Flash floods are one of the most devastating natural hazards. Every year, they cost thousands of lives and millions of dollars in damaged infrastructure. They can occur in large or small catchments, rural or urban areas, close or away from rivers, and with little to no warning. Some regions might have adapted to protect infrastructure and people against this hazard; however, with climate projections suggesting that extreme rainfall might increase in intensity and frequency, "residual risk" might increase in protected areas while unprotected ones might experience unseen severe losses. Hence, relying on forecasts that offer good predictions of areas at risk of flash floods, with enough lead time to extend preparedness and action time windows, is becoming increasingly important.
This presentation will show the most recent developments in data-driven prediction of areas at risk of flash floods, over a continuous global domain and up to one week ahead. The method uses global reanalysis and medium-range post-processed rainfall forecasts to improve the detection of extreme localised rainfall events. It then tests different machine learning algorithms to learn the complex, non-linear relationships between hydro-meteorological parameters to determine the areas at risk of flash floods. The presentation will also focus on cross-validation, hyperparameter tuning, and ensemble approaches to address the issues that arose due to the severely imbalanced dataset we had to work with.
We will finally explore the added value of these data-driven forecasts and reflect on what this all might mean for decision-makers to extend their preparedness and action time window when the next low-probability, high-impact flash flood event strikes.

How to cite: Pillosu, F., Claire, M., Pappenberger, F., Prudhomme, C., and Cloke, H.: Outrunning flash floods: improving forecasts for better preparedness, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-676, https://doi.org/10.5194/ems2025-676, 2025.

The increasing frequency and devastating impacts of compound hydro-meteorological extreme events underscore the urgent need for a deeper understanding of their dynamics and occurrence. Compound events involve the simultaneous or sequential occurrence of multiple natural hazards or drivers which may exacerbate impacts compared to individual events alone. Recent striking cases is the exceptional sequence of heavy rainfalls that struck northern Italy in 2023-2024, culminating in widespread flooding across the Emilia Romagna region. The severity and extent of the floods were amplified by antecedent precipitation, which saturated the soil, a pre-condition that resulted in substantial aggravation of the impacts. However, the rarity and unprecedented nature of such events pose major challenges for their robust characterization using observational records alone due to the limited temporal coverage, which can introduce substantial uncertainties. To address this, the UNSEEN (Unprecedented Simulated Extremes using ENsembles) approach has emerged as a powerful method, leveraging large ensembles of seasonal re-forecasts from numerical weather prediction models. By pooling ensemble members, this approach effectively generates surrogate time series spanning thousands of years, enabling the exploration of low-probability, high-impact events within a statistically robust framework.
This study applies the UNSEEN methodology to compound flood events within a multivariate framework, focusing on the coupling between precipitation and soil moisture as a key preconditioning driver. The seasonal re-forecasts from SEAS5 (ECMWF) dataset for the period 1994-2023 are considered to characterize unprecedented compound extreme severe floodings in northern Italy, with particular focus on the Emilia-Romagna region. The ability of the UNSEEN ensemble to represent univariate extremes is firstly evaluated—an essential preliminary step to ensure the reliability of the pooled surrogate time series. Subsequently, an event coincidence analysis is conducted on the surrogate series of precipitation-soil moisture extremes to investigate the dominant spatio-temporal patterns arising from their interaction. Results show that the UNSEEN ensemble realistically captures the occurrence of historical extreme flood events, offering a more robust representation compared to observational climatologies alone. These findings underscore the potential of ensemble-based modeling approaches to improve our understanding of rare compound events and support the development of more effective adaptation and mitigation strategies in flood-prone regions.

How to cite: Giordani, A., Bianco, E., Ruggieri, P., and Di Sabatino, S.: Characterizing compound floods in Italy by pooling precipitation and soil moisture seasonal ensemble re-forecasts, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-580, https://doi.org/10.5194/ems2025-580, 2025.

Soil moisture-precipitation feedback is an important factor in the water and energy cycles. But how important is it on the time scale of an atmospheric extreme event? We are investigating this question using the example of heavy precipitation in July 2021, which led to destructive flash floods in Western Europe.

