This study employed wind profile radar data from inland stations in Fujian to analyze changes in characteristic quantities near the time of extreme winds for typhoons that made landfall or affected Fujian during 2012-2019, using 265 station-hours of extreme wind samples. A retrospective analysis of Typhoon Mekkhala (2020) was conducted as validation. A low-level strong wind belt (LLSW) was present within 3 h prior to and at the time of over 95% of extreme wind station-hours, with maximum wind speeds below 2 km (Vmax) ranging from 8 to 36 m/s in over 95% of observations and predominant directions of NE, SE, and NW. During the 3 h preceding event onset, Vmax and low-level wind intensity index I progressively increased while the LLSW base H12 descended, peaking at onset. As extreme wind approached, 500–850 hPa vertical wind shear remained relatively small. Mann-Whitney U tests confirmed that Vmax, LLSW base, low-level wind intensity index, and vertical wind shear differed significantly between pre-gale and non-gale samples (p<0.001). Subsequent ROC analysis revealed that parameters related to LLSW had the highest discriminative power (AUC: 0.883–0.923). Using median thresholds for 1 h lead time (Vmax: 20.4 m/s; H12: 0.79 km; I: 30×10-3 s-1; vertical wind shear: 2.1×10-3 s-1), Vmax achieved the highest POD (50.2%) with a low FAR (3.0%), while index I showed the highest AUC (0.923) but a lower POD (43.4%). These quantified thresholds and process insights derived from wind profiler data can provide objective guidance for short-term forecasting of inland typhoon gales in operational settings.
Remote sensing provides indispensable observations for monitoring tropical cyclone (TC) surface winds, yet traditional numerical or parametric models often underestimate structural details and remain under-sampled. Leveraging more than 300 Sentinel-1 (S-1) synthetic aperture radar (SAR)-derived CyclObs wind products and the International Best Track Archive for Climate Stewardship (IBTrACS) best-track data from more than 200 TCs (2018–2024), we trained eXtreme Gradient Boosting (XGBoost) and Transformer models to reconstruct two-dimensional TC wind fields, and evaluated them on 100 independent TCs (2024–2025) against measurements from microwave radiometers onboard Soil Moisture Active Passive (SMAP), Advanced Microwave Scanning Radiometer-2 (AMSR-2) and Stepped-Frequency Microwave Radiometers (SFMR) measurements. The Transformer achieves the highest point-wise accuracy, with root-mean-square errors (RMSEs) of 2.85 m/s against SMAP/AMSR-2 and 3.06 m/s against SFMR, as well as superior correlation and structural similarity compared with XGBoost and the Generalized Asymmetric Holland Model (GAHM). However, its along-track reconstructions exhibit pronounced small-scale oscillations absent in the physically smooth GAHM, revealing that numerical closeness to observations does not guarantee physical authenticity. This purely data-driven approach lacks dynamical constraints and risks unphysical artifacts between tracks, thereby demanding independent two-dimensional validation. Structural validation against IBTrACS shows excellent correlations (r ≥ 0.90) for 34-kt (R34), 50-kt (R50), and 64-kt (R64), with RMSE decreasing from 42.71 to 11.03 km and biases converging toward zero, confirming faithful reproduction of inner-core gradients despite slight smoothing of the outer rainbands. Across 2–10 km resolutions, RMSE remains nearly constant (∼2.35 m/s) and maximum wind speed shows no systematic change, indicating that finer grids only sharpen asymmetry without improving error metrics, likely due to overfitting or interpolation rather than real oceanic signals. Performance degrades sharply above 50 m/s (sparse extreme-wind data) and for translation speeds larger than 11 m/s (inadequate learning of motion-induced asymmetry), yet these limitations are purely statistical, lacking mechanistic insight. Thus, while the Transformer outperforms conventional models and is viable for routine operational TC warning applications, its application to extreme wind speeds (>50 m/s) or fast-moving storms (>11 m/s) is inadvisable without rigorous physical validation, ideally coupled with dynamical models and bias-corrected training data.
