Le Giornate dell'Idrologia 2025 della Società Idrologica Italiana: "Affrontare le sfide climatiche e territoriali: scenari, rischi e strategie di adattamento"

Le Giornate dell’Idrologia 2025 Bari, 8-10 settembre Book of Abstract Edito da Vito Iacobellis, Vincenzo Totaro, Elena Toth Data curators Cinzia Albertini, Maria Rosaria Alfio, Virginia Rosa Coletta, Beatrice Lioi, Federica Mesto, Claudia Panciera, Stefania Santoro doi:10.5281/zenodo.17298437 Progetto grafico: Stefania Santoro Si ringraziano Le Giornate dell’Idrologia 2025 Politecnico di Bari 8-10 Settembre 2025 COMITATO ORGANIZZATORE Vito Iacobellis, Vincenzo Totaro (coordinatore), Gabriella Balacco, Andrea Gioia, Alessandro Pagano, Stefania Santoro, Cinzia Albertini, Maria Rosaria Alfio, Virginia Rosa Coletta, Gaetano Fiorese, Aras Izzaddin, Beatrice Lioi, Audrey Maria Noemi Martellotta, Federica Mesto, Claudia Panciera, Pasquale Perrini, Marwah Yaseen COMITATO SCIENTIFICO Elisa Arnone, Günter Blöschl, Brunella Bonaccorso, Giorgio Boni, Martina Bussettini, Luigi Cimorelli, Angela Corina, Elena Cristiano, Davide De Luca, Umberto Fratino, Francesco Gentile, Vito Iacobellis, Christian Massari, Valerio Noto, Marco Peli, Daniele Penna, Maurizio Polemio, Ivan Portoghese, Riccardo Rigon, Ezio Todini, Elena Toth 1 INDICE PREFAZIONE 11 AIDA (Agricultural Irrigation Demand Analyzer): distribuzione e idroesigenza di colture irrigate e non irrigate a 1km di risoluzione per l’Italia Nikolas Galli, Harsh Nanesha, Arianna Tolazzi, Davide Danilo Chiarelli, Maria Cristina Rulli 12 Analisi di scarsità idrica e competizioni intersettoriali per l’uso dell’acqua attraverso dati di domanda idrica ad alta risoluzione Nikolas Galli, Francesco Capone, Camilla Govoni, Harsh Nanesha, Davide Danilo Chiarelli, Maria Cristina Rulli 14 Quantifying the impacts – from paddies to dry rice fields Francesca Padoan, Maria Cristina Rulli, Davide Danilo Chiarelli 15 Modeling the effects of shifting planting dates and irrigation management strategies in maize cropping systems for sustainable water resource management Adrian Chummac, Brunella Bonaccorso 17 On the hydrological sustainability of agricultural areas of the Tigris–Euphrates river basin Hiba Mohammad, Muhammad Faisal Hanif, Marco Peli, Roberto Ranzi, Stefano Barontini 19 Influenza sull’indice di aridità delle formule per la stima dell’evapotraspirazione potenziale Giovanni Selleri, Mattia Neri, Elena Toth 21 A remote sensing-based evapotranspiration dataset for Europe using MSG images Carmelo Cammalleri, Samuele Maffei, Chiara Corbari, Nicola Paciolla, Marco Mancini, Emanuel Dutra, Isabel Trigo, Francoise Meulenberghs, José Miguel Barrios 23 Sensitivity analysis of Evapotranspiration - A comparison of Northern and Southern climate of Italy Komal Jabeen, Anna Palla 25 Modeling evapotranspiration dynamics in southern Italy using a grid-based hydrological approach Htay Htay Aung, Biagio Sileo, Mauro Fiorentino, Silvano F. Dal Sasso 27 Toward better hydrological drought representation: optimizing vegetation dynamics into the NOAH-MP land surface model for Southern Italy Sara Modanesi, Domenico De Santis, Daniela Dalmonech, Alessio Collalti, Francesco Avanzi, Gabrielle De Lannoy, Tommaso Moramarco, Nunzio Romano, Paolo Nasta, Fabio Massimo Grasso, Martina Natali, Christian Massari 29 Driver-based classification identifies snowmelt as the primary cause of severe droughts in the Upper Adige basin Andrea Galletti, Susen Shrestha, Mariapina Castelli, Massimiliano Pittore, Stefano Terzi, Giacomo Bertoldi 31 Multivariate Risk Modeling of Drought Impact on Water Supply in the Camastra Reservoir Mohamed-Amine Lahkim-Bennani, Brunella Bonaccorso 33 2 Piú energia per gli utenti, piú acqua per l’ambiente: scenari di ibridizzazione idroelettrico-fotovoltaico per un invaso pre-alpino Domenico Micocci, Cristiana Bragalli, Elena Toth, Tobias Wechsler, Massimiliano Zappa 35 Assessment of regional-scale freshwater availability and sustainability of anthropogenic withdrawals: the GOV4WATER project Alessia Flammini, Jacopo Dari, Carla Saltalippi, Francesco Leopardi, Martina Natali, Renato Morbidelli 37 A framework for ultra-high resolution satellite-based irrigation monitoring to support water allocation managers Jacopo Dari, Stefano Lo Presti, Luca Brocca 39 Presenting the Optimizing Water Resources in Coastal Areas Using Artificial Intelligence (AI4Water) PRIMA project Stefano Barontini, Muhammad Faisal Hanif, Hiba Mohammad, Paolo Colosio, Marco Peli, Domenico Ventrella, Roberto Ranzi, Ivan Serina 40 Un modello semplificato per lo studio del bilancio idrologico della zona umida costiera di Torre Guaceto (Br) Claudia Panciera, Vito Iacobellis, Alessandro Pagano, Rita Masciale, Giuseppe Passarella, Ivan Portoghese 42 Modellazione idrologica avanzata del bacino dell’Arno per valutare la risposta delle risorse idriche agli scenari di cambiamento climatico Francesco Bressi, Stefano Tasin, Marco Lompi, Enrica Caporali, Marco Borga, Fabio Pilotti, Anna Paola Lonati, Matteo Dall’Amico 44 Applicazione del sistema idrologico GEOframe per la stima di tendenze di siccità in un contesto mediterraneo: applicazione sul bacino del Crati nell'ambito del progetto Agriclima Riccardo Busti, Daniela Biondi, Giuseppe Formetta 46 The GEOframe system deployment for the water budget quantification in the eastern mountain area of the Po river basin Gaia Roati, Marco Brian, Giuseppe Formetta, Shima Azimi, Hossein Salehi, Daniele Andreis, John Mohd Wani, Francesco Tornatore, Riccardo Rigon 47 RIVERTEMP classifier: un nuovo web tool per identificare e classificare i fiumi non perenni Paolo Vezza, Isabelle Brichetto, Carmela Cavallo, Christina Dolianidi, Almudena González Costas, Anastasios Karakostas, Nikos Nikolaidis, Maria Lilli, Giammarco Manfreda, Giovanni Negro, Guillermo Palau-Salvador, Maria Nicolina Papa, Carles Sanchis-Ibor, Spiros Tsalageorgos 49 Comparing the performance of statistical models in seasonal streamflow forecasting: the case study of the Imera Meridionale River Basin, Sicily, Italy Shewandagn Lemma Tekle, Brunella Bonaccorso, Mohamed Naim 51 An Open-Source Tool for Generating Hourly Synthetic Streamflow Series in Ungauged Basins Using Regional Flow-Duration Curves Alan Spadoni, Rosanna Foraci, Michele Di Lorenzo, Tommaso Simonelli, Attilio Castellarin 53 Setting an environmental flow regime under climate change in a data-limited Mediterranean basin with temporary river Marianna Leone, Francesco Gentile, Antonio Lo Porto, Giovanni Francesco Ricci, Cristoph Schürz, Michael Strauch, Martin Volk, Anna Maria De Girolamo 55 3 Monitoring deep unsaturated zone in an Australian Mediterranean area to reveal crucial insights for water resources management Simone Gelsinari, Richard Silberstain, Sally Thompson 56 Assessing groundwater sustainability through physics-based and machine learning models: a comparative study in Emilia-Romagna region (Italy) Ilaria Delfini, Daniel Zamrsky, Alberto Montanari 57 Modelling groundwater flow and contaminant transport in karst aquifers: a case study from southern Italy Alessandra Campobasso, Gabriella Balacco, Vito Specchio, Alberto Ferruccio Piccinni 59 Urban aquifer recharge and contamination by leaking sewers: a comparative analysis of different modeling approaches Andrea D’Aniello, Dina Pirone, Luigi Cimorelli, Domenico Pianese 61 Individuazione delle perdite fognarie tramite l’analisi dei contaminanti emergenti negli acquiferi urbani: sviluppo di un approccio basato sull’intelligenza artificiale Dina Pirone, Luigi Cimorelli, Daniele Martino, Domenico Pianese, Andrea D’Aniello 63 Irrigation