Conference Agenda

Multi-­ year droughts pose a significant threat to the security of water resources, putting stress on the resilience of hydrological, ecological and socioeconomic systems. Motivated by the recent multi-­ year drought that affected Southwestern Europe and Italy from 2021 to 2023, here we utilise two indices—the Standardised Precipitation Evapotranspiration Index (SPEI) and the Standardised Precipitation Index (SPI)—to quantify the temporal evolution of the percentage of Italian territory experiencing drought conditions in the period 1901–2023 and to identify Widespread Multi-­ Year Drought (WMYD) events, defined as multi-year droughts affecting at least 30% of Italy. Seven WMYD events are identified using two different precipitation datasets: 1921–1922, 1942–1944, 1945–1946, 2006–2008, 2011–2013, 2017–2018 and 2021–2023. Correlation analysis between the time series of Italian drought areas and atmospheric circulation indicates that the onset and spread of droughts in Italy are related to specific phases of the winter North Atlantic Oscillation (NAO), the Scandinavian Pattern (SCAND), East Atlantic/Western Russia (EAWR) pattern and the summer East Atlantic (EA) and East Atlantic/Western Russia (EAWR) patterns. Event-­ based analysis of these drought episodes reveals a variety of atmospheric patterns and combinations of the four teleconnection modes that contribute to persistently dry conditions in Italy during both winter and summer. This study offers new insights into the identification and understanding of the meteorological drivers of Italian WMYD events and serves as a first step toward a better understanding of the impacts of anthropogenic climate change on them.

Coastal lagoons, such as the Venice Lagoon, are ecologically significant yet highly vulnerable ecosystems facing increasing pressure from climate change. Accurate projections of key hydrographic variables—such as water temperature and salinity—are crucial for developing effective adaptation and management strategies. However, traditional process-based hydrodynamic models, while physically robust, are often computationally prohibitive for performing the long-term, large-ensemble simulations required for comprehensive climate impact assessments. This study addresses this challenge by introducing a novel data-driven framework that leverages a Convolutional Neural Network (CNN) to efficiently simulate and project monthly temperature and salinity dynamics at key locations within the Venice Lagoon, representing distinct marine, riverine, and intermediate regimes. The core of the methodology is a CNN architecture specifically designed to capture the complex, non-linear relationships between large-scale environmental drivers and localized lagoon responses. A major methodological hurdle was the limited observational dataset, comprising only four years of irregular, approximately monthly measurements. To overcome this data scarcity, we implemented a non-standard training protocol based on a sequential optimization strategy, which enhanced the model’s ability to learn robust dependencies and generalize effectively. The model was trained using a minimal yet physically meaningful set of predictors: 2 m air temperature, precipitation, offshore sea level, and offshore sea surface salinity. For future projections, the validated CNN was forced with a set of synthetic climate scenarios representing global warming levels (GWLs) of 1.5, 2.0, and 3.0 °C relative to pre-industrial conditions. These scenarios were constructed by perturbing the historical driver data according to established climate sensitivities for the Mediterranean region. The results demonstrate the framework’s high predictive accuracy, with the CNN successfully reproducing historical observations (R-squared > 0.96 for temperature; R-squared> 0.85 for salinity). Sensitivity analyses confirmed that the model learned physically plausible dynamics, correctly identifying atmospheric forcing as the primary driver for temperature and recognizing the distinct roles of oceanic exchange and terrestrial freshwater input in controlling salinity across the lagoon’s spatial gradient. Projections under the 3.0 °C GWL scenario reveal substantial future changes: lagoon water temperature is projected to increase by up to 6 °C in summer, while salinity is expected to rise by more than 4 psu at the riverine station. These changes are not uniform throughout the year, leading to a pronounced amplification of the annual cycle for both variables and, consequently, to increased seasonal stress on the ecosystem. In conclusion, this work highlights the potential of tailored CNNs as powerful and computationally efficient tools for downscaling climate information and generating actionable projections in complex coastal systems. The proposed framework provides a viable alternative to resource-intensive models and offers critical insights into the future hydrographic evolution of the Venice Lagoon, underscoring the urgent need for climate-resilient management.

Acknowledgments: FB was funded from NBFC – National Biodiversity Future Center, funded by European Union – NextGenerationEU, Project code CN_00000033, CUP F87G22000290

Reference: Bozzeda, F.; Sigovini, M.; Lionello, P. Neural Network Modelling of Temperature and Salinity in the Venice Lagoon. Climate 2025, 13, 189. https://doi.org/10.3390/cli13090189

In recent decades, the frequency and intensity of extreme weather events have increased. Heavy precipitation is of major scientific and societal relevance, yet its analysis is especially challenging due to spatiotemporal variability and multiscale interactions among processes of different nature. In this work, precipitation and its extremes are investigated using data from heterogeneous sources, including weather-station observations, reanalysis products, and outputs from downscaling techniques based on Artificial Intelligence algorithms. In particular, the latter comprise results from new machine-learning methods that employ convolutional architectures for climate downscaling, developed through a collaboration between FBK and ENEA. Designed to provide high spatial resolution forecasts, these methods yielded four historical precipitation datasets that form the basis of the present analysis. This study examines seasonal mean precipitation by comparing these datasets with the ERA5 reanalysis after regridding to a common reference grid and assessing the reproduced spatial patterns. In doing so, it quantifies the differences between the reanalysis and the AI-generated datasets. Finally, return period estimates obtained via extreme-value methods highlight the strengths and limitations of the various datasets relative to the reference reanalysis.