This study integrates data on tropical cyclones, sea surface temperature (SST), and fishing activity derived from Automatic Identification System (AIS) records from 2013 to 2024. Using hotspot analysis, empirical orthogonal function (EOF) decomposition, and wavelet coherence analysis, it investigates the interactions among tropical cyclones, SST, and fishing activity in the South China Sea from both spatial and temporal perspectives. The results indicate that fishing hotspots in the South China Sea are primarily distributed along the western coastal regions, particularly concentrated in the Pearl River Estuary, the Beibu Gulf, and the central-western parts of the basin. Among these areas, the Pearl River Estuary is most strongly affected by tropical cyclones, whereas the Beibu Gulf and the central-western South China Sea experience relatively weaker impacts. In terms of spatial interactions, higher SST in the eastern South China Sea during summer and autumn provides substantial heat flux and moisture that facilitate the formation and intensification of tropical cyclones. As these cyclones move toward the western coastal regions, they enhance upwelling processes, leading to the formation of seasonal upwelling fishing grounds and consequently promoting the aggregation of fishing hotspots. Meanwhile, the upwelling effect also contributes to localized cooling in the western coastal waters. From a temporal perspective, the interactions among tropical cyclones, SST, and fishing effort are likely governed by an air–sea coupling mechanism jointly driven by the summer monsoon and the El Niño–Southern Oscillation (ENSO). Tropical cyclones, primarily modulated by the summer monsoon, enhance upwelling effects and primary productivity in the short term, thereby increasing fishing effort, whereas both tropical cyclones and fishing effort exhibit a lagged response to SST variability associated with the combined effects of ENSO and the summer monsoon. These findings provide a scientific basis for fisheries production and management in the South China Sea.
Using the TC precipitation database from the Yearbook of Tropical Cyclones (TCs) as the reference dataset, this study combines traditional statistical verification, error analysis, and the Object-Based Diagnostic Evaluation (MODE) spatial verification method to evaluate the performance of TC daily precipitation and total precipitation accumulation(TPA) forecasts. The evaluation covers nine TCs that affected mainland China in 2022, with forecasts from global and regional numerical weather prediction models, an ensemble-mean method, and the CMA Intelligent Grid Forecast System (CMA-NDFS). For TC daily precipitation, the seven forecast methods yielded average TS of 0.27 for heavy rain (daily precipitation ≥50 mm) and 0.18 for extreme heavy rain (daily precipitation ≥100 mm). Using a TS of 0.10 as the benchmark for a practically useful forecast, CMA-NDFS provided skillful TC daily precipitation forecasts with lead times of up to 120 h during the 2022 evaluation period. For TPA forecasts, the average TS across the seven methods were 0.36 (≥10 mm), 0.23(≥25 mm), 0.16(≥50 mm), 0.09(≥100 mm), and 0.01(≥250 mm), respectively. Among these, CMA-NDFS achieved the highest TS at most precipitation thresholds across the nine cases. The MODE method analysis revealed that the main source of systematic errors lied in underestimations of precipitation area for TPA forecasts. In addition, as the precipitation threshold increased, the total interest decreased and the likelihood of large angular differences increased.Forecast performance also varied with TC track: westward- and northwestward-moving TCs were generally better predicted than right-recurving TCs. Because the analysis is based on only nine cases from a single La Niña year, the results may be influenced by year-specific circulation anomalies. Further evaluation using a larger and more diverse sample of TCs is needed to consolidate and generalize these findings.