To quantify the importance of land-atmosphere coupling and continental moisture sources for the precipitation, we perform numerical simulations with different levels of soil moisture. In one set of simulations, we prescribe the initial conditions only, in another set of simulations, we constantly force the soil moisture to specific values. In this way, we can distinguish between the effect of water availability at the surface and the role of the land-atmosphere coupling. Furthermore, we add moisture tracking to our analysis to see if the modifications also impact the spatial distribution of moisture source regions.

Ensembles of simulations are performed using a global set-up of the ICON numerical weather prediction model. This allows the system to evolve without prescribed lateral boundary conditions. However, the predictability of the event then depends exclusively on the initial conditions. In order for ICON to predict the extreme event, an initial state close in time to the event is required. However, this time may be too short for the moisture changes to take effect. For this reason, we are developing a modified simulation set-up, which uses data assimilation to constrain the evolution of the system into an extreme just enough so that our modifications to the soil are not undone.

Our work is part of the German Research Foundation (DFG) Collaborative Research Center 1502 DETECT. In DETECT we aim to answer the question of whether regional changes in land and water use impact the onset and evolution of extreme events. Our coarse approach to changes in water availability gives us a reference of the changes we can expect as a result of human influence.

How to cite: Fohrmann, T., Valmassoi, A., and Friederichs, P.: The influence of soil moisture on the extreme precipitation event in July 2021 in Western Europe – A storyline approach, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-238, https://doi.org/10.5194/ems2025-238, 2025.

In southeastern Austria, the summer season is often characterized by the occurrence of heavy thunderstorms. Typical for these rainfall events is their rapid development, with only a few minutes to hours between the formation of the first clouds and the end of the event. The intense precipitation during these events often results in severe damage, but is difficult to predict. A profound understanding of the life cycle of such events, from formation to dissipation, is therefore crucial to increasing natural hazard resilience and improving forecasting skills.

Here, we investigate the life cycle of 94 heavy rainfall events in high-resolution observational data provided by the WegenerNet 3D Open-Air Laboratory for Climate Change Research (WEGN3D Open-Air Lab) located around Feldbach, Austria. With its 156 ground stations, one X-band radar, two radiometers, and six Global Navigation Satellite System (GNSS) stations, the WEGN3D Open-Air Lab provides high-resolution observations of key atmospheric parameters. By tracking changes in 10 atmospheric parameters connected to heavy precipitation, we gain insights into characteristic features of the different stages of the precipitation life cycle of small-scale rainfall events.

Beginning with the event formation stage (i.e., the 8 h before the event), we find distinct patterns in air temperature, integrated water vapor, liquid water path, and wind speed that are directly linked to the formation of the first storm clouds. During the precipitation stage the highly localized character of these events is clearly visible in the spatial variability of temperature, liquid water path, and cloud cover. In the 16 h after the event (i.e., dissipation stage), we observe the slow return of the atmospheric parameters to pre-event conditions.

Besides being well in-line with the expected physical processes connected to small-scale rainfall extremes, our findings also show that the WEGN3D Open-Air Lab is very skilled in monitoring heavy rainfall events and their characteristics in high spatial and temporal resolution. This illustrates the dataset’s high potential for applications in the improvement and verification of weather and climate models.

How to cite: Haas, S., Kvas, A., and Fuchsberger, J.: Precipitation life cycle analysis of heavy rainfall events in high-resolution observational data in the southeastern Alpine forelands, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-215, https://doi.org/10.5194/ems2025-215, 2025.

In the Mediterranean region the presence of warm sea surface temperatures and of complex topography favour per se the development of deep convection. This study shows how the presence of Mediterranean cyclones (MEDCs) further enhances the frequency of thunderstorms and how the synoptic and mesoscale features around the cyclones’ low-pressure centres organise the distribution of convective environments. The results are based on ERA5 reanalysis environmental variables, hail and lightning probabilities (modeled from ERA5 predictors) and a lightning detection dataset called ATDNet. Furthermore, a recent classification of MEDCs into nine clusters based on upper-level dynamical structure (Givon et al., 2024) serves as a framework for assessing the relationship between cyclone type and convection.

