The U.S. National Weather Service’s (NWS) mission includes providing impact-based decision support services to its core partners for the protection of lives and property. Unlike traditional forecast products that focus on accumulation totals, the NWS Weather Prediction Center (WPC) has developed a cutting-edge impact-based winter forecast product – the Winter Storm Severity Index (WSSI) – to translate the forecast to societal impacts. The WSSI is a suite of operational forecast tools that integrates the official gridded forecast data with climatological and non-meteorological information to create a situational awareness forecast tool that communicates the location, timing, and severity of anticipated societal impacts from winter hazards.

WSSI classifies impacts for several winter hazards including snow and ice accumulations, blowing snow, and snow load. Impacts captured in the WSSI include but are not limited to: impacts to transportation, damages to vegetation, disruptions to utilities, damages to infrastructure, and the overall disruption to daily life. Accurate depiction of the societal impacts from winter weather hazards involves an interdisciplinary approach that leverages not only improved forecasts but new and existing research on how winter hazards interact with the natural and built environment. This presentation will include: 1) An overview of the WSSI suite of products (i.e., the deterministic and probabilistic operational versions, in addition to hourly and probabilistic hourly products in development), 2) Examples of how the WSSI scientists and developers approach improvements and additions to the underlying algorithms, and 3) Examples of how the WSSI product is used in the field with forecasters and core partners.

How to cite: Tobin, D., Harnos, K., and Nelson, J.: The Winter Storm Severity Index: Forecasting and Communicating Winter Hazards, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-243, https://doi.org/10.5194/ems2026-243, 2026.

How to cite: Gascón, E., Haiden, T., and Vannière, B.: Forecasting extremes: what AIFS and the physics-based km-scale global IFS model can (and can’t) do , EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-212, https://doi.org/10.5194/ems2026-212, 2026.

Approximation of vertical turbulence depends on the ability of the forecast model to resolve as many vertical levels as possible. There might be times when forecast inaccuracy relies on a certain atmospheric level, especially, under complex terrain cases such as orographic flows. Therefore, this is the point when vertical resolution becomes vital in predicting phenomena such as turbulence, convective transport and clouds that are partially resolved otherwise by coarser vertical resolution.

It is here investigated how by altering the number of vertical levels at specific altitudes above earth surface impacts forecast accuracy. The global non-hydrostatic Conformal Cubic Atmospheric Model (CCAM) is used, with a global stretched grid targeting high resolution over part of North Greece (1-km) and the wider region (8-60-km). The grid configuration in CCAM allows the placement of a high-resolution face centred over the domain of interest with gradually reduced resolution away from this region. This more seamless grid configuration allows coupling between the global and regional spatial scales on the same grid and may provide benefits for the representation of storms as they approach the domain of interest. CCAM offers the ability of selecting different number of vertical levels ranging between 0-43,900m from surface.

The importance of vertical levels resolution was made obvious during an extreme rainfall event in Northern Greece when coarse approximation failed to predict the heavy rain event but only became evident when the number of vertical levels was improved. The area is characterized by complex morphology with orographic flows adjacent to sea breezes from Aegean Sea. In the first instance 27 vertical levels (0-35,000m) were selected underestimating intense rainfall of the following three days. On the contrary, the 54 vertical levels (0-43,700m) run predicted efficiently the rain event due to finer vertical resolution at certain altitudes. This was proved by comparing daily and hourly rain levels by differing forecast intervals before the event.

Acknowledgements: Marcus Thatcher,^, Commonwealth Scientific and Industrial Research Organisation (CSIRO) Oceans and Atmosphere, Aspendale, VIC, Australia

How to cite: Zoras, S.: Understanding the importance of vertical levels resolution in forecasting an orographic heavy rainfall event, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-770, https://doi.org/10.5194/ems2026-770, 2026.

This study analyzed the integration of Local Climate Zones (LCZ) with machine learning (ML) has advanced urban heat research and quantifying intra-zonal heatwave variability within built LCZ classes (1–10). However, most studies remain constrained to single-city analyses and treat LCZ classes as internally homogeneous. Heatwave events (2013–2025) were identified from the weather stations data through applying a WMO definition for Bari and KNMI local criteria for Amsterdam. LCZ maps for 2025 were derived at 500 m resolution, while morphological subclasses were defined at 100 m and aggregated to capture intra-zonal heterogeneity using a WUDAPT-GIS approach. For Amsterdam, quantify intra-zonal thermal heterogeneity in built-up LCZ classes, parameterized by key parameters like built-up index, impervious surface fraction, and surface albedo. In parallel, the heatwave-driven thermal behaviour and relative UHI dominance of natural LCZ subclasses were explicitly analysed to capture vegetation-mediated cooling dynamics and their interaction with urban heat island and heatwaves events. A Random Forest model estimated LCZ-specific land surface temperature (LST) during heatwaves, with SHAP analysis used to quantify predictor importance.