This study presents an integrated modelling approach to assess cyclone-induced tide–surge–wave compound flooding along the Bangladesh coast, focusing on Cyclones Sidr, Amphan and Sitrang. In this study, compound flooding is considered in the restricted coastal context, where astronomical tide, cyclone-induced storm surge and wave action interact to intensify inundation. Rainfall-runoff and upstream river discharge are not dynamically coupled in the present framework and are acknowledged as limitations and future extensions. A coupled Delft3D-FLOW–SWAN model, supported by Delft Dashboard, was used to simulate storm-induced water levels, wind fields and wave processes using historical cyclone-track data from the Indian Meteorological Department. The model was calibrated and validated using observed tidal water levels and significant wave heights from available coastal stations, showing good agreement with recorded events. Sensitivity analysis indicated that modelled surge elevations were strongly influenced by wind drag and nearshore roughness, while the separate contribution of atmospheric pressure requires further controlled simulations. The results show that Cyclone Amphan generated the largest inundated area, affecting approximately 9076 km2, equivalent to about 19.29% of the coastal belt, whereas Cyclone Sidr produced the highest simulated inundation depth, reaching up to 6 m. The novelty of this study lies in developing a calibrated Delft3D-FLOW–SWAN tide–surge–wave modelling framework and an event-specific cyclone inundation dataset for the Bangladesh coast. The generated cyclone footprints provide spatially explicit information on inundation depth and extent and can support future coastal risk mapping, early-warning applications and decision-support system development.
Accurate prediction of tropical cyclone (TC) wind fields is crucial for disaster mitigation, yet remains challenging, particularly for extreme events. Existing deep learning models often struggle to capture the diverse dynamics across the TC lifecycle and tend to under-predict intense storms. To address these issues, we propose a Multi-Attention Spatio-Temporal Generative Adversarial Network (MAST-GAN), which couples a multi-source feature encoder that jointly processes environmental fields via an CNN,Transformer backbone and TC attributes via an MLP, then fuses them with adaptive gating, a triple parallel attention fusion module that simultaneously emphasizes channel, spatial and temporal cues, and a multi-branch dynamic prediction head that adapts across the lifecycle by balancing global and TC-core forecasts while using an auxiliary branch to enhance extreme-event prediction. Experiments on a large dataset built from ERA5 and IBTrACS show that MAST-GAN improves TC wind field reconstruction and forecasting accuracy over strong baselines, with clear gains on an extreme-typhoon test set measured by RMSE and MAE, and case studies on Typhoons Bebinca and Pulasan further confirm realistic and robust wind field evolution.
The Hurricane Weather Research and Forecasting (HWRF) model has shown significant advancements in predicting tropical cyclone (TC) intensity and intensity changes over the past decades. However, forecasting the rapid intensification (RI) of TC remains challenging, with notable inconsistencies in HWRF cyclic runs. This study investigates the causative factors of RI prediction errors in the HWRF cyclical simulations of two consecutive intense TCs i.e., Super Cyclonic Storm (SuCS) Kyarr and Extremely Severe Cyclonic Storm (ESCS) Maha that occurred over the Arabian Sea (AS) in 2019. Most HWRF cyclic runs predicted delayed RI for Kyarr and early RI for Maha. For Kyarr, RI began at a higher intensity and with a weaker magnitude than observed. For Maha, RI began at a lower intensity and predicted a stronger magnitude than observed. These contrasting performances are attributed to differences in the evolution of upper-level winds, which affected Kyarr and Maha differently. For the Kyarr, enhanced upper-level winds increased vertical wind shear, tilting the vortex and disrupting symmetric convection in the inner core. This hindered the formation of an upper-level warm core, ultimately delaying the RI onset in model cyclic runs. In the case of Maha, large-scale deformation steering flows over the Arabian Sea played a key role. Differences in the evolution of upper-level anticyclones resulted in diametrically opposite forecasted tracks. Some initial model cyclic runs for Maha incorrectly predicted a westward track, while later runs successfully captured the sharp eastward recurvature. Kyarr and Maha traveled along the same path with a three-day interval during their intensification phases. The forecasted westward trajectories kept the Maha away from the region of upper ocean cooling (3.5°C) caused by Kyarr. Consequently, these cyclical runs predicted earlier RI, resulting in higher intensity. This research highlights the necessity of accurately depicting upper-atmospheric dynamics to enhance RI forecasts in coupled models.