Viene presentato un nuovo dataset operativo climatico di temperature minime e massime giornaliere esteso all'Italia centro-settentrionale per il periodo dal 1991 ad oggi, creato dal Gruppo di lavoro ArCIS (Archivio Climatologico per l’Italia Centro-Settentrionale). Il dataset copre l’area di studio con una griglia regolare di circa 5 km di risoluzione ed è costruito a partire da dati osservativi raccolti e controllati per la qualità dai servizi meteorologici regionali locali. I dati sono stati controllati per qualità e omogeneità statistica. Ai fini dell’interpolazione, il ruolo delle stazioni è stato determinato in base alla loro rappresentatività dal punto di vista termico e in base alla loro appartenenza a serie climatiche storiche; inoltre, il territorio è stato diviso in 28 macroaree, con caratteristiche meteorologiche e geografiche specifiche e, partendo dalle mappe di uso del suolo rese disponibili dal Servizio Corine, sono state create mappe di frazione urbana e una mappa statica dei corpi idrici principali. In ogni macroarea l'interpolazione è stata quindi eseguita prima identificando la dipendenza della temperatura dalla quota, poi identificando la dipendenza lineare dei residui dalla frazione urbana e dalla distanza dai corpi idrici; il prodotto finale è stato completato interpolando i residui sull’intero territorio in base alla distanza tra le stazioni.
Il dataset è descritto mostrando mappe giornaliere per eventi meteorologici specifici, anomalie mensili, descrivendo le caratteristiche dei valori climatologici e delle tendenze annuali e stagionali sul periodo coperto dal dataset. Particolare attenzione sarà dedicata alla descrizione dei valori medi e delle tendenze di specifici indici climatici, costruiti a partire dai dati giornalieri di temperatura minima e massima, che descrivono la frequenza degli eventi estremi nell'area di studio.

An increase in precipitation extremes is one of the most robust aspects of anthropogenic climate change, but the latest assessment of the IPCC still reported low confidence on projected changes in the Mediterranean region. Yet, a number of Mediterranean cyclones, i.e intense mid-latitude storms, have caused considerable precipitation extremes and economic damage in the most recent years, including in Italy. The role played by climate change in these events remain poorly quantified.

In this work we first show that the CERRA regional reanalyses show a significant upward trend (1985-2024) in the number of annual daily precipitation extremes within the Warning Areas of the Italian Civil Protection, and that the upward trend is largely robust to the driving large-scale weather type. To better understand these trends, we then take a storyline approach and by looking at circulation analogs we analyse the response to climate change of selected past high-impact Italian storms, such as storm Vaia and Cyclone Minerva, in large ensembles of regional and global climate model simulations. This enables us to cleanly separate the contribution of internal climate variability from the forced response to climate change. A probabilistic framework is introduced to isolate the role of changes in the large-scale atmospheric circulation, cyclone-development and precipitation intensity in the risk of precipitation extremes. Results show a clear increase in the risk of intense cyclone-associated Italian precipitation extremes, though internal variability is large, and it can mask the climate change signal at individual grid points in single climate realisations. We conclude suggesting new storyline-based and statistical approaches that might help to generalise the results to the different weather types that cause Italian precipitation extremes.

The atmospheric monitoring station at Plateau Rosa, situated in in the north-western Italian Alps near Mt. Cervino, is part of the WMO/GAW (World Meteorological Organisation/Global Atmospheric Watch, Identification Code: PRS) program since 1989 and part of the ICOS (Integrated Carbon Observation System) framework since 2021. At the station carbon dioxide (CO₂) and methane (CH₄) mole fractions have been measured since 2018 with a cavity ring down spectrometer (Picarro G2301). Concentration measurements at this site, 3480 meter AMSL, are particularly valuable for tracking the atmospheric background and global trend of greenhouse gases but are also impacted by various source areas in Europe.

In this study, we analyzed the seven years (2018-2024) record of CO₂ and CH₄ mole fractions at the station. We focused on the past five years, since the station has been part of the ICOS network, to analyse periods of enhanced CO₂ and CH₄ levels over the background that are associated with pollution events at regional scale. We identified 30 pollution events, when air masses were coming mainly from the Po Valley and central Europe. We used the FLEXPART atmospheric transport model coupled to the high resolution (1 km x 1 km) output of the numerical weather prediction model COSMO to produce concentration footprints and simulate regional CO₂ and CH₄ contributions. We assessed how well this transport model, coupled with different surface fluxes (EDGAR, CAMS), captures the selected pollution events and reproduces the continuous CO₂ and CH₄ record at the station.

We finally demonstrate how CO₂ and CH₄ mole fraction data measured continuously at the station at Plateau Rosa can be used to attribute pollution events to specific regional source areas in Europe that might not be accounted by the inventories.