volumes monitoring coupling an energy-water balance model with ground and satellite data Nicola Paciolla, Chiara Corbari 65 Bridging knowledge and practice in the WEFE Nexus: multi-level participatory modelling for sustainable irrigation in the mediterranean Virginia Rosa Coletta, Stefania Santoro, Laura Selicato, Claudia Panciera, Ivan Portoghese, Alessandro Pagano, Raffaele Giordano 67 The quantification of energy for irrigation Davide Danilo Chiarelli, Paolo D’Odorico, Aldo Fiori, Akhil Unnikrishnan, Ivan Lombardich, Maria Cristina Rulli 69 Prioritizing Drought Resilience: Identifying High-Vulnerability Irrigated Areas in Italy Harsh Nanesha, Lorenza Cappellato, Benedetta Moccia, Flavia Marconi, Francesco Napolitano, Elena Ridolfi, Maria Cristina Rulli, Davide Danilo Chiarelli 71 Quantifying vegetation transpiration from groundwater in Mediterranean forest Simone Gelsinari, Andrea Alessandri, Daniele Penna 73 Monitoraggio da satellite della intermittenza del fiume Tagliamento negli ultimi 40 anni Lucio Iantorno, Carmela Cavallo, Isabelle Brichetto, Giammarco Manfreda, Giovanni Negro, Paolo Vezza, Maria Nicolina Papa 74 Is Today Drier? A Metastatistical framework for drought frequency across Eras Maria Francesca Caruso, Gabriele Villarini, Marco Marani 76 Assessing snow water equivalent reconstruction using multi-source highresolution satellite data and hydrological models in alpine regions Michele Bozzoli, Valentina Premier, Cristian Tonelli, Giuseppe Formetta, Giacomo Bertoldi, Carlo Marin, Mathias Bavay 77 4 Use of Snow Water Equivalent Products for Improving the Parameterisation of Hydrological Models in Northern Italy Gokhan Sarigil, Mattia Neri, Elena Toth 79 Previsione della resa mensile di tre sorgenti mediante un modello ibrido CNNLSTM e SF Francesco Castaldo, Claudio Arena, Antonio Francipane, Leonardo Valerio Noto 81 Exploring Validation Methodologies for Modelled Soil Moisture Using Ground Measurements and Remote Sensing Based Data Fabio Delogu, Francesco Silvestro, Fabio Gardella, Giorgio Boni, Luca Repetto 83 Dynamic Soil-Atmosphere Interactions in Mediterranean Forests: Insights from Wavelet Analysis Ilenia Murgia, Christian Massari, Matteo Verdone, Konstantinos Kaffas, David Labat, Daniele Penna 85 Regionalization of an Hourly Lumped Hydrological Model Using a Decision-Tree Approach Giuditta Smerilli, Luca Lombardo, Anna Basso, Alberto Viglione, Alberto Montanari, Attilio Castellarin 86 Problematiche di calibrazione di trasformazioni afflussi-deflussi in bacini non strumentati a monte del Lago Maggiore Pietro Bogoni, Giulia Evangelista, Pierluigi Claps 88 Calibrazione di modelli afflussi-deflussi attraverso i livelli idrometrici per la ricostruzione delle serie storiche di portata nella regione Campania Giacomo Nicoletti, Antonia Longobardi, Giovanni Braca, Paolo Villani 90 Modeling Streamflow under Data Scarcity: Regionalization of IHACRES Using Catchment Attributes in a Mediterranean Setting Francesco Alongi, Caterina Alonzo, Antonio Francipane, Leonardo Valerio Noto 92 Modellazione idrologica-idrodinamica integrata per la simulazione della risposta idrologica di un bacino di testata Pierfranco Costabile, Carmelina Costanzo, Luca Furnari, Francesco Greco, Margherita Lombardo, Alfonso Senatore, Giuseppe Mendicino 94 A cost-effective approach for estimating flood hazardscenarios induced by climate change Leonardo Mancusi, Arianna Trevisiol, Andrea Abbate, Paola Faggian, Raffaele Albano 96 An index-based approach to identify areas with low flood risk sustainability Daniela Biondi, Danilo Spina, Giovanna Capparelli, Francesco Cruscomagno 98 L’interazione tra la dinamica fluviale e del mare in aree antropizzate prone al rischio idrogeologico Antonia Longobardi, Fabio Dentale, Albina Cuomo, Roberta D’Ambrosio, Angela Di Leo, Giacomo Nicoletti, Michele Pisani, Anna Sansanelli, Domenico Guida 100 Exploring the potential impact of rainfall extreme dynamics on river intersections Gianluca Lelli, Paola Mazzoglio, Alessio Domeneghetti, Serena Ceola 102 5 Pericolosità vs. Rischio: Un Cambio di Prospettiva Domenico Denora 104 Convolutional neural network model for rapid prediction of urban flash flood: a case study in Matera, Italy Muhammad Asif, Roberto Bentivoglio, Raffaele Albano 106 Analisi morfo-idrologica per lo studio degli allagamenti pluviali urbani Giulio Paradiso, Daniele Ganora 108 Assessment of damage scenarios due to pluvial flooding using remote sensing and object detection models Arianna Cauteruccio, Giorgio Boni, Roozbeh Rajabi , Gabriele Moser 110 Vehicles instability criterion for flood risk assessment in urban areas Omayma Amellah, Raffaele Albano, Aurelia Sole 112 Analisi multi-rischio e multi-livello dello stato dei Ponti Fluviali della Rete Stradale nel Bacino dell’Agri gestita dalla Provincia di Potenza: La Classe di Attenzione per Rischio Idraulico Aurelia Sole, Raffaele Albano, Carmine Limongi, Pietro Vuono, Beniamino Onorati, Giuseppe Francesco Cesare Lama, Ruggero Ermini, Annamaria De Vincenzo, Domenica Mirauda 113 Advanced Multi-Risk Analysis of the River Bridges Status in the Provincial Road Network of Potenza (Basilicata): Segmentation-Guided Machine Learning for Bridge Defect Detection in Wide-Angle Surface Images Aurelia Sole, Raffaele Albano, Gilda Manfredi, Giuseppe Santarsiero, Valentina Picciano 116 Analisi multi-rischio e multi-livello dello stato dei Ponti Fluviali della Rete Stradale nel Bacino dell’Agri gestita dalla Provincia di Potenza: Stabilità dei versanti Aurelia Sole, Raffaele Albano, Giuseppe Francesco Cesare Lama, Mario Bentivenga, Salvatore Ivo Giano, Vincenzo Siervo, Francesco Sdao 118 Assessment of bridge overtopping hazard at different spatial scales Michele Amaddii, Fabio Castelli, Mario di Bacco, Chiara Arrighi 120 Prioritizing bridges at flood risk: a large-scale index based on overtopping and traffic impact Michele Amaddii, Francesco Paolo Deflorio, Amirehsan Charlang Bakhtyari, Fabio Castelli, Chiara Arrighi 122 Flow charateristics evaluation in gravel bed rivers Donatella Termini, Peyman Peykani 124 Un approccio basato sui dati per il miglioramento della previsione delle piene: il caso del progetto TECH4You nel bacino del Crati Marco Dionigi, Stefania Camici, Melissa Sessa, Domenico De Santis, Christian Massari, Silvia Barbetta, Tommaso Moramarco 126 Enhancing flood forecasting through multi-model hydrological simulation and Bayesian uncertainty estimation with remote and modelled precipitation forcings Domenico De Santis, Elenio Avolio, Silvia Barbetta, Stefania Camici, Marco Dionigi, Sara Modanesi, Daniela Biondi, Tommaso Moramarco, Christian Massari 127 12 AIDA (Agricultural Irrigation Demand Analyzer): distribuzione e idroesigenza di colture irrigate e non irrigate a 1km di risoluzione per l’Italia Nikolas Galli*, Harsh Nanesha, Arianna Tolazzi, Davide Danilo Chiarelli, Maria Cristina Rulli Dipartimento di Ingegneria Civile e Ambientale, Politecnico di Milano, Milano, Italia *e-mail: [email protected] Sommario In Italia, l’agricoltura occupa un ruolo centrale non solo nella produzione alimentare, ma anche nella salvaguardia ambientale, nell’economia e nel patrimonio culturale del Paese. Al tempo stesso, la vulnerabilità del settore agricolo italiano agli eventi idroclimatici è stata evidenziata da episodi estremi come le siccità degli ultimi anni. In questo contesto, risulta essenziale disporre di una base modellistica che quantifichi modo accurato e spazialmente esplicito, ma al contempo consistente su tutto il territorio nazionale, i fabbisogni idrici agricoli, distinguendo tra colture irrigue e asciutte, e