Tropical cyclones (TCs) are infrequent but potentially high-impact hazards for Sri Lanka. Using IBTrACS best-track data and ERA5 reanalysis for 1980–2024, a Sri Lanka-centered climatology was developed to characterize storms approaching the island and the large-scale environmental conditions associated with their occurrence. Storms were identified using proximity thresholds within a 700 km screening domain; 67 TCs satisfied the ≤100 km direct-impact criterion, corresponding to approximately 1.5 events yr-1. Sri Lanka-influencing tracks were dominated by Bay of Bengal systems propagating westward to west-northwestward at low latitudes. Peak-intensity timing near Sri Lanka was strongly seasonal, with a primary maximum in October–December and a secondary peak in May, consistent with the broader bimodal seasonality of North Indian Ocean TC activity. Storm-event rainfall over Sri Lanka, accumulated within a fixed ±48 h window centered on closest approach, showed substantial overlap between direct-impact storms and near-miss storms, defined here as systems whose minimum center-to-coast distance was >100 km and ≤300 km. This result indicates that offshore storms can still produce notable rainfall over land even when their centers remain away from the coast. Trend analysis indicated no statistically significant long-term change in direct-impact frequency or genesis location over 1980–2024. Wavelet analysis further suggested that variability in annual direct-impact counts is weighted more strongly toward interannual than lower-frequency timescales, although no robust long-period oscillation is inferred because of the short record length and edge effects. Environmental changes were selective rather than uniform: storm-peak deep-layer shear increased significantly; OND mean SST and RH700 also increased significantly, whereas OND lower-tropospheric wind extremes, low-level vorticity, and upper-level divergence showed no significant long-term trends. Overall, the results show that near-miss storms remain rainfall-relevant for Sri Lanka and provide a Sri Lanka-focused hazard climatology linking storm proximity, rainfall-footprint relevance, and environmental conditions.
Emanuel’s potential intensity theory is widely accepted as providing a useful upper bound on tropical cyclone intensity for both forecasting and climate assessment purposes. However, recent revised observational and laboratory estimates of the mean value for the enthalpy and momentum exchange coefficients at near-surface wind speeds in major hurricanes have reduced the ratio of these coefficients to such an extent that potential intensity estimates, which depend on this ratio, are significantly reduced. In view of evidence that such estimates are already up to two intensity categories too low, a re-appraisal of the theory is called for. Such is the purpose of this paper. We have identified a range of issues with the theory that call into question its physical integrity. Some of the issues include the lack of a rotational constraint on the predicted intensity, the silence on the radial distance that air parcels are drawn inwards above the boundary layer, as well as the lack of dependence on the gravitational acceleration of the planet. Arguably, the most major issue is that the assumed steady-state flow configuration, with outflow everywhere above the boundary layer and exactly moist neutral ascent, is not dynamically consistent and could not emerge from any physically realistic initial-value problem. Some implications of these findings are discussed.
The characteristics of low level turbulence at the Hong Kong International Airport (HKIA) under the influence of Severe Tropical Storm Wutip in June 2025 is studied. Multiple sources of turbulence observations were studied and intercomparison analysis was performed. There were more than 80 reports of windshear and turbulence received from pilots in HKIA. The aircraft data derived Eddy Dissipation Rate (EDR) data and the airline downlink EDR data was compared with those obtained from Doppler Light Detection And Ranging (LIDARs) at HKIA. The LIDAR EDR data includes both the plan position indicator (PPI) scan EDR and glide path scan EDR. It is found that there is a positive relationship between the aircraft mean EDR values with LIDAR PPI scan when considering the values of LIDAR PPI scan over the corridor, with the correlation coefficient reaching a maximum of 0.38. Due to the time window required for the calculation of LIDAR EDR data, the higher frequency peak EDR values cannot be well captured. Besides, further analysis found that the turbulence reports are found to mainly occur at times when the crosswind strength is peaked. Studying the wind strength and period suggests that during the peak wind speed, the shedding period is shorter, indicating possible vortex shedding or periodic waves. This gives better understanding on the wind characteristics when it flow across the Lantau island mountains in the vicinity of HKIA. It also demonstrates the usefulness of aircraft turbulence data for developing and verifying EDR products from meteorological instruments.