We find that model performance is robust (R² = 0.79 to 0.85) in different LCZs during heatwaves events in the both cities. In Bari, LCZ-based modelling demonstrates strong performance (R² ≈ 0.81; RMSE ≈ 2.0 °C), with compact classes showing pronounced heat storage and nocturnal heat retention. In Amsterdam, the inclusion of LCZ subclasses reveals additional intra-zonal variability (up to 0.7–2 °C) and alters the spatial extent of high-temperature areas (15% to 20%). These results indicate that LCZ classes are not thermally homogeneous and that incorporating intra-zonal morphological detail improves to understand the heatwave hazards in built-up areas and spatial heat-risk assessment.

How to cite: Nadeem, F., Milošević, D., Berardi, U., and Steeneveld, G.-J.: Quantifying Intra-Zonal Heatwave Variability Using Local Climate Zone Subclasses and Machine Learning: A Cross-Climate Study of Bari (Italy) and Amsterdam (Netherlands), EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-662, https://doi.org/10.5194/ems2026-662, 2026.

The rapid urbanization occurring in the Maritime Continent—an expansive geographical domain encompassing Southeast Asia and Oceania—has intensified the exposure of population centers to extreme heat. However, the seasonal drivers and physiological consequences of this thermal stress exhibit spatial heterogeneity. This study investigates the seasonal variation of 2-m air temperature and human biometeorological indices across three major Indonesian cities—Medan, Jakarta, and Surabaya—representing diverse geographical conditions, monsoon onsets, and other climatological regimes.

Meteorological surface observations reveal a distinct bimodal temperature distribution in all three cities, with peak heat stress intensities typically occurring during the monsoonal transition periods in April and October. However, the magnitude and primary peak month vary substantially due to regional climatological influences. In Medan (Northern Sumatra), April is identified as the hottest month, where high ambient temperatures are exacerbated by high specific humidity. In contrast, Surabaya (East Java) experiences its maximum thermal intensity in October, coinciding with the late dry season's clear-sky conditions. Meanwhile, Jakarta (Western Java) exhibits nearly symmetrical peaks during both transitional periods, reflecting its intermediate position relative to the shifting monsoonal air masses.

To quantify the localized impact, canyon layer Urban Heat Island (UHI) intensities are estimated by comparing urban-rural station pairs. Furthermore, human thermal comfort is evaluated using two robust biometeorological indices: the Universal Thermal Climate Index (UTCI) and the Physiologically Equivalent Temperature (PET). By employing both indices, the study estimates day and night-time heat stress levels and the resulting physiological strain on the urban population. Preliminary results indicate that high humidity levels in cities like Medan and Jakarta substantially sustain elevated UTCI and PET values into the night, effectively hindering nocturnal cooling. These findings underscore the importance of incorporating accurate meteorological factors for localized heat-health warning systems and urban planning to effectively mitigate the escalating risks of heat-related disasters in tropical environments.

How to cite: Subasita, N., Ferdiansyah, M. R., Milošević, D., and Steeneveld, G.-J.: Understanding of the Atmospheric Dynamics of Extreme Heat Events in Tropical Cities of Indonesia: An Analysis from Observation Networks in Medan, Jakarta, and Surabaya, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-531, https://doi.org/10.5194/ems2026-531, 2026.

In this paper, conventional observation data, Fengyun-4 satellite, wind profile radar, S-band dual-polarization Doppler radar, lightning positioning system data, automatic weather stations combined with radar four-dimensional variational assimilation system are used to comprehensively analyze a supercell storm in Beijing on 4 June, 2022. The results show that this supercell is triggered by a cold front. Based on the fact that the observation data has the forecasting potential for the occurrence of severe convective weather, the triggering mechanism and development process of supercell is analyzed by using multi-source data. Vertical wind shear and the change of warm and cold advection in vertical direction based on wind profile radar can help forecasters understand the change characteristics of mesoscale environment before the occurrence of strong convection.S-band dual polarization radar can effectively identify the types and characteristics of severe convection. The simulation of different stages of the occurrence and development of supercell by Variational Doppler Radar Analysis System (VDRAS for short)  shows that the collision between thunderstorm outflow and ambient warm and humid air leads to the rapid development of supercell and the right shift feature.The analysis of vdras data can effectively judge the characteristics of thunderstorm structure and development stage, which provides an important reference for the future occurrence and development trend of thunderstorm. The increase time of lightning frequency detected by lightning positioning system is advanced to the occurrence time of severe convective weather, with the advance is 7 min to 58min. The joint application of the above multi-source data has a good guiding significance for the nowcasting and early warning of sudden severe convective weather.