For each MEDC cluster, we characterise the frequency, intensity, spatial distribution and time evolution of convective environments and hazards. Convective activity typically develops to the northeast of the cyclone centre and within the warm sector, peaking before the cyclone reaches its minimum central pressure. Among the various cyclone types, small and deep systems occurring during autumn in the Northern Mediterranean exhibit the highest potential for the development of severe convection, followed by weaker cyclone systems propagating mainly in the Southern Mediterranean during transition seasons and summer. We further examine feature objects that correspond with different dynamical processes within the cyclones, finding that regions of warm conveyor belt ascent are more strongly associated with deep convection than cold frontal zones. The pattern holds across all cyclone clusters. These findings advance the understanding of mesoscale processes associated with MEDCs and offer useful insights for improving operational weather forecasting and risk communication regarding MEDC-related hazards.

Givon, Y., Hess, O., Flaounas, E., Catto, J. L., Sprenger, M., and Raveh-Rubin, S.: Process-based classification of Mediterranean cyclones using potential vorticity, Weather Clim. Dynam., 5, 133–162, https://doi.org/10.5194/wcd-5-133-2024, 2024.

How to cite: Portal, A., Angelidou, A., Rousseau-Rizzi, R., Raveh-Rubin, S., Givon, Y., Catto, J. L., Battaglioli, F., Taszarek, M., Flaounas, E., and Martius, O.: Convective environments and hazards in Mediterranean cyclones, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-664, https://doi.org/10.5194/ems2025-664, 2025.

Convective initiation (CI) nowcasting in subtropical regions, such as South China, is often hindered by complex atmospheric processes and the imbalanced distribution of CI events, leading to high false alarm ratios (FAR). To address these challenges, this study develops a Storm Warning System with Physics-Augmentation (SWASP), which integrates a random forest (RF) algorithm with cloud physical conditions and auxiliary geospatial information. The model leverages multi-channel data from the Himawari-8 Advanced Himawari Imager (H08/AHI) during the warm seasons (April to September) of 2019.

The SWASP model incorporates critical physical indicators associated with CI triggering mechanisms, including cloud-top cooling rates, cloud-top height relative to the tropopause, and the temporal evolution of cloud-top height. Given that the physical thresholds for convective development vary regionally, we recalculated six threshold criteria based on these physical variables, evaluating multiple percentile-based schemes. The most effective performance was achieved when applying the 85th percentile criterion to define pre-convective cloud conditions, which was then used as input to the RF model. This configuration yielded the highest critical success index (CSI) and the lowest FAR compared to traditional threshold-based methods.

In addition to cloud physical parameters, the model also integrates topographic elevation, satellite zenith angle (SAZ), and latitude (LAT) to account for regional and observational variability. Compared with conventional methods, the SWASP model improves the probability of detection (POD) by 0.11 and 0.08, while reducing FAR by 0.38 and 0.44 during daytime and nighttime, respectively. Moreover, the system demonstrates the capability to detect local convective storm systems approximately 30 minutes to 1 hour ahead of radar-based detection in typical CI cases.

This study highlights the advantage of integrating physical knowledge into machine learning-based nowcasting frameworks and demonstrates the potential of geostationary satellite observations in enabling timely and accurate convective early warnings in subtropical regions.

How to cite: Yuan, H. and Yang, C.: Convective Initiation Nowcasting in South China using Physics-augmented Random Forest Models and Geostationary Satellites, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-291, https://doi.org/10.5194/ems2025-291, 2025.