In 2025, the Hong Kong Observatory introduced, for the first time in operation, artificial intelligence (AI) global weather prediction models for operational forecasting of tropical cyclones (TCs) over its area of responsibility, including the South China Sea and part of the western north Pacific. This paper documents the first-year operational experience of using such AI models and provides verification results for their track forecasting, temporal consistency and genesis. The AI models are found to have reduced forecast track errors, converge towards the actual tracks earlier, and have higher temporal consistency among successive forecast runs, relative to the selected traditional global numerical weather prediction (NWP) models. When combined the traditional global NWP models with AI models, a grand ensemble is possible to enhance the robustness of tropical cyclone warning service for Hong Kong, supporting earlier and more consistent forecast to facilitate early preparation work of the public and the emergency preparedness parties. While AI models are expected to be an indispensable tool for operational TC warning, their limitations on TC intensity and wind-structure prediction should be acknowledged. It should also be noted that present evaluation covers a single TC season, over a regional subset of TCs and a selected set of AI models; multi-year and multi-basin verification is needed to assess generality.
The formation and evolution of heavy rainfall associated with landfalling typhoons are closely related to their internal precipitation microphysical structures. However, the differences in precipitation microphysical characteristics among typhoons with different structural types and their underlying causes remain inadequately understood through systematic research. In this study, two landfalling typhoon cases, namely Typhoon "In-Fa" (2021) and Typhoon "Co-may" (2025), were selected for their distinct structural morphologies and evolutionary characteristics. Comprehensive utilization of ERA5 reanalysis data, ground automatic station precipitation observations, dual-polarization Doppler weather radar, and disdrometer data was adopted to conduct a comparative analysis of the precipitation structure and raindrop spectrum microphysical characteristics during the two typhoon events affecting the southern region of Jiangsu Province. The results are as follows. (1) The precipitation of Typhoon "In-Fa" was dominated by continuous and stable stratiform clouds and mixed-phase precipitation, with radar echo intensity mostly concentrated in the range of 30–45 dBZ and a relatively shallow vertical structure. In contrast, Typhoon "Co-may" exhibited significant differences across different impact phases: during its convective development phase, the radar echo intensity reached 45–55 dBZ, with a notably increased vertical extension height. (2) The raindrop spectrum characteristics of the two processes showed obvious divergence. The raindrop spectrum of "In-Fa" precipitation featured a broad distribution, with a large mass-weighted mean diameter and lower Nw, reflecting the important role of melting and collision-coalescence processes. In the remnant-vortex phase of Co-may, the opposite characteristics were observed: a narrow spectrum and a high total number concentration (NT), with warm rain processes playing a dominant role.(3) Environmental vertical wind shear and water vapor transport conditions modulate the intensity of convection development, thereby influencing the type of precipitation microphysical processes and the distribution characteristics of raindrop spectra. The study provides valuable insights for understanding the diversity of microphysical processes in landfalling typhoon rainfall and improving regional precipitation forecasting.