tra componenti di domanda soddisfatte da acqua verde (pioggia e umidità del suolo) e acqua blu (irrigazione). AIDA (Agricultural Irrigation Demand Analyzer) (Galli et al., in press) risponde a queste esigenze, fornendo mappe ad alta risoluzione della distribuzione colturale e dei fabbisogni irrigui sull’intero territorio nazionale. Per la creazione di mappe dettagliate della copertura agricola in Italia, si è partiti integrando, i dataset EUCROPMAP 2022 (risoluzione 10 m) (Ghassemi et al., 2024) e CORINE (EEA, 2020) e i censimenti ISTAT a livello comunale (ISTAT, 2020), attraverso un’armonizzazione spaziale e semantica e tenendo conto dei limiti di ciascun dataset. Per quanto riguarda la gestione dell’irrigazione, sono stati utilizzati i dataset GMIE (Tian et al., 2024) e GMIA (Siebert et al., 2013) attribuendo le aree irrigate alle singole colture a seconda della loro probabilità di irrigazione basata su dati ISTAT. Le statistiche provinciali di ISTAT sono state invece usate per la validazione del dato. Il prodotto così ottenuto alimenta quindi il modello idrologico WATNEEDS (Chiarelli et al., 2020), che simula il bilancio idrico agricolo su base giornaliera. WATNEEDS si focalizza sulla stima della domanda idrica delle colture, distinguendo tra acqua verde (proveniente dalle precipitazioni) e acqua blu (fornita dall’irrigazione). Il modello si basa su una serie di parametri, tra cui le caratteristiche del suolo, i dati climatici giornalieri provenienti, per AIDA, dal dataset MERIDA (Bonanno et al., 2019) e le informazioni specifiche sulle colture (come i coefficienti colturali e i calendari di crescita). Questo permette di calcolare in modo dinamico l’evapotraspirazione effettiva (ETa) e la domanda di irrigazione necessaria per soddisfare le esigenze delle colture, in funzione delle condizioni ambientali. I risultati della distribuzione colturale mostrano, rispetto alle statistiche provinciali ISTAT, correlazioni lineari superiori al 75% ed errori inferiori al 20% in più dell’80% delle aree analizzate. Le domande irrigue stimate da AIDA mostrano un buon accordo con i volumi al campo riportati da ISTAT, al netto di applicare un coefficiente di efficienza di campo alle risaie, che tenga conto della pratica di sommersione. AIDA presenta alcune incertezze, in particolare a scala locale, legate a errori di classificazione nelle mappe colturali e alla risoluzione relativamente grossolana dei dati di validazione, influenzata da pratiche come la rotazione colturale. Per i flussi idrici, le principali fonti di incertezza derivano dai dati e dai parametri di input e da scelte modellistiche inevitabilmente arbitrarie in WATNEEDS. Il dataset non include pratiche agricole come coltivazioni in serra o raccolti multipli, e la domanda irrigua blu è stimata come fabbisogno biofisico, non come prelievo effettivo. Inoltre, non sono considerati contributi da risalita capillare, potenzialmente rilevanti in aree con falda superficiale, come la Pianura Padana. Pertanto, eventuali versioni future di AIDA potranno prevedere nuovi input meteorologici, una migliore armonizzazione dei dati colturali e di irrigazione che tenga conto anche di dati disponibili a livello regionale, e la modellizzazione diretta in WATNEEDS di pratiche agricole come l'agricoltura in serra e la sommersione, attraverso l’inclusione di processi come la risalita capillare e il ristagno delle acque. Nonostante questi limiti, AIDA fornisce stime coerenti sulla domanda idrica agricola e sull’uso del suolo in Italia, distinguendo tra aree e fabbisogni millimetrici per permettere una maggiore flessibilità d’uso anche a scala locale. Rappresenta così un contributo solido per analisi, pianificazione e gestione delle risorse idriche nel settore agricolo, particolarmente rilevante in un contesto esposto a pressioni climatiche e idrologiche crescenti. Bibliografia Bonanno, R., Lacavalla, M. & Sperati, S., 2019. A new high‐resolution Meteorological Reanalysis Italian Dataset: MERIDA. Quarterly Journal of the Royal Meteorological Society 145, 1756–1779 13 Chiarelli, D. D., et al., 2020. The green and blue crop water requirement WATNEEDS model and its global gridded outputs. Scientific Data 7, 273 European Environment Agency, 2019. CORINE Land Cover 2018 (raster 100 m), Europe, 6-yearly - version 2020_20u1, May 2020 Galli, N., Nanesha, H., Tolazzi, A., Chiarelli, D. D. and Rulli, M. C., 2024. The AIDA 1km resolution crop-specific rainfed and irrigated areas and green and blue water demands for Italy. Scientific Data (in press) Ghassemi, B., et al., 2024. European Union crop map 2022: Earth observation’s 10-meter dive into Europe’s crop tapestry. Scientific Data, 11, 1048 ISTAT. Statistiche Istat --- dati.istat.it. Siebert, S., Henrich, V., Frenken, K. and Burke, J., 2013. Global Map of Irrigation Areas version 5. Rheinische Friedrich-Wilhelms-University, Bonn, Germany Food and Agriculture Organization of the United Nations, Rome, Italy Tian, F., et al., 2024. GMIE-100: a global maximum irrigation extent and irrigation type dataset derived through irrigation performance during drought stress and machine learning method. Earth System Science Data Discussion, 1–33 14 Analisi di scarsità idrica e competizioni intersettoriali per l’uso dell’acqua attraverso dati di domanda idrica ad alta risoluzione Nikolas Galli, Francesco Capone*, Camilla Govoni, Harsh Nanesha, Davide Danilo Chiarelli, Maria Cristina Rulli Dipartimento di Ingegneria Civile e Ambientale, Politecnico di Milano, Milano *e-mail: [email protected] Sommario La scarsità idrica rappresenta una delle principali criticità ambientali e socioeconomiche globali, con conseguenze crescenti sulla sicurezza alimentare, sugli ecosistemi e sulla gestione delle risorse. Anche in l’Italia il cambiamento climatico aggrava la pressione sulle risorse idriche, insieme a una domanda idrica elevata e a eventi siccitosi sempre più frequenti, in particolare nelle regioni meridionali e nel bacino del Po. In tale contesto, risulta essenziale disporre di strumenti capaci di quantificare in modo dettagliato e su base spaziale la scarsità idrica e i suoi impatti sul settore agricolo. Nel contesto della quantificazione della scarsità idrica, due tra i più comuni indicatori sono la scarsità di acqua verde, che misura la quota di domanda evapotraspirativa delle colture soddisfatta dalle precipitazioni, e la scarsità di acqua blu, definita come il rapporto tra domanda complessiva e disponibilità di risorse idriche prelevabili, al netto dei flussi ambientali. Il primo è direttamente legato alla perdita di resa per stress idrico, mentre il secondo consente di valutare la sostenibilità dell’uso irriguo e la pressione sugli ecosistemi. Questo studio si propone di utilizzare questi indicatori per esplorare l’esposizione dei diversi settori produttivi italiani alla scarsità idrica, con un occhio di riguardo per l’agricoltura, e le competizioni intersettoriali che possono derivarne. L’analisi è stata condotta sul territorio italiano per il periodo 2013–2023 utilizzando AIDA (Agricultural Irrigation Demand Analyzer) (Galli et al, in press), che combina un’armonizzazione di mappe colturali alla simulazione agro-idrologica del modello WATNEEDS (Chiarelli et al., 2020), calcolando il bilancio idrico del suolo con risoluzione spaziale di 30 arcosecondi. AIDA distingue, per ciascuna cella, tra fabbisogni colturali coperti da acqua verde e quelli soddisfatti tramite irrigazione, generando stime di domanda blu associate a mappe colturali aggiornate e validate sul territorio italiano. Le domande irrigue sono state integrate con