How to cite: He, N., Yang, Y., Tian, Y., and Liu, K.: Integrated Analysis of Multi-Source Observations during a Supercell Storm Event in Beijing, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-221, https://doi.org/10.5194/ems2026-221, 2026.

Understanding the spatial organization and transitions of convective cloud systems over complex terrain remains a major challenge for both observation-based analysis and numerical weather prediction (NWP). This limitation affects our ability to accurately represent and predict extreme precipitation and hail events, which are expected to intensify over the Alpine region under climate change.

In this work, we present a data-driven framework based on self-supervised learning (SSL) to characterize convective cloud structures and their transitions using long-term geostationary satellite data records. By learning representations directly from infrared imagery without manual labeling, the model organizes cloud scenes based on their spatial and morphological similarities, providing a physically interpretable feature space of mesoscale cloud features. This learned representation is exploited in two complementary ways. 

First, it enables the analysis of convective cloud structures and their transitions associated with severe precipitation events by identifying dominant pathways linked to intense rainfall and hail over the Alpine region. The most severe precipitation is predominantly observed either in association with long-lasting deep convective systems or during transitions from early-stage convection to a more mature regime, highlighting two key modes linked to extreme precipitation.

Second, the framework provides a novel tool for evaluating convection-permitting NWP simulations. By projecting synthetic satellite channels into the same feature space, it enables a direct comparison between modeled and observed cloud structures, both in terms of continuous embeddings and discrete cloud classes. Preliminary results indicate that ICON tends to overestimate the spatial extent of convective cells during their early development stages, suggesting biases in the representation of convective initiation and growth. 

Overall, this approach demonstrates the potential of self-supervised learning to bridge satellite observations and model simulations, providing a flexible framework for studying convective processes and assessing their representation in high-resolution weather models over complex terrain.

How to cite: Corradini, D., Cattani, E., Chatterjee, D., and Acquistapace, C.: Learning Convective Cloud Structures from Satellite Data for Severe Weather Analysis and Model Evaluation over the Alps, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-176, https://doi.org/10.5194/ems2026-176, 2026.

Mesoscale convective systems (MCSs) experience complex changes in intensity as they move from land to the ocean. Predicting the offshore evolution of these systems poses a significant forecasting challenge. Using radar observations and ERA5 reanalysis data from the warm seasons of 2016–2018, this study investigates the characteristics and environmental controlling factors of land-to-sea (LS) MCSs over the Yangtze-Huai River Basin (YHRB). These systems are classified into offshore-intensifying and offshore-weakening types, accounting for approximately 40% and 60% of the total cases, respectively. Spatiotemporally, intensifying MCSs predominantly occur in spring (April–May; 56%) at lower latitudes, whereas weakening MCSs peak in summer (June–July; 45%) at higher latitudes. Composite analysis reveals that intensifying MCSs occur in environments with weaker thermodynamic instability but stronger dynamic forcing. Specifically, most unstable convective available potential energy (MUCAPE) is ~43% lower with reduced total column water (TCW), yet they are characterized by enhanced upper- and lower-level jets and greater vertical wind shear (VWS). A critical factor facilitating intensification is the cooler marine atmospheric boundary layer (MABL), which promotes enhanced isentropic lifting and lowers the lifting condensation level (LCL) via increased relative humidity. Crucially, offshore intensity evolution is governed by environmental change rather than absolute coastal conditions. Weakening MCSs originate in thermodynamically and dynamically superior environments over land but undergo considerable deterioration in thermodynamic and moisture conditions (MUCAPE drops ~47%, TCW decreases) upon moving offshore. Conversely, intensifying MCSs experience modest MUCAPE reduction (~26%) but benefit from significant enhancement of dynamic conditions (low-level VWS strengthening) and moisture enhancement (TCW increases in 60% of cases) during transit, enabling their intensification over the ocean despite lower absolute instability at the coast. These findings provide crucial insights into how coastal environments alter storm intensity, and offer practical insights to improve offshore weather forecasting.

How to cite: Wei, Y. and Xu, X.: Offshore Transition of Mesoscale Convective Systems over the Yangtze-Huai River Basin: Intensity Change and Environmental Controls, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-121, https://doi.org/10.5194/ems2026-121, 2026.