Convectively forced gravity waves (CGWs) are atmospheric phenomena that influence convection through feedback mechanisms. These waves, generated by convective heating, can either suppress or enhance convection. In this study, we explore the generation mechanisms of CGWs and their impacts on convection using a combination of analytical models and numerical simulations. We introduce a new linear analytical model for CGWs, driven by periodic heating to represent both short-lived and long-lived convection. The model shows that deep convection generates n=1 gravity waves, which stabilize the atmosphere by increasing Convective Inhibition (CIN) and decreasing CAPE, thereby suppressing convection. In contrast, stratiform convection produces n=2 waves, destabilizing the atmosphere and promoting elevated convection by lowering the Level of Free Convection (LFC). We analyze the sensitivity of wave propagation to the vertical and horizontal scales of convective heating. The results indicate that the vertical scale of heating primarily determines wave characteristics, with larger horizontal scales leading to slower propagation speeds and longer wavelengths. The presence of a tropopause further enhances wave propagation by reflecting waves, allowing them to travel longer distances. Numerical simulations of squall lines under vertical wind shear reveal that CGWs contribute to the asymmetric development of squall lines. n=1 waves suppress convection on the downshear side, while n=2 waves enhance cloud formation and destabilize the environment. A real case from South China demonstrates that CGWs generated by frontal rainbands can influence warm-sector convection, increasing low-level humidity and reducing CIN, thus aiding convection initiation. This study enhances our understanding of the interactions between convection and CGWs, with implications for weather prediction, particularly in squall lines and warm-sector rainfall.

How to cite: Du, Y., Yang, H., and Rotunno, R.: Generation of Convectively Forced Gravity Waves and Their Impacts on Convection, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-75, https://doi.org/10.5194/ems2025-75, 2025.

Mesoscale Convective Systems (MCSs), with length scales of 100 to 1000 km or more, fall into the "grey zone" of global models with grid spacings of 10s of km. Their under-resolved nature leads to model deficiencies in representing MCS latent heating, whose vertical structure critically shapes large-scale circulations. To address this challenge, we use analysis increments—the corrections applied by Data Assimilation (DA) to the model's prior state—from a 10 km Met Office operational forecast model to inform the development of a stochastic parameterization for MCS latent heating. To focus on errors in MCS feedback rather than errors due to a missing MCS, we select analysis increments from 1037 MCS tracks that the model successfully captures at the start of the DA cycle.

A Machine Learning–based Gaussian Mixture Model reveals that the vertical structure of temperature analysis increments is probabilistically linked to the atmospheric environment. Bottom-heavy heating increments tend to occur in low Total Column Water Vapor (TCWV) conditions, suggesting that the model underestimates low-level convective heating in relatively dry environments. In contrast, top-heavy heating increments are linked to a moist layer overturning structure—characterized by high TCWV and strong vertical wind shear—indicating model underestimation of upper-level condensate detrainment in such environments. This probabilistic relationship is implemented in the Met Office operational forecast model as part of the MCS: PRIME stochastic scheme, which corrects MCS-related uncertainties during model integration. By enhancing top-heavy heating, the scheme backscatters kinetic energy from the mesoscale to larger scales, improving predictions of Indian seasonal rainfall and the Madden–Julian Oscillation (MJO). Future work will assess its impact on forecast busts and its potential to extend predictability.

How to cite: Zhang, Z., Christensen, H., Plant, R., Tennant, W., Muetzelfeldt, M., Whitall, M., Woollings, T., and Stirling, A.: Data-Driven Stochastic Parameterization of MCS Latent Heating in the Grey Zone, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-406, https://doi.org/10.5194/ems2025-406, 2025.

In recent years, extreme droughts – both on spatial extent and duration – have become more frequent and tend to occur in all continents. Some examples are the consecutive droughts in California in the 2010s and the 2020s, the long-lasting drought in the La Plata Basin in South America in the 2020s, and the pan-European summer drought of 2022. Such droughts, often named mega-droughts, also have multiple impacts over a wide set of sectors, including health, agriculture, forestry, ecosystems, and infrastructures, to the point that they may cause permanent land degradation and disruption of life-needed facilities, as the provision of clean potable water.

