S‐band (GD‐SPOL) and C‐band (NJU‐CPOL) radar data sets, FURECast achieves a 14.1% improvement in 90‐min critical success index (35 dBZ threshold) over the 2D reflectivity‐only baseline, while reducing physical inconsistency by two orders of magnitude. These results underscore the value of 3D polarimetric structure and physics‐guided learning in advancing convective precipitation nowcasting.

Accurate typhoon quantitative precipitation estimation (QPE) with polarimetric radar depends on the assumed drop-shape relation (DSR), but typhoon DSRs derived from surface 2D video disdrometer (2DVD) observations may not represent elevated radar sampling volumes. To address this issue, we apply the established polarimetric self-consistency framework to estimate radar-constrained effective linear DSRs from quality-controlled horizontal reflectivity factor (ZH), differential reflectivity (ZDR), and specific differential phase (KDP) observations using multi-sample optimization. The method estimates βeff, the slope parameter of the effective linear DSR, and is first evaluated using a separate non-typhoon event, for which the radar-constrained relation agrees well with the median axis ratios measured by a nearby 2DVD. It is then applied to six landfalling typhoons observed by the Guangzhou S-band polarimetric radar. The resulting βeff values range from 0.046 to 0.049 mm−1, indicating stable event-to-event behavior and greater effective oblateness than two published surface 2DVD-derived typhoon DSRs. Scattering simulations show that DSR choice has little effect on ZH but substantially affects ZDR and KDP, causing overly spherical DSRs to underestimate R(ZH, ZDR) and overestimate R(KDP). Gauge validation with more than 400 gauges shows that the radar-constrained effective and Brandes et al. DSRs provide more consistent rainfall estimates than the two surface-derived typhoon DSRs. Application without recalibration to Typhoon Lekima, observed by the independent Wenzhou S-band radar, reproduces the same overall performance grouping. These results identify a transferability limitation of surface-derived DSRs and demonstrate the value of radar-volume constraints for typhoon polarimetric QPE.

This study uses three double-moment bulk microphysics schemes to simulate four rainstorm events in China, aiming to evaluate the simulated raindrop size distribution (RSD) in heavy rainfall. Both the Thompson and Milbrandt-Yau schemes exhibit nearly constant near-surface raindrop mean diameter Dmwhen the rainwater content exceeds 2 g m23, noticeably diverging from Dm values retrieved by polarimetric radar. The variability in Dm between deep and shallow convection also steadily decreases from the top of warm-cloud layer to the surface. The National Severe Storms Laboratory (NSSL) scheme shows an unrealistic increase in the near-surface Dmvalues with increasing rainwater content. By investigating the relative contribution to the total raindrop number transfer rate, the raindrop self-collection and breakup processes are identified to control all the simulated RSDs in heavy rainfall. In these schemes, the raindrop selfcollection/breakup rate is anomalously positively correlated with the rainwater content. The self-collection/breakup efficiency Ec is a simple function of Dm in the schemes. In Thompson and Milbrandt-Yau, Ec is positive below a cutoff Dm value, indicating net raindrop self-collection, and negative above it, indicating net raindrop breakup. In contrast, Ec is always nonnegative in NSSL, suggesting net raindrop self-collection rather than breakup. Two sensitivity experiments are designed by swapping the Ec values between Thompson and NSSL. Simulated RSD characteristics from the two schemes are effectively exchanged. Such changes in self-collection and breakup settings directly affect extreme rainfall forecasting by altering raindrop fall velocities and affecting other warm-rain processes, with an impact of up to 30% on the maximum hourly rainfall forecasts. SIGNIFICANCE STATEMENT: Accurate parameterization of cloud microphysical processes and representation of raindrop size distribution characteristics are essential for high-accuracy quantitative precipitation forecast in numerical models. This study evaluates the simulated raindrop size distribution in multiple rainstorm events and the ability of key microphysical process parameterizations in heavy rainfall. Deficiencies in cloud parameterizations and raindrop size distribution simulations are revealed and found to obviously affect the quantitative forecasting of heavy rainfall. The findings also provide valuable insights into improving heavy rainfall forecasting from the perspective of cloud microphysics parameterization.

Abstract. Urban surveillance cameras offer a valuable resource for high spatiotemporal resolution observations of ground hydrometeor phase (GHP), with significant implications for sectors such as transportation, agriculture, and meteorology. However, distinguishing between common GHPs—rain, snow, and graupel—present considerable challenges due to their visual similarities in surveillance videos. This study addresses these challenges by analyzing both daytime and nighttime videos, leveraging meteorological, optical, and imaging principles to identify distinguishing features for each GHP. Considering both computational accuracy and efficiency, a new deep learning framework is proposed. It leverages transfer learning with a pre-trained MobileNet V2 for spatial feature extraction and incorporates a Gated Recurrent Unit network to model temporal dependencies between video frames. Using the newly developed 94-hour Hydrometeor Phase Surveillance Video (HSV) dataset, the proposed model is trained and evaluated alongside 24 comparative algorithms. Results show that our proposed method achieves an accuracy of 0.9677 on the HSV dataset, outperforming all other relevant algorithms. Furthermore, in real-world experiments, the proposed model achieves an accuracy of 0.9301, as validated against manually corrected Two-Dimensional Video Disdrometer measurements. It remains robust against variations in camera parameters, maintaining consistent performance in both daytime and nighttime conditions, and demonstrates wind resistance with satisfactory results when wind speeds are below 5 m/s. These findings highlight the model's suitability for large-scale, practical deployment in urban environments. Overall, this study demonstrates the feasibility of using low-cost surveillance cameras to build an efficient GHP monitoring network, potentially enhancing urban precipitation observation capabilities in a cost-effective manner.