quelle domestiche e industriali, ricavate dal dataset di Khan et al. (2023) e redistribuite spazialmente mediante mappe di insediamento residenziale e non residenziale a 100 metri di risoluzione. La scarsità di acqua blu è stata calcolata considerando anche i deflussi locali e a monte, corretti per requisiti ambientali e usi preesistenti. I risultati mostrano che vaste aree agricole e urbane del Paese si trovano in condizioni di scarsità idrica per periodi prolungati, con particolare intensità nella pianura padana e nelle aree meridionali durante la primavera e l’estate. Pur con l’agricoltura come fattore determinante nell’intensità della scarsità idrica, inquanto utilizzatore maggioritario di acqua, l’industria dimostra di avere un ruolo importante nella persistenza della scarsità. Queste criticità riflettono gradienti climatici e morfologici, ma anche socioeconomici, nella sostenibilità dell’uso della risorsa idrica. Il modello, basato sull’idromorfologia naturale, non include attualmente gli effetti delle infrastrutture irrigue, il che rappresenta un limite, in assenza di dati certi sui punti di prelievo e sulle effettive consegne idriche. Nonostante ciò, l’approccio consente di identificare le aree più esposte e i settori più vulnerabili, evidenziando l’importanza di una lettura integrata tra disponibilità, domanda e dinamiche idroclimatiche e socioeconomiche. Bibliografia Chiarelli, D. D. et al. The green and blue crop water requirement WATNEEDS model and its global gridded outputs. Sci. Data 7, 273 (2020). Galli, N., Nanesha, H., Tolazzi, A., Chiarelli, D. D. & Rulli, M. C. The AIDA 1km resolution crop-specific rainfed and irrigated areas and green and blue water demands for Italy. Scientific Data (in press). Khan, Z. et al. Global monthly sectoral water use for 2010–2100 at 0.5° resolution across alternative futures. Sci. Data 10, 1–16 (2023). Nieves, J. J., Sorichetta, A., Linard, C., Bondarenko, M., Steele, J. E., Stevens, F. R., Gaughan, A. E., Carioli, A., Clarke, D. J., Esch, T., & Tatem, A. J. (2020). Annually modelling built-settlements between remotelysensed observations using relative changes in subnational populations and lights at night. Computers, Environment and Urban Systems, 80, 101444. https://doi.org/10.1016/j.compenvurbsys.2019.101444 15 Quantifying the impacts – from paddies to dry rice fields Francesca Padoan*, Maria Cristina Rulli, Davide Danilo Chiarelli Department of Civil and Environmental Engineering, Politecnico di Milano, 20133, Milan, Italy. *e-mail: [email protected] Abstract Rice is the main user of irrigation water worldwide, yet field-scale studies that link local practice to global water footprints are still limited. The aim of this thesis is to evaluate and compare the hydrology of wet and dry rice cultivation systems from individual fields to a global scale. We adapted the WATNEEDS soil-water model to represent three vertical layers – ponded water, saturated root zone, unsaturated soil – and applied it to a continuously flooded paddy near Novara (Piedmont, Italy). Simulations (2008-2023) show that wet rice requires a nearly constant 0.1 km3/year of effective irrigation, of which more than 80% bypasses roots as deep percolation. In contrast, dry rice reduces blue water withdrawals by almost 20%, at the cost of higher sensitivity to rainfall variability. Analyzing the model on a global scale using a 5-arc-minute resolution shows that flooded rice utilizes approximately 752 km3/year of irrigation on average - four times that of dry rice - and loses approximately 810 km3/year to deep percolation. China and India alone account for more than half of these flows. Experiments reveal that implementing Alternate Wetting and Drying practices in paddy fields in Asia could save up to 110 km3/year of blue water and decrease methane emissions by nearly 30%, with minimal yield loss. Three insights emerge. First, prioritising AWD where blue-water withdrawals already exceed sustainable limits leads to rapid savings. Furthermore, reframing deep percolation as managed aquifer recharge shifts it from “loss” to shared resource. And last, because a substantial share of paddy evapotranspiration returns as downwind rainfall, atmospheric moisture – together with rivers and aquifers – must be included in basin water accounts. Acting on these strategies can relieve pressure among agriculture, cities, and ecosystems, keeping rice production within sustainable blueand green-water boundaries, as illustrated locally by the historic marcite meadows of northern Italy: winter-flooded forage fields that recycle water through shallow groundwater and evaporation, sustain cold-season baseflow, and demonstrate how flood-based irrigation can reconcile agricultural production with regional hydrology. References Bouman, B. A. M., Lampayan, R. M., and Tuong, T. P. (2007). Water management in irrigated rice: Coping with water scarcity. Manual, International Rice Research Institute, Los Ba˜nos, Philippines. 54 pages. Carrijo, D. R., Lundy, M. E., and Linquist, B. A. (2017). Rice yields and water use under alternate wetting and drying irrigation: A meta-analysis. Field Crops Research, 203:173–180. Chiarelli, D. D., Passera, C., Rosa, L., Davis, K. F., D’Odorico, P., and Rulli, M. C. (2020). The green and blue crop water requirement watneeds model and its global gridded outputs. Scientific Data, 7:273. Devanand, A., Huang, M., Ashfaq, M., Barik, B., and Ghosh, S. (2019). Choice of irrigation water management practice affects indian summer monsoon rainfall and its extremes. Geophysical Research Letters, 46(15):9126–9135. Food and Agriculture Organization of the United Nations (2021). AQUASTAT Main Database – Water Use by Sector. https://www.fao.org/aquastat/en/databases/maindatabase. Dataset last updated 2021, accessed 9 July 2025. Food and Agriculture Organization of the United Nations (2025). Rice production. Retrieved June 22, 2025, from Our World in Data. Gilardi, G. L. C., Mayer, A., Rienzner, M., Romani, M., and Facchi, A. (2023). Effect of alternate wetting and drying (awd) and other irrigation management strategies on water resources in rice-producing areas of northern italy. Water, 15(12):2150. Greve, P., Schmitt, A. U., Miralles, D. G., McDermid, S. P., Findell, K. L., Garcìa-Garcìa, A., and Peng, J. (2025). Observational evidence of increased afternoon rainfall downwind of irrigated areas. Nature Communications, 16:3415. Kebede, E., Oluoch, K. O., Siebert, S., Mehta, P., Hartman, S., Jagermeyr, J., Ray, D., Ali, T., Brauman, K. A., Deng, Q., Xie, W., and Davis, K. F. (2024). A global open-source dataset of monthly irrigated and rainfed cropped areas (mirca-os) for the 21st century. HydroShare. Keys, P., van der Ent, R., Gordon, L., Hoff, H., Nikoli, R., and Savenije, H. (2012). Analyzing precipitationsheds 16 to understand the vulnerability of rainfall dependent regions. Biogeosciences, 9(2):733–746. Lampayan, R., Rejesus, R., Singleton, G., and Bouman, B. (2015). Adoption and economics of alternate wetting and drying water management for irrigated lowland rice. Field Crops Research, 170:95–108. Linquist, B., van Groenigen, K. J., Adviento-Borbe, M. A., Pittelkow, C., and van Kessel, C. (2012). An agronomic assessment of greenhouse gas emissions from major cereal crops. Global Change Biology, 18(1):194– 209. Mekonnen, M. M. and Hoekstra, A. Y. (2011). The green, blue and grey water footprint of crops and derived crop products. Hydrology and Earth System Sciences, 15(5):1577–1600. Posada-Marìn, J., Salazar, J., Rulli, M. C., Wang-Erlandsson, L., and Jaramillo, F. (2024). Upwind moisture supply increases risk to water security. Nature Water, 2:875–888. Tuong, T. and Bouman, B. (2003). Rice production in water-scarce environments. In Kijne, J. W., Barker, R., and Molden, D., editors, Water Productivity in Agriculture: Limits and Opportunities for Improvement, pages 53–67. CABI Publishing, Wallingford. 