A better understanding of landfalling tropical cyclone (LTC) precipitation is crucial for accurate severe weather forecasting. However, it remains uncertain whether explicit convection or cumulus parameterization schemes (CPSs) yield superior performance in high-resolution models, highlighting the need for further research to identify optimal configurations for regional precipitation simulation. To address this issue, this study utilizes the Weather Research and Forecasting (WRF) model to conduct numerical simulations of precipitation associated with LTCs Lekima (2019)and In-Fa (2021)during the periods around landfall. The study systematically compares the influences of six different CPSs (the Kain-Fritsch (new Eta) scheme, the Betts-Miller-Janjic scheme, the Grell-Freitas ensemble scheme, etc.) and horizontal resolutions (18, 6 and 2 km) on the simulated precipitation characteristics for these two events. The results show that: (1) The model reasonably captures key precipitation features; scale-aware CPSs effectively reduce relative errors, whereas traditional CPSs perform better in reproducing spatial patterns; (2) High-resolution simulations improve spatial correlation (by approximately 0.1) over the concentrated precipitation areas and better resolve localized structures of heavy rainfall; (3) Scale-aware CPSs significantly enhance the simulation of intense precipitation by optimizing critical physical processes, such as moisture convergence, vertical motion, and ice-phase microphysics. Moreover, (4) they exhibit negligible convective contribution (<5%), in contrast to traditional CPSs, which display a pronounced dependence on grid resolution. These findings demonstrate that scale-aware CPSs are more sensitive to changes in model resolution, and when combined with high-resolution grids, substantially improve the accuracy of heavy rainfall simulations in LTCs.
When Severe Typhoon Podul affected southern Taiwan in mid-August 2025, a number of landing aircraft at Taoyuan International Airport in northern Taiwan conducted missed approaches (MAPs). This paper studies the potential meteorological factors contributing to the MAPs. From the flight data analysis, the MAPs were likely related to significant headwind changes, significant crosswind changes, and severe low-level turbulence. Such factors may be related to the boundary layer structure of the typhoon and the terrain upstream of the airport. To study these factors further, high resolution numerical weather prediction modelling using a mesoscale meteorological model has been performed by incorporating the terrain upstream of the airport. It is found that headwind and crosswind changes are more sensitive to the terrain, though the simulation results do not accurately reproduce the absolute magnitudes of headwind and crosswind changes as recorded by the aircraft, due to limitations of the turbulence parameterization scheme. The severe turbulence in the MAP cases may be related to mainly the boundary layer of the typhoon, and enhanced by terrain. Further research directions from this case are also discussed.
A summary is provided of aircraft missions into Super Typhoon Ragasa (2025), the most intense typhoon of the 2025 season that impacted the Philippines and south China coast. The aircraft missions occurred as Ragasa was approaching Hong Kong, with the Challenger aircraft from the Hong Kong Observatory and the King Air aircraft from the Asia-Pacific Typhoon Collaborative Research Center conducting joint missions into the typhoon, marking only the second time that such two-plane missions have occurred in the South China Sea. Unique measurements from these aircraft, including deep-layer dropsondes within the moat region between Ragasa’s primary and outer eyewalls and between the outer eyewall and spiral rainband, are described. Additionally, surface wind radii as measured by the dropsondes, and how they compare with satellite measurements provided ∼12 h after the aircraft missions, are presented. The temporal separation between the aircraft and satellite measurements provides the ability to monitor the evolution of the surface wind as the typhoon approaches land, a valuable observation for forecasters and storm surge modelers. An evaluation of the Shanghai Typhoon Model wind field based on a comparison with dropsonde observations is also performed, identifying potential sources of biases in the model and pointing toward ways to improve the model.Using these missions as a guide, suggestions for more fully exploiting the complementary capabilities of each aircraft are provided. These suggestions include optimizing the coordinated sampling to provide simultaneous measurements of surface winds in various quadrants and radii and providing dropsonde profiles of temperature, moisture, and pressure in the vicinity of in situ measurements of microphysics properties from the King Air. The use of composite datasets of dropsondes and Stepped Frequency Microwave Radiometer is also discussed as a way to obtain robust, quantifiable characterizations of tropical cyclone structure and how it relates to intensity change and associated hazards such as storm surge and rainfall, providing a way to advance the understanding and prediction of hazards and impacts from landfalling tropical cyclones.