Extreme convective events are increasingly challenging aviation operations, particularly in vulnerable regions such as the central Mediterranean. This study investigates the atmospheric conditions associated with the severe late-autumn thunderstorm that disrupted operations at Rome–Fiumicino Airport on 13 December 2024. Using a conditional analogue-based framework applied to ERA5 reanalysis data, we identify past events characterized by similar synoptic and mesoscale circulation patterns over the satellite era (1979–present), and compare their associated environmental conditions between a historical baseline (1979–2000) and a recent period (2002–2023).

Our results suggest that the large-scale dynamical structure of circulation patterns has broadly remained unchanged over time. However, the thermodynamic background has undergone significant changes, marked by increased near-surface air and warmer sea surface temperatures across the Mediterranean basin. These shifts favor low-level moisture availability and atmospheric instability, leading to more favorable conditions for intense convection. Correspondingly, similar events in the recent period are associated with increased precipitation, stronger wind shear, and higher convective potential. Furthermore, observations at Rome–Fiumicino Airport reveal statistically significant increases in mean wind speed and maximum wind gusts, along with higher dew-point temperatures and reduced visibility in recent decades. These factors are critical for aviation safety, as they can affect aircraft handling and increase the likelihood of operational disruptions. Finally, the absence of significant changes in large-scale climate variability modes suggests that these trends are primarily driven by gradual thermodynamic warming rather than internal variability.
Overall, the findings support the hypothesis that climate change is amplifying the impacts of convective events by intensifying the environmental conditions in which they develop, even when the underlying circulation patterns remain similar. This has important implications for aviation risk assessment and for understanding the evolving nature of late-season convection in the Mediterranean region.

How to cite: Alberti, T. and Faranda, D.: Mediterranean warming amplified the convective environment during the December 2024 severe thunderstorm at Rome-Fiumicino Airport, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-125, https://doi.org/10.5194/ems2026-125, 2026.

Bow echoes, identifiable by their distinct bow-like shape on radar, are often associated with intense specific convective processes. The T-initiation Mesoscale Convective System (MCS) exemplifies this phenomenon, appearing as a quasi-linear bow echo with downstream nascent leading convection (LC), forming a characteristic 'T' shape on radar displays. The LC presents significant challenges for operational forecasts and can lead to severe weather conditions, including damaging straight-line winds and extreme precipitation, particularly when they rapidly merge with the parent bow echoes. Focusing on a high-impact T-initiation MCS over the Bohai Sea in China, this study investigates the underlying physical mechanisms governing LC initiation in a coastal environment. Utilizing high-resolution, cloud-resolving simulations from the WRF-ARW model, backward trajectory tracking within a Lagrangian framework and detailed vertical momentum budget analysis are employed to explore the dynamic factors driving the uplift of air parcels within the LC cells from their pre-convective state. Results reveal that a mid-level mesoscale anticyclonic vortex, associated with the bow echo’s upper-level outflow, significantly modulated the local environmental wind field. This resulted in stronger vertical wind shear between 2.5–5 km AGL downstream of the bow-echo. Interactions between this locally enhanced shear and pre-existing vertical motion led to low perturbation pressures at midlevels. This effect subsequently generated an upward-directed nonhydrostatic vertical perturbation pressure gradient force that aided parcel ascent, directly contributing to the initiation of new LC cells. Thus, this study helps elucidate the complex self-organizing and sustaining processes of bow echoes by identifying a novel, dynamically driven initiation mechanism for LC cells in a T-initiation MCS.

How to cite: Huang, L., Zhang, S., and Xu, X.: Dynamics Governing a T-initiation Mesoscale Convective System, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-166, https://doi.org/10.5194/ems2026-166, 2026.

Intense extratropical cyclones contain a range of mesoscale features capable of producing hazardous surface wind gusts. Well‑established examples include the cold conveyor belt, the evocatively named sting jet, and lines of locally enhanced convection along narrow cold frontal zones. These features are now routinely represented in convection‑permitting numerical weather prediction models, whose horizontal resolution approaches kilometre scale, and have been extensively studied using both idealised simulations and case‑study analyses.

How to cite: Ramsdale, S., Flack, D., Ackerley, D., Volonte, A., and Gray, S.: The Shallow Convective Zone – enhanced gusts in operational UKV forecasts of developing mid latitude windstorms., EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-700, https://doi.org/10.5194/ems2026-700, 2026.

This talk will focus on the devastating small-scale cyclonic windstorm “Kristin” that hit central Portugal in the early hours of 28^th January 2026. This was reportedly the most costly European windstorm since “Lothar” hit France 26 years ago.