17 Modeling the effects of shifting planting dates and irrigation management strategies in maize cropping systems for sustainable water resource management Adrian Chummac1,2,3*, Brunella Bonaccorso2 1 University School for Advanced Studies IUSS Pavia, Piazza della Vittoria 15, 27100, Pavia 2 Department of Engineering, University of Messina, Contrada di Dio, Villiaggio Sant’Agata, 98166, Messina 3 Central Luzon State University, Science City of Muñoz, Nueva Ecija 3120, Philippines *e-mail: [email protected] Abstract The availability of water resources in agriculture plays a vital role in ensuring productivity, given the sector’s importance to food security. Assessing in-season crop water requirements and evaluating irrigation management strategies provide essential insights into water demand, offering adaptation tools for managing limited water resources. This study evaluates crop water demand and adaptation strategies for managing water resources in a maize cropping system in the two provinces of Basilicata region (Matera and Potenza), Italy. The methodology employs a calibrated crop model (Mereu et al., 2019), namely CSM-CERES-Maize, initialized with crop management practices, soil characteristics, and weather data from 1991 to 2023. Two adaptation strategies were evaluated: (1) shifting planting dates (early, mid, and late planting) and (2) irrigation management (full irrigation, 75% deficit, 50% deficit, 25% deficit, and rainfed/no irrigation). The model simulates plant-soil-water interactions, providing a quantitative assessment of crop yield, seasonal water demand, water productivity, and the effectiveness of the adaptation strategies. Results of a mixed-effects analysis of variance using the lme4 package in R (Bates et al., 2015) showed that both planting date and irrigation management had a significant effect on crop yield and water productivity in both provinces; however, their interaction was not significant. An exception was observed in Matera, where the interaction between planting date and irrigation management had a significant effect on water productivity. Overall, a declining trend in seasonal water requirements was observed from early to late planting at most irrigation levels, indicating that later planting dates require less irrigation. Conversely, water productivity increased from early to late planting across all irrigation levels, with the highest values observed at late planting combined with 25% deficit irrigation. Post-hoc analysis revealed that, across all irrigation strategies, crop yields were highest for late planting in both provinces. However, the difference in yields between mid and late planting dates was not statistically significant, possibly due to better alignment with precipitation patterns. Full irrigation consistently resulted in the highest yields at all planting dates, though these yields were not significantly different from those under 75% deficit irrigation. This indicates a potential 25% water saving without substantial yield loss if adopting 75% deficit irrigation. The findings also suggest that in scenarios where water resources are limited or irrigation is not feasible, late planting may serve as a viable strategy, as it better aligns the crop’s growing season with precipitation patterns, as evidenced by increased water productivity and reduced irrigation requirements. The study is currently being extended to incorporate an evaluation of drought hazard impacts on crop yields, alongside an assessment of the efficacy of the proposed adaptation strategies in mitigating drought-induced yield risks. 18 Figure 1. Simulated yields in Matera and Potenza at different planting dates and irrigation management strategies over 33 years (1991-2023) Acknowledgments This study was partially carried out within the cascading funding project WaterWISE- “Water Management Strategies and Climate Change Adaptation in Southern Italy”– RETURN extended Partnership funded by NextGeneration EU (National Plan for Recovery and Resilience—PNRR, Mission 4, Component 2, Investment 1.3— D.D. 1243 of 2 August, 2022, PE0000005), CUP D43C22003030002 References Bates D, Mächler M, Bolker B, Walker S (2015). “Fitting Linear Mixed-Effects Models Using lme4.” Journal of Statistical Software, 67(1), 1–48. doi:10.18637/jss.v067.i01. Mereu, V., Gallo, A., & Spano, D. (2019). Optimizing Genetic Parameters of CSM-CERES Wheat and CSMCERES Maize for Durum Wheat, Common Wheat, and Maize in Italy. Agronomy, 9(10), 665. https://doi.org/10.3390/agronomy9100665 19 On the hydrological sustainability of agricultural areas of the Tigris–Euphrates river basin Hiba Mohammad*, Muhammad Faisal Hanif, Marco Peli, Roberto Ranzi, Stefano Barontini Department of Civil, Environmental, Architectural Engineering and Mathematics, University of Brescia, Brescia, Italy. *e-mail: [email protected] Abstract The Tigris–Euphrates river basin, spanning across Türkiye, Syria, and Iraq, is one of the most hydro-politically sensitive and ecologically stressed transboundary basins in the Middle East and North Africa (MENA) region (N. Al-Ansari, 2016). It plays a central role in supporting agricultural production, energy, and livelihoods, especially in Iraq and Syria, where dependence on the basin exceeds 90% of freshwater supply (FAO, 2009). However, dam construction, climate variability, and aggressive irrigation expansion have significantly altered the natural hydrological regime, particularly downstream (Al-Quraishi & Kaplan, 2021). Given the basin’s high dependence on agriculture and irrigation, a comprehensive analysis of its hydrological sustainability under cropand climate-specific conditions is urgently needed. This study aims to assess the long-term hydrological sustainability of the Tigris–Euphrates river basin between 1975 and 2022. The analysis focuses on crop-specific evapotranspiration demands (𝐸𝑇𝑚𝑎𝑥), and their influence on the regional water balance. The study used monthly temperature and precipitation data from 54 meteorological stations across the basin, sourced from the Mashreq domain dataset (1975–2022) (UNESCWA & ACSAD, 2021). 𝐸𝑇𝑚𝑎𝑥 was computed via the Thornthwaite formula and the FAO56-Hargreaves method, integrating crop coefficients (𝐾𝑐) (Allen et al., 1998). Based on the precipitation availability and on the 𝐸𝑇𝑚𝑎𝑥 demand, Melisenda’s aridity index (𝑎) was used to classify the areas as “wet” (a < -1) or “potentially dry” (-1 < a < 0; see Figure 1), based on their ability to refill the soil field capacity during the wet season, and to estimate the annual deficit and superavit (Melisenda, 1964). Then the Budyko framework was employed to assess the sustainability of the long-term agroclimatic water balance using the Budyko Aridity Index (𝐵𝐴𝐼 = 𝐸𝑇𝑚𝑎𝑥 𝑃 ⁄) and Budyko Dryness Index (𝐵𝐷𝐼 = 𝐴𝐸𝑇 𝑃 ⁄) (Budyko, 1974). Six scenarios were analysed for 𝐸𝑇𝑚𝑎𝑥 demand: Thornthwaite’s exosystemic demand, the actual land use/ land cover (LULC) and four standard crop-intercrop scenarios, i.e., Millet-Lentils (ML), Barley-Lentils (BL), Winter Wheat-Beans (WB), and Rice-Lentils (RL), to determine spatial water deficits, actual evapotranspiration, and sustainability under different field capacities. Our results indicate: Chronic aridity across Iraq and Syria, where water deficits are too great to sustain water-intensive crops like rice and wheat without extensive irrigation infrastructure. Millet consistently emerges as the optimal crop; halving irrigation demands compared to rice and improving resilience in dryland regions. Strategic crop selection and water-efficient irrigation practices are critical adaptation pathways. Adoption of drought-tolerant cereals, paired with localized water harvesting, could reduce stress on surface water systems and groundwater depletion. 