In situ observations during tropical cyclone (TC) landfall remain limited in the Philippines despite being one of the world's most tropical cyclone-prone countries. This study documents a storm-chasing mission conducted by the Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) during Super Typhoon Fung-Wong (Local name Uwan), which made landfall over Aurora Province on 09 November 2025. The storm chaser team was deployed on 08 November 2025 at the Aurora Provincial Disaster Risk Reduction and Management Office (PDRRMO) in Baler (15.75°N, 121.54°E), enabling near-real-time monitoring within the storm’s inner-core environment. The mission integrated high-resolution surface observations, doppler radar analyses, and rapid post-event storm surge surveys. Surface observations obtained near the inner eyewall (∼45 km from the circulation center) captured a minimum pressure of 963.2 hPa, peak sustained winds of 86.5 km h-1, and a maximum 3-s gust of 128.96 km h-1. Comparative observations across Aurora Province revealed coherent pressure evolution but notable spatial variability in wind speeds, with stronger winds in the right-front quadrant and at more exposed, elevated stations, highlighting the influence of storm geometry and local terrain exposure. Radar reflectivity mosaics and vertical cross sections indicated a well-defined warm-core structure prior to landfall, followed by increasing eyewall asymmetry and structural degradation associated with terrain interaction and a possible incomplete eyewall replacement process. Post-event surveys conducted across 14 coastal communities documented high-water marks ranging from 2.4 to 7.6 meters above sea level, with highest values observed along embayed coastlines. Survey results indicate that wave setup and coastal geometry played a dominant role in amplifying total water levels. Overall, this study highlights the value of integrated storm-chasing activities in bridging real-time forecasting, in situ observation, post-event hazard assessment, and underscores the importance of high-resolution field data for improving TC forecasting, storm surge modeling, and risk communication in the Philippines.
An oft quoted benchmark for understanding the pattern of boundary-layer induced vertical motion in a translating tropical cyclone is based on approximate solutions for a steady slab boundary layer beneath a prescribed, uniformly translating axisymmetric vortex in gradient wind balance. The fidelity of this framework is investigated here in the context of a recent, idealized, three-dimensional, convection-permitting, numerical model simulation of a translating tropical cyclone in a deep uniform flow. These new analyses show that the slab boundary layer model captures rather well the spatial structure of the inner-core wind asymmetries in the boundary-layer, but overestimates the magnitude of these asymmetries. We investigate also to what extent the flow emanating from the boundary layer influences the pattern of deep convection in the intensifying vortex in the numerical simulation. In this simulation, the location of maximum ascent during the mature stage is displaced azimuthally upstream in the mid-troposphere compared with that at the boundary layer top, consistent with the differential rotation in a warm cored vortex where the azimuthal velocity decreases with height. This vertical-motion tilt pattern is a robust feature also of innermost arcs of subsidence inside the eye region. During the period of intensification and maturity, the strength of ascent in the numerical simulation is larger in the mid-troposphere than at the boundary layer top as would be expected in a region of buoyant updraughts. However, this velocity difference is reversed in sign during the later decay period, suggesting the decay is linked to the inability of deep convection to remove mass at the rate that it is being expelled by the boundary layer. Some implications of these findings are discussed.