The indirect radar reflectivity assimilation method, which assimilates retrieved hydrometeors from radar reflectivity data, is simple and efficient in severe weather forecasting applications. However, it suffers from retrieval errors due to the uncertainties in discerning multiple hydrometeor types based solely on reflectivity observations. To mitigate these inaccuracies, dual-polarization radar data are incorporated into the background-dependent indirect reflectivity assimilation method in this study. First, the contribution of multiple hydrometeor species to the whole reflectivity is estimated using the observed reflectivity and background microphysical information; then, the hydrometeor classification algorithm (HCA) product from dual-polarization radar observations is introduced to correct the dominant hydrometeor type if in error; and finally, the contribution factors are adjusted and used to retrieve multiple hydrometeor species from reflectivity data. Through a single squall line case, it is demonstrated that the incorporation of the HCA product from dual-polarization radar data leads to more reasonable hydrometeor identification, with more supercooled rainwater above the melting layer and more graupel at low levels, thereby refining the hydrometeor analysis. With the 15-min rapid update cycling configuration, the changes in the analysis field enable more cold rain processes, resulting in more intense latent heat release at higher levels and stronger cooling near the surface in the forecast. This in turn strengthens updraft motion and cold pools in the convective regions, thereby improving the reflectivity and precipitation forecasts. Four cases' quantitative evaluations of the 0-3-hr reflectivity and precipitation forecasts further validate the effectiveness of incorporating dual-polarization radar data in the assimilation process.

The mechanisms linking raindrop size distributions (DSDs) to environmental conditions remain poorly understood, limiting their practical application. We develop a unique fine‐scale vertical in situ data set to reveal the evolution of near‐surface DSDs and quantify how environmental factors modulate raindrop microphysics. Near‐surface raindrop breakup is identified as a common feature during the East Asian summer, with an average threshold diameter of 1.16 mm for breakup initialization. Further analysis reveals that relative humidity and wind speed exert opposing influences on raindrop microphysical processes, with coalescence favored in humid monsoon environments and breakup intensified within typhoon outer rainbands. By incorporating empirical relationships between these two environmental factors and microphysical processes, we derive observational constraints that significantly reduce biases in near‐surface rainfall estimates. For heavy rainfall cases the bias is reduced by up to 75%. These findings improve understanding of raindrop microphysics in boundary layer and help improve quantitative precipitation estimation.

The Hong Kong Observatory (HKO) installed an X-band dual-polarization Phased Array Weather Radar (PAWR) at its wind profiler station at Sha Lo Wan (SLW) in 2021 to monitor high-impact weather in Hong Kong. The PAWR could complete a volume scan in one minute with a spatial resolution of 30 meters. Dual polarimetric variables from the SLW PAWR, including differential reflectivity (ZDR), specific differential phase (KDP), and hydro-classification (HCL) products, were used to diagnose the vertical motion and lightning characteristics of mesoscale convective storms (MCS). Through variational data assimilation, three-dimensional (3-D) wind fields were constructed to validate the SLW PAWR observations. Two MCS events that occurred on 18 September 2022 and 17 June 2023 are central to this study. The findings include (1) negative ZDR serves as a good indicator of the occurrence of intense downdrafts associated with an MCS, a premise further supported by the 3-D wind field analysis results, (2) negative KDP suggested the formation of vertically aligned ice crystals which facilitated cloud electrification, and (3) HCL products indicated the presence of mixed ice crystals and graupel above the 0°C melting layer which promoted active cloud-to-cloud and cloud-to-ground lightning strokes. These results show that the SLW PAWR provides essential observations, which, when coupled with 3-D wind field analysis, can aid in enhancing the understanding of the dynamics and electrification processes within an MCS.

The impacts of the anthropogenic heat (AH) effect on the evolution of a merger‐formation bow echo over the Guangdong‐Hong Kong‐Macao Greater Bay Area are documented. The utilization of radar data assimilation greatly improves the simulated results comparing against observations, strengthening the robustness of analyses in this work. The simulation with AH effect produces the most accurate results compared to observations, exhibiting approximately 62% larger spatial extent of heavy rainfall (>30 mm) and twice the area of strong winds (>10.8 m s −1 ) compared to the non‐AH simulation. Additionally, the top 1% rain rates and surface winds from the AH‐included simulation are about 25% stronger and 23% greater, respectively, relative to the non‐AH counterpart. On the one hand, higher AH flux tends to enhance the values of convective available potential energy and vertical wind shear within urban areas on average, providing favorable thermodynamic environmental conditions for convective development. On the other hand, greater AH effect triggers stronger convective cell, leading to a more intense merged system. This cell plays a crucial role in the merger process and the formation of bow echo, but it does not persist sufficiently in the non‐AH simulation. A third sensitivity simulation, excluding the urban land cover, produces results comparable to those of the non‐AH simulation. This study quantifies the relative contribution of the AH effect to the evolution of convective systems and the associated weather‐related hazards over the Greater Bay Area, underscoring the significant impacts of AH forcing on the regional flow patterns and the corresponding convection dynamics.