20 Figure 1. Maps of the Melisenda’s aridity index for the Tigris-Euphrates river basin. The top left scenario is based on Thornthwaite’s potential evapotranspiration. The other five are based on the FAO56 procedure to estimate the maximum evapotranspirative demand for the following crop scenarios: actual ESA Land Cover map (LC), millet and lentils (ML), barley and lentils (BL), winter wheat and green beans (WB), and rice and lentils (RL). References Al-Ansari, N., 2016. Hydro-Politics of the Tigris and Euphrates Basins. Engineering, 08(03), 140–172 Allen, R., Pereira, L. and Smith, M., 1998. Crop evapotranspiration-Guidelines for computing crop water requirements-FAO Irrigation and drainage paper 56 Al-Quraishi, A. K. and Kaplan, D. A., 2021. Connecting changes in Euphrates River flow to hydropattern of the Western Mesopotamian Marshes. Science of The Total Environment, 768, 144445 Budyko, M. I., 1974. Climate and Life. Academic Press FAO, 2009. AQUASTAT Transboundary River Basins – Euphrates-Tigris River Basin (p. 14). Food and Agriculture Organization of the United Nations (FAO) Melisenda, I., 1964. Sui calcoli idrologici per il terreno agrario: Influenza del ‘clima’. L’acqua, 4, 3–20 UNESCWA, & ACSAD., 2021. Impact of Climate Change on Shared Water Resources in the Euphrates River Basin (E/ESCWA/CL1.CCS/2021/RICCAR/TECHNICAL REPORT.11; RICCAR Technical Report, p. 37). Arab Center for the Studies of Arid Zones and Dry Lands (ACSAD) and United Nations Economic and Social Commission for Western Asia (ESCWA) UN-ESCWA, & BGR., 2013. Inventory of Shared Water Resources in Western Asia. United Nations Economic and Social Commission for Western Asia; Bundesanstalt für Geowissenschaften und Rohstoffe. 21 Influenza sull’indice di aridità delle formule per la stima dell’evapotraspirazione potenziale Giovanni Selleri*, Mattia Neri, Elena Toth DICAM, Università di Bologna, Bologna, Italy *e-mail: [email protected] Sommario L’effetto combinato di tutti i processi evapotraspirativi svolge un ruolo chiave nel bilancio idrico per la modellazione idrologica a scala di bacino idrografico. Tuttavia, la misurazione diretta del tasso di evapotraspirazione è molto complessa e tipicamente vengono utilizzate formule che esprimono la cosiddetta evapotraspirazione potenziale (Ep, ovvero il flusso massimo in condizioni di disponibilità idrica illimitata), in funzione di alcune variabili meteorologiche. Oltre che per l’utilizzo come forzante dei modelli afflussi-deflussi, l’andamento e l’entità della Ep sono fattori cruciali per caratterizzare le dinamiche idrologiche e la similarità tra bacini, e in particolare la sua stima è utilizzata per il calcolo degli indici di aridità. Sono state proposte svariate formule per il calcolo della Ep e diverse di queste sono comunemente utilizzate con ottimi risultati; ma la scelta della formula rimane comunque a discrezione del singolo modellista, il quale solitamente si basa sui dati meteorologici disponibili e sulla propria esperienza. In questo studio vengono analizzate le differenze tra i valori stimati, riferiti ad un periodo esteso di osservazione, da otto diverse formule per un ampio insieme di bacini situati in diverse regioni idro-climatiche del mondo. Per considerare nell’analisi un vasto insieme di bacini in tutto il mondo, con disponibilità omogenea di tutte le forzanti meteorologiche necessarie, sono stati selezionati oltre tremila bacini, che coprono vaste aree geografiche di Stati Uniti, Brasile, Cile, Australia, Regno Unito ed Europa centrale, dal recente dataset Caravan (Kratzert et al. 2023). Per ogni bacino, Caravan raccoglie le serie delle portate giornaliere osservate ottenute da un insieme di dataset idrologici nazionali e regionali chiamati “CAMELS-type”. Nei dataset CAMELS le variabili meteorologiche mediate sul bacino sono derivate a partire da interpolazioni delle misure al suolo, mentre in CARAVAN, per mantenere uniformità sull’intero continente, i dati meteorologici giornalieri medi areali sono derivati dal prodotto di rianalisi globale ERA5-Land (Muñoz-Sabater et al., 2021), che include anche numerose variabili non presenti nei prodotti CAMELS (che in genere si limitano a precipitazione e temperatura). Tra le numerose formule proposte per la stima della Ep, sono state selezionate e applicate otto fra le più utilizzate in letteratura e nella pratica idrologica. In questo sommario sono presentati, per brevità, solo i risultati relativi a quattro di esse, riportate in Tabella 1. Tabella 1. Alcune formule adoperate in questo studio e mostrate nell’esempio in Figura 1 Formula Variabili richieste Penman-Monteith (versione FAO, Allen et al. 1998) Radiazione, pressione, vento, temperature media, massima, minima e di rugiada Priestley-Taylor Radiazione, temperatura media Hargreaves Temperatura media, massima e minima Thornthwaite Temperatura media Le stime medie areali di Ep risultanti dall’applicazione delle formule sono state confrontate analizzando la media annuale cumulata di lungo periodo e la massima escursione stagionale. Oltre ad analizzare la variabilità delle stime ottenute da ciascuna formula all’interno di una stessa area geografica, si sono confrontati i risultati ottenuti sulle diverse regioni per la stessa formula, per valutare l’impatto delle differenze idro-climatiche. Inoltre, per i bacini aventi almeno 30 anni di portate giornaliere continue, sono state raccolte le medie areali di lungo periodo P (precipitazione) e Q (portata osservata), in modo da determinare l’indice di aridità Ep/P ed l’indice evaporativo (P-Q)/P dei singoli bacini e rappresentarli nel diagramma di Budyko. A titolo di esempio, in figura 1 sono illustrati alcuni risultati per i bacini del Brasile. I risultati illustrati nei diagrammi a violino nel pannello 1a) mostrano come ci siano differenze non trascurabili tra le stime ottenute dalle diverse formule. In generale, per tutte le otto formule, risulta netta variabilità tra le stime ottenute con le diverse formule, con differenze significative a livello regionale; è quasi sempre possibile individuare le formule semplificate che localmente permettono di avvicinarsi maggiormente alle stime di riferimento (Penman- 28 Figure 1. a) Study area; b) Percent differences on monthly AET the reference to ETMonitor (2001-2019), GLASS (20012023), MODIS (2000-2022) [grey color referred to comparison with ETMonitor, red to GLASS and green to MODIS] References Avino, A., Cimorelli, L., Furcolo, P., Noto, L. V., Pelosi, A., Pianese, D., Villani, P., and Manfreda, S., 2024. Are rainfall extremes increasing in southern Italy? Journal of Hydrology, 631. Braca, G. and Ducci, D., 2018. Development of a GIS Based Procedure (BIGBANG 1.0) for evaluating groundwater balances at national scale and comparison with groundwater resources evaluation at local scale. Groundwater and Global Change in the Western Mediterranean Area. pp. 53-61. Hargreaves, G. H., and Samani, Z. A., 1985. Reference Crop Evapotranspiration from Temperature. Applied Engineering in Agriculture, 1(2), 96-99. Hlavčová, K., Kohnová, S., Szolgay, J., Parajka, J., & Blöschl, G. (2008). Assessment