Accurate estimation and prediction of rainfall during the tropical cyclones (TCs) are vital for disaster preparedness and inundation modeling. Reanalysis products provide an opportunity to understand the physical mechanism of rainfall variations associated with TCs through several surface, sub-surface and environmental parameters. This study evaluates mean rainfall characteristics from one of the best global reanalysis products – ERA5 and one of the best regional reanalysis products for the South Asian monsoon region – Indian Monsoon Data Assimilation and Analysis (IMDAA) for different intensity categories of TCs over both basins of the North Indian Ocean (NIO) during pre- and post-monsoon seasons against a merged satellite-gauge rainfall product. This analysis has been done for 41 TCs formed over the NIO between October 2015 and December 2023 due to availability of the reference rainfall data. Although mean TC rainfall patterns captured by both reanalysis products, they underestimate intense mean rainfall over both basins of the NIO for all TC intensity categories. Both reanalysis products underestimated mean rainfall by about 50% in TC inner core region during intensification stage. IMDAA showed 5-15% improvement in error than ERA5 in daily peak rainfall estimation over both basins of the NIO, except for the post-monsoon TCs over the Arabian Sea. As TC intensity is one of the major factors determining rainfall characteristics and associated hazards, TC intensity computed through the maximum sustained wind (MSW) and estimated central pressure (ECP) estimates from both reanalyses has also been assessed for these 41 TCs against the best-track data of India Meteorological Department (IMD). IMDAA showed improvement over ERA5 in bias of ECP by 1-3 hPa over the Bay of Bengal, while ERA5 showed better performance than IMDAA for pre-monsoon TCs over the Arabian Sea. The largest error in ECP of 19-20 hPa was observed for post-monsoon TCs over the Arabian Sea by both reanalysis products. Furthermore, IMDAA showed notable improvement in bias and errors over ERA5 in MSW estimation for TCs over both basins of the NIO. Results of this study would be vital for historical analysis of TCs over the NIO using global and regional reanalysis products and also provide a framework to assess projected TC rainfall and associated hazards for disaster risk reduction planning.
This study develops a deep-learning framework for detecting tropical cyclone (TC) life-cycle stages from infrared satellite imagery across six global ocean basins. You Only Look Once (YOLO) v8 models were trained separately to detect genesis, intensification, and decay stages within individual basins as well as in generalized settings. Detection accuracy was consistently highest during intensification, with F1 scores exceeding 0.80 in several basins, while genesis and decay remained more challenging due to weaker or more disorganized storm structures. Cross-basin experiments revealed substantial performance degradation when Northern Hemisphere models were applied to Southern Hemisphere storms, highlighting the influence of hemispheric rotation and storm asymmetry. Rotation-based augmentation (180 degrees rotation via image flipping) improved detection in the South Pacific and South Indian basins, particularly for genesis and intensification. Nonetheless, cross-basin transferability was constrained even within the same hemisphere, with notable weaknesses in the North Atlantic and variability across other basins. To interpret these disparities, three structural indices were evaluated; symmetry, texture, and shape complexity and Detected versus Non-detected distributions were statistically tested. Detected cases were generally associated with stronger texture and more symmetric structure, whereas shape-complexity contrasts were more basin-and stage-dependent. Spatial localization was evaluated using a YOLO box-midpoint center proxy, which can be biased for asymmetric systems. Overall, the results indicate that detection performance depends on life cycle stage, hemispheric orientation, and storm structural organization, underscoring the need for regionally adapted, structure-aware-specific generalization strategies for robust global TC detection. (c) 2026 The Authors. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Super Typhoon Ragasa in September 2025 ranked the second-strongest tropical cyclone (TC) in the South China Sea since records began in 1950. Its closest approach to Hong Kong was only about 120 km yet its impact on the territory was not particularly severe. This work analysed the meteorological perspective using observations and model forecasts, where Ragasa's high wind region skirted just around the door step of the territory of Hong Kong. This work also serves as a valuable case study for near-coast eyewall replacement of TC; and for meteorological authorities that pursue early warnings to intense TC by showing that (i) observational data, especially near-sea-surface dropsonde data, are crucial for determining the spatial extent of hurricane winds associated with Ragasa, and (ii) high-resolution regional model was used to capture subtle changes in Ragasa's track, intensity, and radius of maximum wind as it approached Hong Kong, thereby providing essential information for assessing its potential impact on the territory. Furthermore, the performance of location-specific wind forecasts over Hong Kong is discussed. A regional ensemble prediction system with data assimilation of all-sky radiance from satellites is shown to provide the smallest error on local wind forecasts among other global and regional deterministic models for Ragasa. (c) 2026 The Authors. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