After describing the dynamical setting, including critical juxtaposition of an upper level jet and the surface feature, and the storm‘s morphology, including strong radar-based evidence of a sting jet, the surface impacts in both Portugal and Spain will be highlighted. These included 15 fatalities, Portugal’s biggest ever power outage, with over 5000 km of power lines damaged or destroyed, 15 million trees uprooted or snapped and an estimated €7.2 billion in losses in Iberia.

Next the handling by various forecast models, and reanalysis products, will be elaborated. Due in part to Kristin’s small size there was a strong link between the accuracy of model output and model spatial resolution. But even at higher resolution other problems are apparent, such as the generation of some unrealistic initial conditions in mean sea level pressure via the ECMWF EDA (ensemble of data assimilations). Also there was high uncertainty in cyclone track, such that red warnings issued by the Portuguese Met Service (IPMA) initially covered an area much larger than the area eventually impacted.

Machine learning model output, from ECMWF’s AIFS systems, will be included in discussion, highlighting the extra challenges of using such systems. These include temporal discretization (currently 6h), lack of a gust diagnostic and the potential for noisy output.

The practical implications of all these issues, for forecasting in a life-critical operational setting, will be discussed.

Similarities between storm Kristin and other notable small-scale European windstorms, such as Xola from 2009, will also be touched on.

How to cite: Hewson, T., Prates, F., Gascón, E., Vanniere, B., and Pinto, P.:  Extreme Windstorm “Kristin”: Representation across the model spectrum, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-745, https://doi.org/10.5194/ems2026-745, 2026.

On 29 October 2024, an extreme precipitation event affected eastern Spain, triggered by a cut-off low, with daily rainfall totals exceeding annual climatological values in several locations. The province of Valencia was particularly impacted, where widespread accumulations surpassed 300 mm and local maxima reached up to 771 mm. Given the magnitude of the event, its dynamical and thermodynamical drivers, as well as the potential influence of climate change, have been the subject of extensive investigation. Several media reports and recent studies (Campos et al., 2025) have highlighted the possible role of an upper-level tropospheric moisture plume resembling an atmospheric river (hereafter, AR-like structure) connecting the Mediterranean region with the tropical Atlantic via North Africa, suggesting a potential contribution to the event’s intensification. However, the quantitative contribution of this structure to the observed precipitation remains unclear.

To address this question, we employ the WRF with Water Vapor Tracers (WRF-WVTs) model, which enables tracking moisture from predefined source regions to precipitation while fully resolving the event dynamics. Our results indicate that moisture evaporated from the Mediterranean Sea constitutes the dominant contribution, whereas moisture associated with the AR-like structure accounts for approximately 20-30% of the total precipitation. To further assess the role of this remote moisture transport, we introduce a methodology to quantify its indirect impact via enhanced latent heat release and the associated increase in atmospheric instability. This methodology is further applied to several additional events in the western Mediterranean characterized by the coexistence of a cut-off low and an AR-like structure producing substantial precipitation. Preliminary results indicate that this indirect mechanism is considerably more important than the direct moisture contribution, highlighting the key role of mid-tropospheric moisture supplied by these structures in intensifying such events.

Campos, D. A., Grayson, K., Saurral, R. I., Beyer, S., John, A., Olmo, M., and Doblas-Reyes, F.: The October 2024 Extreme Precipitation Event over Valencia: Storyline Attribution of the Synoptic-Scale Thermodynamic Drivers, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-5929, 2025.

How to cite: Crespo-Otero, A., Insua-Costa, D., and Míguez-Macho, G.: The role of AR-like moisture plumes in the October 2024 Valencia precipitation event and other western Mediterranean extremes, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-523, https://doi.org/10.5194/ems2026-523, 2026.

Thunderstorms with climatologically extreme rainfall rates produced flash flooding in the city of San Diego and adjacent suburban areas to the east during the mid-late morning of 22 January 2024. Local rainfall rates from ALERT and ASOS precipitation gauges between 50 and 75 mm in one hour resulted in rainfall return intervals (based on NOAA Atlas 14) of 500 to 1000 years. This resulted in reported damage to around 1000 homes and businesses, numerous road closures and nearly 200 swift-water rescues. An atmospheric river along and just south of the US-Mexico border, associated with a jet stream which circumnavigated the entire North Pacific had estimated synoptic-scale integrated water vapor transport (IVT) reaching the coast of 400 kg m^-1s^-1. However, a southerly low-level jet over San Diego with 850 hPa winds of nearly 20 ms^-1 and associated with precipitable water of 28-30 mm brought locally higher IVT of over 600 kg m^-1s^-1. The low-level jet also combined with west to southwest winds of similar speeds at 700 hPa to bring strong shear in the unstable environment. Thunderstorms trained along an approximate west to east line from Coronado through southern San Diego to Spring Valley, maximizing rainfall amounts locally to 100 mm during a 3-hour period.