of climate change impact on the hydrologic regime in the upper Hron River basin, Slovakia. Journal of Hydrology and Hydromechanics, 56(3), 163–175. McCabe, G. J. and Wolock, D. M., 2011. Independent effects of temperature and precipitation on modeled runoff in the conterminous united states. Water Resources Research 47 (11). Mendicino, G., and Senatore, A., 2013. Regionalization of the Hargreaves Coefficient for the Assessment of Distributed Reference Evapotranspiration in Southern Italy. Journal of Irrigation and Drainage Engineering, 139(5), 349-362. Peres, D. J., Bonaccorso, B., Palazzolo, N., Cancelliere, A., Mendicino, G., and Senatore, A., 2023. A dynamic approach for assessing climate change impacts on drought: an analysis in Southern Italy. Hydrological Sciences Journal, 68(9), pp.1213-1228. Thornthwaite, C. W., 1948. An approach toward a rational classification of climate. Geographical review, 38(1), 55-94. Xie, Z., Yao, Y., Zhang, X., Liang, S., Fisher, J. B., Chen, J., Jia, K., Shang, K., Yang, J., Yu, R., Guo, X., Liu, L., Ning, J., and Zhang, L., 2022. The Global LAnd Surface Satellite (GLASS) evapotranspiration product Version 5.0: Algorithm development and preliminary validation. Journal of Hydrology, 610. Zheng, C., Jia, L. and Hu G., 2022. Global land surface evapotranspiration monitoring by ETMonitor model driven by multi-source satellite earth observations. Journal of Hydrology 613. Running, S., Mu, Q., Zhao, M., and Moreno, A., 2021. MODIS/Terra Net Evapotranspiration Gap-Filled 8-Day L4 Global 500m SIN Grid V061 [Data set]. NASA Land Processes Distributed Active Archive Center. https://doi.org/10.5067/MODIS/MOD16A2GF.061 Date Accessed: 2025-06-18 29 Toward better hydrological drought representation: optimizing vegetation dynamics into the NOAH-MP land surface model for Southern Italy Sara Modanesi1*, Domenico De Santis2, Daniela Dalmonech3, Alessio Collalti3, Francesco Avanzi4, Gabrielle De Lannoy5, Tommaso Moramarco1, Nunzio Romano6, Paolo Nasta6, Fabio Massimo Grasso7, Martina Natali1,8, Christian Massari1 1 Research Institute for the Geo-Hydrological Protection, National Research Council, Perugia, Italy 2 Research Institute for Geo-Hydrological Protection, National Research Council, Rende, Italy 3 Institute for Agriculture and Forestry Systems in the Mediterranean, National Research Council, Perugia, Italy 4 CIMA Research Foundation, Savona, Italy 5 Department of Earth and Environmental Sciences, KU Leuven, Heverlee, Belgium 6 Department of Agricultural Sciences, University of Naples Federico II, Portici, Italy 7 Institute of Atmospheric Sciences and Climate, National Research Council, Lecce, Italy 8 Department of Civil and Environmental Engineering, University of Perugia, Perugia, Italy *e-mail: [email protected] Abstract Meteorological droughts, driven by prolonged precipitation deficits and often amplified by increased atmospheric water demand, significantly affect the partitioning of the water budget in river basins. In this context, the complex interplay among land surface evaporation (ET), basin storage capacity, and vegetation dynamics plays a key role in controlling the transition from meteorological drought (precipitation deficit) to hydrological drought (reduced streamflow and groundwater recharge). Accurately modelling these processes in large-scale Land Surface Models (LSMs) requires a realistic representation of ET, soil moisture (SM), and their coupling with carbon fluxes, such as gross primary productivity (GPP). However, the performance of LSMs is highly sensitive to parameterization choices. In particular, the selection of runoff schemes and vegetation dynamics parameters jointly influence the coupling between SM and ET. This coupling is critical for regulating the partitioning of water fluxes and directly impacts streamflow simulation, water availability assessments, and drought severity. Despite their importance, the combined effects of runoff schemes and vegetation parameterization on drought representation remain underexplored. This study investigates the impact of vegetation parameters optimization in the Noah-MP LSM (version 4.0.1; Niu et al., 2011), implemented within the NASA Land Information System (LIS; Kumar et al. 2006; Peters-Lidard et al. 2007), with a specific focus on its influence on runoff dynamics and hydrological drought characterization. A series of experiments were conducted using four different runoff schemes incorporated into the Noah-MP LSM (Niu et al., 2007; Niu et al., 2005; Schaake et al., 1996; Dickinson et al.,1993), while satellite-derived datasets of GOSIF GPP, SMAP and in situ SM, and LSA SAF ET (Li and Xiao, 2019; O'Neill et al., 2021; Trigo et al. 2011), were used to evaluate model outputs. After selecting the best-performing runoff scheme, vegetation-related remote sensing products (e.g., MODIS leaf area index [LAI], Myneni et al., 2015), were used to calibrate vegetation parameters and assess their impact on the simulation of hydrological drought conditions. The analysis focused on catchments in Southern Italy, with particular emphasis on the Crati River basin, investigated as part of the “Tech4You – Technologies for Climate Change Adaptation and Quality of Life Improvement” project. Special attention was given to drought years, identified using the 12-month Standardized Precipitation Index (SPI-12). Results show that calibrating LAI, particularly in combination with the best performing runoff scheme, can improve the representation of key variables such as runoff and SM, without compromising the accuracy of ET and GPP. These findings highlight the crucial role of vegetation parameterization in modulating water fluxes and SM-ET coupling and offer valuable insights for improving drought representation in LSMs. This has direct implications for water resource management and agroecosystem resilience in drought-prone Mediterranean regions. 30 Acknowledgements This work was funded by the Next Generation EU - Italian NRRP, Mission 4, Component 2, Investment 1.5, call for the creation and strengthening of ‘Innovation Ecosystems’, building ‘Territorial R&D Leaders’ (Directorial Decree n. 2021/3277) - project Tech4You - Technologies for climate change adaptation and quality of life improvement, n. ECS0000009. This work reflects only the authors’ views and opinions, neither the Ministry for University and Research nor the European Commission can be considered responsible for them. References Dickinson, R. E., A. Henderson‐Sellers, and P. J. Kennedy, 1993. Biosphere‐Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Climate Model, NCAR Tech. Note NCAR/TN - 387+STR, 80 pp., Natl. Cent. for Atmos. Res., Boulder, Colo. Dickinson, R. E., M. Shaikh, R. Bryant, and L. Graumlich,1998. Interactive canopies for a climate model, J. Clim., 11, 2823–2836, doi:10.1175/1520-0442(1998)011<2823:ICFACM>2.0.CO;2. Kumar, S.V., and Coauthors, 2006. Land information system: An interoperable framework for high resolution land surface modeling. Environmental modelling & software, 21 (10), pp. 1402–1415. Li, X., and J. Xiao, 2019. Mapping photosynthesis solely from solar-induced chlorophyll fluorescence: A global, fine-resolution dataset of gross primary production derived fromoco-2. Remote Sensing, 11 (21), 2563. Myneni, R., Knyazikhin, Y., & Park, T., 2015. MOD15A2H MODIS/Terra Leaf Area Index/FPAR 8-Day L4 Global 500m SIN Grid V006. NASA EOSDIS Land Processes DAAC. Niu, G.‐Y., and Z.‐L. Yang, 2007. An observation‐based formulation of snow cover fraction and its evaluation over large North American river basins, J. Geophys. Res., 112, D21101. Niu, G.