In the first part of the study, we collected and ranked the largest drought events from 1901 to 2024, dividing them by country and macro-region, and using a scoring system that accounts for drought extent, severity, intensity, peak, extreme conditions, and other 15 parameters. To do that, we used a combination of climate datasets (ERA5, GPCC, CRU, and Berkeley Earth) and two meteorological drought indicators (the SPI and the SPEI) at multiple accumulation scales. In this presentation, we present the classification system (5-class plus a 0-100 score) and we detail on some specific events.

In the second part, we assigned, to each single event part of our new database, a set of 25 additional parameters regarding the exposure of population, land-use (forests, croplands, pastures, and urban areas), and macro-economic indexes (GDP, GDP per capita, GDP-PPP). We do that to answer a key question: are the drought usually considered the biggest in terms of hazard also the biggest considering the exposure of specific categories? We therefore present another classification scheme to incorporate population, land-use, and GDP exposure to drought events, showing that sometimes the drought hazard is not so effective in capturing how drought affects critical parts of the Earth.

We present new lists of mega-droughts divided by hazard, by exposure of single classes, and eventually all grouped under a single score that accounts for both hazard, and exposure. Regarding population, we also included a special section focusing on the exposure of segments at high risk, the young people (below 5 years old) and the old people (above 69 years old).

How to cite: Spinoni, J., Mastropietro, M., Cammalleri, C., Dosio, A., and Tavoni, M.: Classifying mega-droughts according to population, land-use, and economic exposure, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-426, https://doi.org/10.5194/ems2025-426, 2025.

How to cite: Gargiulo, F., Ruggieri, P., and Famooss Paolini, L.: On the connection between key thermodynamic drivers of heatwave onset and the seasonal prediction of heat extremes, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-405, https://doi.org/10.5194/ems2025-405, 2025.

The Svalbard archipelago in the high Arctic is one of the fastest warming regions on Earth, warming approximately five times faster than the global average. The most dramatic warming in recent decades has been observed in the coldest months, that occur during the winter and early spring. Despite an evident warming trend, the warming in summer has been less pronounced. Still, recent summers have also seen increasingly frequent and extreme warm anomalies, with 2024 being the third consecutive record warm summer. During this summer, both seasonal and monthly summer temperature records were shattered by exceptional margins at various weather stations across the archipelago. We present the atmospheric drivers behind these extreme conditions, with focus on the warmest month in Svalbard’s measurement history: August 2024.

Synoptic analyses, based on ERA5 and ORAS5 reanalysis products and LAGRANTO (Sprenger & Wernli, 2015) back trajectories reveal an unusual persistence of pressure systems and an exceptionally strong and persistent southerly flow. Sea surface temperatures were exceptionally high too, both amplifying and being amplified by the synoptic heat transport. Together, these conditions sustained pronounced warmth over the region far beyond what would be expected from the long-term warming trend alone.

These findings highlight the importance of atmospheric circulation, persistence of weather and ocean–atmosphere interactions in driving extreme Arctic summer temperatures. Accordingly, we recommend research to the potential increase in persistence of summer weather at high latitudes, the extreme events it may cause, and the consequences thereof.

Sprenger, M., & Wernli, H. (2015). The LAGRANTO Lagrangian analysis tool–version 2.0. Geoscientific Model Development, 8(8), 2569-2586.

How to cite: van den Broek, D., Urbancic, G., Rantanen, M., and Vihma, T.: 2024 Summer Heat in Svalbard: Atmospheric & Marine Drivers Behind Extreme Temperatures, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-589, https://doi.org/10.5194/ems2025-589, 2025.

Citizen science has already been used for hazard-based data collection globally including storm, flood and air-pollution. Here we present a citizen science demonstration project of HIWeather (High-Impact Weather project) under WWRP, which aimed on hailstone collection from 2016 to 2024, reviewing its scientific objectives and results, public motivation and participation, challenge and future prospects.

The project was setup to fully understand how aerosols interact with moisture in the growth process of hailstone from 2016. As hailstorm are mesoscale deep convection with high social-impact and low predictability, it is hard for researcher to collect nature hailstone before melting. However, citizen science provide a possible solution in China, considering its dense population. The online platform of WeChat was leveraged to broadcast the collection needs, engage the public, and provide detailed instructions for sample collection. The project has successfully gathered 3,063 freshly fallen hailstones from 15 provinces in China by 100 times of voluntary collections.