While details of an extreme precipitation event like this are difficult to predict, a forecaster can detect and communicate the possibility of this occurring. For this event, high-resolution models, including the HRRR and a WRF-ARW model run locally by the National Weather Service in San Diego, showed local hourly rainfall amounts of 50 mm or more in one hour up to 36 hours in advance of the rainfall, though not in the exact location. Ensemble models also showed a high probability of heavy rainfall rates. As a result, forecasters at the National Weather Service Forecast Office in San Diego were able to issue a flood watch the day before the event and produce messaging for emergency managers and other National Weather Service partners mentioning a risk of heavy rain five days before the event.

How to cite: Maxwell, B.: Extreme Rainfall Rates and Flash Flooding in San Diego, California, USA on 22 January 2024, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-255, https://doi.org/10.5194/ems2026-255, 2026.

Accurate precipitation data plays a central role in ensuring the reliability of flood prediction, especially in catchments where observation networks are sparse. When relying on a single measurement source such as rain gauges, weather radar or satellite products surface runoff predictions often become less reliable. This limitation largely arises from topographic effects such as radar beam blockages as well as the inherent constraints of each sensor in capturing extreme rainfall events.

To address these challenges, this study proposes a multisensor precipitation data fusion framework based on the Random Forest machine learning algorithm with the aim of improving flood prediction lead times in the Busan city, South Korea. The approach combines weather radar observations, GPM IMERG satellite estimates, and ground-based gauge data along with topographic information derived from a Digital Elevation Model.

The resulting fused precipitation product was evaluated using a range of statistical metrics and further tested within a hydrological modeling framework. The results show that integrating remote sensing data with ground observations can substantially improve predictive performance. In particular, the Random Forest model demonstrates strong capability in capturing complex nonlinear relationships and in reducing orographic biases, leading to more accurate spatial rainfall estimates.

When applied to hydrological simulations, the multisensor framework enables earlier detection of upstream rainfall signals, which in turn supports longer and more reliable evacuation lead times. This improvement is also reflected in more stable simulations of hydrograph peak timing. Overall, the findings highlight the practical value of multisensor data integration for supporting more informed and timely decision-making in flood risk management, particularly in regions with complex terrain.

Acknowledgments: The research for this paper was carried out under the KICT Research Program (Project no. 20260161–001, Development of Digital Urban Flood Control Technology for the Realization of Flood Safety City) funded by the Ministry of Science and ICT.

How to cite: Nugraha, B., Seong-Sim, Y., and Dong Sop, R.: Multisensor Precipitation Fusion Based on Random Forest for Enhancing Flood Prediction Lead Time, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-437, https://doi.org/10.5194/ems2026-437, 2026.

Wave-flow interactions within tropical cyclones are closely linked to their structural evolution and intensity variability. Using ERA5 reanalysis data and Tropical Regional Assimilation Model for the South China Sea (TRAMS) numerical simulation data, this study adopts wave-flow separation and Fourier decomposition techniques to investigate the interaction between the basic-state vortex  and asymmetric disturbance of Typhoon Hato. Cross-scale energy transfer processes are analyzed via energy spectra and energy budget analyses. The results are shown as follows.

Typhoon Hato intensified primarily due to strong southwesterly and southeasterly moisture convergence, which advected water vapor toward the inner core and facilitated upward vertical transport, sustaining condensation latent heat release and the subsequent intensifying. Vigorous updrafts created favorable dynamic conditions for latent heat release during the rapid intensification (RI) stage, whereas downdrafts dominated the stagnation stage, markedly suppressing latent heat release processes.

Kinetic energy transfer within Typhoon Hato is closely coupled to its intensity evolution and exhibits distinct radial heterogeneity. During the RI stage, a forward (positive) energy cascade from the symmetric basic-state vortex to asymmetric disturbances dominates in the lower troposphere (~1.5 km), while an inverse (negative) energy cascade from asymmetric scales back to the symmetric vortex prevails in the middle troposphere (~5 km). At the peak of RI, a complete energy transfer chain forms in the upper troposphere (~8 km), with energy sequentially conveyed from the symmetric vortex to outer asymmetric small-scale disturbances and then to spiral rainbands. In the lower stratosphere (~13 km), continuous conversion of asymmetric-scale available potential energy to kinetic energy serves as the dominant process.