Hailstones after local hailstorms are collected by volunteers, and stored properly until delivered for further laboratory analysis. After pre-processing, hailstone samples were analyzed to determine their chemical composition, including water-soluble ions, stable isotopes, and insoluble particles from inner embryos to outer shells. Analysis revealed that the major source of water-soluble ions and insoluble particles in hailstones are likely come from local surface aerosols, and multiple hailstone growth trajectories exist in deep convection. The analysis of the collected samples has yielded valuable information on aerosols and the growth trajectories of hailstones. The results provide insights into hailstone formation processes and highlight the importance of atmospheric chemistry in predicting hailstorm in the future.

To foster mutual benefits and encourage future engagement, follow-up communication with contributors is upheld to express gratitude, acknowledge achievement, and gather feedbacks. Surveys indicate that most responded volunteers (79%) are primarily motivated by contributing to science, with a good scientific literacy through general education and science outreach efforts. Motivation from direct experience of hail damage is notably with 67% respondents reported that. Widespread use of smartphones and social media is crucial to extended information network and booster citizen science project mobilization.

How to cite: Zhang, Q.: Mobilizing Citizen Science for Hailstorm Research: Motivations, Challenges and the Outcomes, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-662, https://doi.org/10.5194/ems2025-662, 2025.

In the early morning of 1 August 2021, a supercell developed over the Veneto plain and moved eastward towards Friuli-Venezia Giulia, locally producing large hailstones with diameters up to 9 cm.

In the present work, this event is studied by means of simulations with the Weather Research and Forecasting (WRF) model at 1 km resolution, coupled with the HAILCAST hail growth parameterization, which provides estimates of the maximum hail size at the ground. Several simulations are performed using different initial and boundary conditions (GFS and IFS forecasts), different initialization times and physics options, to study the predictability of the event.

The analysis of the simulations show a weak predictability of the event overall. In fact,  the results highlights a significant sensitivity to the forcing meteorological model and the initialization time. In particular, WRF is not able to properly simulate the development of strong convection over the Veneto and Friuli-Venezia Giulia plain in the early morning of 1 August using GFS forcing, while better results are obtained with IFS initial and boundary conditions, especially when simulations are initialized more than 24 hours before the event. Moreover, results are significantly affected by the microphysics scheme and the land surface model, while the planetary boundary layer parameterization seems to have a minor influence. However, the development of the supercell is properly simulated (with hailstone diameters comparable to observations) only when data from radiosoundings of Udine Rivolto are nudged into the model, highlighting the importance, and at the same time the complexity, of correctly reproducing local thermodynamic conditions for the simulation of extreme convection events. This study shows the significant impact that radiosonde data nudging can have on convective simulations, and demonstrates the capability of HAILCAST to accurately reproduce large hailstone events.

How to cite: Sioni, F., Perbellini, A., Manzato, A., and Giovannini, L.: Numerical simulations of a supercell in northeastern Italy with WRF-HAILCAST, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-293, https://doi.org/10.5194/ems2025-293, 2025.

Hailstorms are among the most damaging convective weather events globally, leading to significant socioeconomic impacts on infrastructure, agriculture, and property. Effective hail prediction and hindcasting require a robust understanding of the environmental conditions under which hail forms. Synoptic and mesoscale atmospheric patterns play a critical role in convective phenomena such as hail formation, with variations in these patterns being closely linked to hailstorm development. The classification of these patterns is essential for identifying region-specific environmental conditions, which is crucial for optimizing modeling strategies and improving the accuracy of hail predictions. Consequently, hail prediction requires an integrated approach that considers multiscale processes, spanning synoptic-scale conditions, mesoscale characteristics, and convective parameters.

This study proposes a framework that combines atmospheric conditions favourable to hail occurrence, accounting for spatial and seasonal variability in hail frequency and physical drivers such as orographic features with numerical model optimisation to improve region-specific hail simulation.