The direction of kinetic energy cascade reverses periodically with fluctuations in typhoon intensity, especially at intensity extrema. Forward cascades dominate the RI stage, while inverse cascades prevail during the re-intensification stage. Compared with the inner core region, kinetic energy transfer in the outer region is relatively simpler, characterized by the predominance of asymmetric-scale kinetic energy.This study elucidates the coupled mechanism between abrupt typhoon intensity changes and multi-scale kinetic energy redistribution, in which nonlinear energy transfer at the 8-km level plays a critical regulatory role in overall energy repartitioning.

The Heat redistribution induced by hydrometeor phase transitions and the drag effect associated with hydrometeor sedimentation enhance multi-scale energy transport within the typhoon at the same time, thereby modulating tropical cyclone intensity. These results may provide a sound theoretical basis for understanding the physical mechanism by which hydrometeor phase changes in spiral rainbands trigger asymmetric disturbances and further regulate the overall intensity of tropical cyclones.

How to cite: Huang, H., Wang, J., Li, X., Wang, T., Gu, S., and Jiang, C.: Multi-scale Energy Transfer during the Offshore Intensification of Typhoon Hato, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-315, https://doi.org/10.5194/ems2026-315, 2026.

In this study, numerical experiments with different initial radius of maximum wind (RMW) are performed to study the effects of tropical cyclone (TC) size combined with land-sea contrast on TC motion and low-level wind structure before landfall. By idealized numerical simulations, we found that when TCs approach the coastline, the translation speed of them will accelerate. The larger TC arrived coastline earlier than smaller TC, when they started moving from the same position. This is because that the larger TCs not only accelerate earlier but also have greater movement speed than smaller TCs when they approach the coastline. The mechanism responsible for this is that the edge of large TCs reach coastline earlier, thus their movement speed accelerated earlier than small TCs, due to the asymmetries in diabatic heating and radial flow generated by the land-sea contrast. Moreover, when TCs in three experiments all affected by the land-sea contrast, the stronger asymmetries generated in larger TC, thus resulting in faster movement in larger TC. The stronger inflow in western quadrant and weaker inflow (even outflow) in eastern quadrant of larger TC deduced apparently difference in vertical motion and diabatic heating between western and eastern quadrant of TC before landfall. An analysis of potential vorticity tendency proved that the diabatic heating terms were important and considered in determining the TC landward drift because asymmetries in vertical motion and relative vorticity developed due to asymmetric flow. This achievement can deepen the understanding of the evolution of TC structures in coastal areas and provide certain reference value for the accurate prediction of TC landing time over land.

How to cite: Liu, L. and Wang, H.: The effect of tropical cyclone size on its movement as TC approach the coastline, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-89, https://doi.org/10.5194/ems2026-89, 2026.

    Accurate multi-step typhoon track and intensity prediction remains a critical yet hightly challenging task due to complex nonlinear atmospheric dynamics and strong spatiotemporal variability of typhoon evolution. Traditional statistical methods and numerical modeling approaches often struggle to sustain predictive accuracy, particularly for medium- to long-range forecasts. In this study, we propose a Transformer-based deep learning approach to jointly predict typhoon trajectories and maximum wind speed using multi-source meteorological data. The model is trained on historical best-track of typhoon datasets across China and reanalysis data from 1949 to 2023, incorporating key environmental variables to characterize large-scale atmospheric conditions that modulate typhoon development. All input features are subjected to rigorous quality control, normalization, and temporal alignment, and are structured  sequential samples to capture the temporal evolutionary  dynamics of typhoon systems. The Transformer architecture is adopted to effectively model long-range dependencies and complex spatiotemporal interactions inherent in typhoon evolution.

   To quantitatively evaluate the performance of the proposed approach, two baseline models, XGBoost and Long Short-Term Memory (LSTM), are implemented for comparison. Experimental results demonstrate that the proposed model significantly outperforms both baselines across all evaluated forecast lead times. At short lead times, the model achieves RMSE values of 200–244 km for track prediction, corresponding to a 30% reduction relative to LSTM and a 58% reduction compared to XGBoost. At longer lead times (48h), the model maintains robust and competitive performance, with track forecasting RMSE ranging from 273–480 km, achieving up to 26% lower errors than LSTM and 52% reduction relative to XGBoost. For intensity prediction, the model also shows strong performance, with an MAE of 2.37 m/s, RMSE of 3.12 m/s, and R² of 0.68, indicating strong agreement with best-track observational records.

    Overall, our results confirm that the proposed framework effectively captures the complex spatiotemporal dynamics of typhoon systems, providing marked improvements in predictive accuracy, stability, and robustness. This framework holds strong application potential for operational typhoon forecasting and typhoon-induced disaster risk management.