To this end, we couple the selected hail-prone environments in Europe, categorised into clusters, with a Genetic Algorithm (GA) designed to optimize the configuration of the Weather Research and Forecasting (WRF) model for simulating various hail conditions. The GA systematically evaluates different combinations of WRF physics schemes to identify those most effective at reproducing observed hail events. By applying the derived optimized WRF configurations to representative cases within each cluster, we assess whether different environmental settings require tailored modelling configurations for accurate hail simulation. This integrated approach has the potential to reveal important links between local terrain, synoptic-scale patterns, and model performance.

The integration of this clustering with model optimization offers a scalable and efficient pathway for improving hail simulations. By linking atmospheric conditions for hail formation with optimised WRF configurations, this framework enables more streamlined, region-specific hail simulations and a better understanding of hail formation processes, ultimately enhancing hail prediction capabilities and hail risk assessments.

How to cite: Guerrero-Calzas, I., Baladima, F., Cortés, A., Hanzich, M., and Miró, J. R.: WRF Optimization for Hail Risk: Coupling Environmental Clustering with Genetic Algorithms, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-281, https://doi.org/10.5194/ems2025-281, 2025.

Hailstorms are among the costliest extreme weather events, particularly for sectors like agriculture. However, physically representing hailstorms in models is challenging, as they depend on high-resolution, sub-grid deep convective processes that are both difficult and expensive to simulate. To address this, some approaches based on artificial intelligence, and especially Machine Learning (ML), have recently emerged to bypass some of these limitations of physical models. These approaches usually rely on hailstorm occurrence data from reports as the model’s target. This data is often scarce and exhibits spatial and temporal biases. Although there is a relatively large and consistent database of hail reports in the US, such data is much scarcer in Europe. The European Severe Weather Database contains tens of thousands of reports, but the hail reporting rate has only increased in recent years and remains low in some countries. Since these ML methodologies require large and consistent datasets, it is challenging to construct statistical model based only on European data. In this study, we explore how to build ML classifiers for hailstorm occurrence in data-scarce regions like Europe, using datasets from different regions and applying domain shift adaptation techniques, alongside local data.

We first describe our baseline ML model, which uses meteorological variables from ERA5 reanalysis associated with deep convection, and hailstorm occurrence data from the US. The model is trained to learn the relationship between these variables and hailstorm occurrence, and is then used to infer the probability of such events in locations and days not seen by the model.

Next, we explore how to build a similar hailstorm occurrence classifier for Europe. We compare several approaches: applying the baseline US classifier in Europe, training a separate classifier on Europe’s limited data, and using domain shift adaptation techniques to combine both datasets. We tested direct dataset mixing and a more advanced approach where the model is pre-trained on US data—where it is more abundant—and fine-tuned on Europe’s limited dataset.

We benchmark these different techniques in Europe and the US, offering insights into how to build generalizable ML models for hailstorm occurrence. We also present a calibration method to ensure model output accuracy and show how the model can be used to construct hailstorm hazard maps for use by various stakeholders.

How to cite: Sanchez Mallorquín, A., Cataldi, M., and Rodriguez Pinilla, M.: Domain shift adaptation methodologies for statistical modeling of hailstorm occurrence in Europe, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-152, https://doi.org/10.5194/ems2025-152, 2025.

Accurate and timely lightning nowcasting is critical for industry safety and public safety.  Based on the data from the self-built Very Low Frequency Long-Range Lightning Location Network (VLF-LLN), combined with radar data and Himawari-8 satellite data, our team researched the lightning nowcasting within 0-1 hours in China using a number of deep learning models.

• A new algorithm for the long-range lightning location.The algorithm is used to extract the ground wave peak points of the lightning sferics based on the waveform bank and extracted the arrival time of the ground wave, and the lighting flashes can been located by using TOA (Time of Arrival) combined with the equivalent propagation velocity method, and the detection efficiency of the VLF-LLN was determined to be up to 70.6% with a median location accuracy of 1.814 km.
• Lightning nowcasting based on high-density area and extrapolation only utilizing VLF-LLN data.More current lightning nowcasting is based on data-driven methods, which mainly focus on the optimization of model structure, ignoring the characteristic factors of the data itself, especially the discontinuity in time and the discreteness in space of lightning location data. Therefore, a Gaussian kernel with a radius of 20 km is first used to diffuse the original lightning strike frequency image, and through the weight allocation between adjacent frames, we obtained a sequence of mutually related lightning image frames, and separately used the frame data for lightning recognition and extrapolation. The results show that the accuracy rates of lightning areas extrapolation within the next 6 minutes, 18 minutes, and 36 minutes are 94.1%, 79.4%, and 65.8%, respectively.
• Lightning nowcasting based on Himawari-8 satellite data, radar data, and VLF-LLN data.Based on the deep learning models, we have solved the problems of predicting the first lightning and extrapolation for the lightning nowcasting. We consider time as an additional channel dimension and combine it with recent advances in attention mechanisms, a lightweight model was proposed to obtain better performance index than other baseline models with minimal computational resource, which consists of an encoder and decoder based on two-dimensional convolution and several temporal translators based on one-dimensional convolution. The data include water vapour in the middle troposphere and cloud top temperature from the Himawari-8 satellite, radar combined reflectivity factor, and VLF-LLN lightning data. The results indicate that the radar data significantly improves the hit rate and reduces the false alarm rate, while satellite data mainly reduces the false alarm rate. The average hit rate, false alarm rate and critical success index for lightning nowcasting are 0.5645, 0.3302 and 0.4441, respectively.

How to cite: Song, L., Li, J., Liu, Y., Zhang, Q., Zhang, G., and Gong, Y.: Application of the VLF-LLN Data and Lightning Nowcasting Based on Multi-Source Fusion Model, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-68, https://doi.org/10.5194/ems2025-68, 2025.

How to cite: Vitali, B., Lacavalla, M., and Bonanno, R.: Towards Probabilistic Methods for Forecasting Snow Load on Power Lines in Italy, EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-210, https://doi.org/10.5194/ems2025-210, 2025.

Turbulence is a major aviation hazard responsible for the majority of weather-related aircraft accidents. The recent fatal turbulence event on a commercial flight from London to Singapore underscores the growing risk turbulence poses to aviation safety. As climate change is expected to strengthen upper-level jet streams, turbulence is projected to intensify in both frequency and severity, making its reliable forecasting more important than ever.

Operational turbulence forecasting often relies on the Richardson number (Ri), a classical diagnostic representing the balance between turbulent kinetic energy (TKE) generation by vertical wind shear and its suppression by stable stratification. However, the conventional Ri formulation neglects horizontal shear, which can be a crucial source of turbulence near jet streams, upper-level fronts, and regions of strong horizontal deformation.

In this study, we develop a new Ri formulation based on the full TKE budget that explicitly includes the effects of horizontal shear. We apply this diagnostic to ERA5 reanalysis dataset and evaluate its performance against in situ turbulence observations from commercial aircraft using eddy dissipation rate (EDR) as a reference. We also compare it with the conventional Ri and other widely used turbulence diagnostics. Our results show that the new Ri performs better than the conventional Ri and other indices. Furthermore, combining the new Ri with additional diagnostics leads to significant improvements in upper-level turbulence forecasting. These findings suggest that integrating this refined Ri formulation into operational forecasting systems may likely improve aviation safety, reduce flight delays, and optimise fuel consumption through better route planning.

This work demonstrates the value of revisiting classical diagnostics using a physically complete framework and highlights the importance of refining turbulence forecasting tools in a changing climate where upper-level aviation hazards are likely to become more frequent and intense.

How to cite: Foudad, M., Teixeira, M., and Williams, P.: A New Diagnostic for Improved Forecasting of Aviation Turbulence Hazards , EMS Annual Meeting 2025, Ljubljana, Slovenia, 7–12 Sep 2025, EMS2025-264, https://doi.org/10.5194/ems2025-264, 2025.