How to cite: Lin, Z., Wang, Y., Li, W., Ma, H., and Zhang, G.: An improved Transformer-based deep learning approach enhances typhoon track and intensity prediction across China , EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-459, https://doi.org/10.5194/ems2026-459, 2026.

Mesoscale coherent vortex structures (CVSs), including polar lows and tropical cyclones, play a central role in generating high-impact weather such as heavy precipitation, strong winds, and extreme air–sea fluxes. However, a consistent and physically interpretable climatology of these systems across different vortex types and regions remains limited, constraining our ability to assess associated atmospheric hazards.

This study presents a unified framework for identifying, tracking, and clustering mesoscale CVSs in the atmosphere. The analysis uses high-resolution numerical simulations over the North Atlantic and the Arctic. Vortex structures are identified with the Eulerian Rortex criterion, which isolates the rigid-body rotation component of the flow in three-dimensional velocity fields. Identified vortices are then tracked with a self-developed algorithm and subsequently classified by K-means clustering based on a minimal set of physically motivated features describing vortex dynamics, thermodynamics, and geometry.

First, to ensure the framework's physical interpretability, clustering is applied to reference sets of vortex tracks manually identified for 2010 (North Atlantic) and 2019 (Arctic) using the same Rortex-based detection. These tracks are linked to known vortex types using existing best-track datasets for tropical cyclones and polar mesocyclones, providing a physically grounded reference for the most intense and best-studied systems. Clustering the manual tracks enables evaluation of how known vortex types are represented in feature space, assessment of class separability, and identification of additional vortex types with distinct dynamical and thermodynamic characteristics. The resulting classes are further analyzed using a range of physically relevant metrics — including intensity, lifetime, and associated surface heat fluxes — allowing comparison of their relative contributions to hazardous weather and air–sea interaction processes.

The validated feature space and clustering strategy are then applied to a 40-year automatically tracked dataset, enabling a consistent large-scale climatological analysis. The results reveal changes in the relative occurrence and dominance of different vortex types over the past four decades, suggesting links to large-scale circulation variability and ongoing climate change. This approach improves the physical interpretability and robustness of mesoscale vortex climatologies and supports more reliable assessment of associated atmospheric hazards.

How to cite: Koshkina, V., Gavrikov, A., Ezhova, E., Nikitenko-Valiakhmetov, M., Kuznetsov, G., and Gulev, S.: Automatic identification, tracking, and classification of mesoscale atmospheric vortices in high-resolution numerical model data, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-719, https://doi.org/10.5194/ems2026-719, 2026.

The Northeast China Cold Vortex (NCCV) is the primary synoptic system causing severe weather in North China and more than 70% of convective events associated with the NCCV occur in its southeast and southwest sides; however, forecasting first convection initiation (first CI) in its rear side remains highly challenging. This study investigates the underlying CI triggering mechanisms and the causes of low forecast skill in this regime. Using 2016–2021 mosaic reflectivity from seven operational Doppler radars and a modified convective tracking algorithm, we map first-CI distributions and evaluate Numerical Weather Prediction model (NWP) simulations to pinpoint low-forecasting-skill regions (the plain region between Beijing and Tianjin). A representative case (8 June 2019) is then diagnosed using a high-resolution rapid-update (3-km, 12-min) analysis assimilating radar and dense-surface-network observations.

Results show that CI is driven by complex interactions among a cold, dry northerly flow, a warm, moist southerly flow, and a modified sea-breeze front (SBF). Near 850 hPa, the converging northerly and southerly flows generate localized weak convergence and high humidity. Simultaneously, the topographically deflected, inland-propagating SBF forces intense near-surface convergence with the prevailing southerly flow. This combined dynamic forcing establishes a deep updraft channel that successfully initiates convection.

Diagnostics reveal that NWP forecast failures primarily stem from inaccuracies in simulating nocturnal boundary layer wind fields. Although the model similarly captures synoptic-scale forcing and the moisture environment, it produces an anomalously strong low-layer northerly flow. This overestimation blocks the inland penetration of the SBF, thereby eliminating the critical near-surface convergence zone. Absent this essential dynamic trigger, accumulated convective instability cannot be released, resulting in missed CI forecasts. This study addresses the critical challenge of forecasting CI events associated with the R-NCCV, which are notoriously difficult to predict due to their weak synoptic forcing. We have identified the key multiscale factors of a representative event and elucidated how their misrepresentation compromises prediction accuracy.

How to cite: xiao, X.: Key Multi-scale Factors Determining Convection Initiation and Forecast Skill on the Rear Side of the Northeast China Cold Vortex, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-98, https://doi.org/10.5194/ems2026-98, 2026.