‐Y., Z.‐L. Yang, R. E. Dickinson, and L. E. Gulden, 2005. A simple TOPMODEL‐based runoff parameterization (SIMTOP) for use in global climate models, J. Geophys. Res., 110, D21106. O'Neill, P. E., S. Chan, E. G. Njoku, T. Jackson, R. Bindlish, and J. Chaubell. 2021. L3 Radiometer Global Daily 36 km EASE-Grid Soil Moisture, Version 8. [Indicate subset used]. Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. Peters-Lidard, C. D., D. M. Mocko, M. Garcia, J. A. Santanello Jr, M. A. Tischler, M. S. Moran, and Y. Wu, 2008. Role of precipitation uncertainty in the estimation of hydrologic soil properties using remotely sensed soil moisture in a semiarid environment. Water Resources Research, 44 (5). Trigo, I. F., and Coauthors, 2011. The satellite application facility for land surface analysis. 1092 International Journal of Remote Sensing, 32 (10), pp. 2725–2744 31 Driver-based classification identifies snowmelt as the primary cause of severe droughts in the Upper Adige basin Andrea Galletti1*, Susen Shrestha1,2, Mariapina Castelli3, Massimiliano Pittore1, Stefano Terzi1, Giacomo Bertoldi4 1 Eurac Research, Center for Climate Change and Transformation, Bolzano, Italy 2 Department of Land, Environment, Agriculture and Forestry, University of Padova, Padova, Italy 3 Eurac Research, Institute for Earth Observation, Bolzano, Italy 4 Eurac Research, Institute for Alpine Environment, Bolzano, Italy *e-mail: [email protected] Abstract Hydrological drought is intensifying across the European Alps. In 2022, northern Italy experienced Po River lows unseen in two centuries: a ~500-year event that severely affected hydropower, irrigation, and aquatic ecosystems (Montanari, 2023; Ramírez-Molina, 2024). The accepted theory typically tracks drought as a series of cascading events, from meteorological anomalies through agricultural and hydrological stages to socioeconomic losses (Van Loon, 2015). However, impact-based monitoring relies on consistent local loss data such as crop yields, economic damage, and ecosystem-service disruptions, yet these records seldom provide the spatial or temporal coverage required for developing early warning systems. Consequently, drought studies should provide framing that translates observable signals into clear triggers for operational response. The qualitative scheme of Van Loon & Van Lanen (2012), along with its automated successor by Brunner et al. (2022), represent a step in this direction, linking each drought to the hydrometeorological driver that initiates it. However, climate, storage, and relief can vary significantly within a single Alpine basin, thus calling for a locally tuned implementation of these frameworks. Therefore, in this study, we developed a driver-based classification framework tailored to the 7,000 km² Upper Adige basin (Italy), analysing 22 nested sub-basins from 1992 to 2023. Our aim was to identify the drivers behind the most severe low-flow events, understand their characteristics, triggering mechanisms and spatial variability, and explore how such knowledge can refine strategic monitoring and early response in mountain catchments. To generate a basin-wide, internally consistent suite of driver variables, we reconstructed 1992–2023 daily precipitation, air temperature, snowfall, and snowmelt with the semi-distributed hydrological model TOPMELTICHYMOD (Norbiato et al., 2008, Zaramella et al., 2019). The model is in operational use at the local water authority for flood forecasting, is forced by spatialized meteorological observation, and is calibrated for optimal flow reconstruction at several gauges. As a first step, we successfully validated the model’s low-flow prediction (>75% of low-flow days). On this foundation, we adapted Brunner et al., (2022)’s framework to our domain, tailoring classification decisions and anomaly thresholds to the Upper-Adige hydroclimate (Figure 1). We tuned rules and parameters iteratively, comparing automated tags with expert manual labelling of every low flow (Q < Q20 for ≥ 20 days) until ≥ 90% concordance was achieved, ensuring that the classifier captured local hydrological nuances without overfitting. We then stress-tested robustness with a Monte Carlo simulation across 1,000 parameter sets, Slight variations in decision parameters do not affect the dominance hierarchy. Only the “composite” share grows under the most extreme parameter settings, reflecting the class’s role as a conservative fallback. The framework isolates 647 drought events across all 22 subcatchments, and melt-related episodes (warmand cold-snow seasons) account for 31% of them; the remaining drivers are composite (28%), rainfall-driven (27%), winter-recession (12%), and heatwave (1%). At the basin closing section, the balance tilts even further toward cryospheric control: melt-related types rise to 46% (warm 26.6%, cold 20%), with composite droughts receding to 17%, confirming that downstream events are primarily governed by snowmelt anomalies. Severity metrics further clarify the distinct characteristics of each drought driver. Warm snow season droughts deliver the highest median daily deficit (i.e., intensity), while composite events combine similarly high intensity with the longest median duration (approximately 55 days), resulting in the highest total-volume deficits (i.e., severity). Although less severe, winter recession droughts coincide with the annual low-flow season and can therefore trigger acute supply conflicts. Classifying low-flows by their governing processes provides actionable triggers instead of after-the-fact labels. Future improvements to this work concern the “composite” category, which will be split by assigning weights to concurrent anomalies. The presented classifier could be already used to translate April snowpack anomalies, May–June melt lags, 32 and persistent summer rainfall shortfalls into tiered alerts, giving operators a driver-specific lead time for action that plain discharge thresholds cannot provide. Figure 1. Schematic of the drought classification framework. The share of events (out of 647 total) assigned to each category is shown in percentage at the bottom. References Brunner, M. I., et al. (2022). Snow and ice-related water resources modeling using a novel multimodal approach. Water Resources Research. https://doi.org/10.1029/2022WR032070 Montanari, A. (2023). Controlli sulla composizione isotopica stabile della precipitazione giornaliera a Sydney (Australia): 9 anni di dati e Radon-222. Journal of Hydrology, 617, 129123. https://doi.org/10.1016/j.jhydrol.2023.129123 Norbiato, D., Borga, M., Degli Esposti, S., Gaume, E., & Anquetin, S. (2008). Flash flood warning based on rainfall thresholds and soil moisture conditions: An assessment for gauged and ungauged basins. Journal of Hydrology, 362(3-4), 274-290. Ramírez Molina, A. A., Tootle, G., Formetta, G., Piechota, T., & Gong, J. (2024). Extraordinary 21st Century Drought in the Po River Basin (Italy). Hydrology, 11(12), Article 219. https://doi.org/10.3390/hydrology11120219 Van Loon, A. F., & Van Lanen, H. A. (2012). A process-based typology of hydrological drought. Hydrology and Earth System Sciences, 16(7), 1915-1946. Van Loon, A. F. (2015). Hydrological drought explained. WIREs Water, 2(4), 359–392. https://doi.org/10.1002/wat2.1085 Zaramella, M., Borga, M., Zoccatelli, D., & Carturan, L. (2019). TOPMELT 1.0: a topography-based distribution function approach to snowmelt simulation for hydrological modelling at basin scale. Geoscientific Model Development, 12(12), 5251-5265. 33 Multivariate Risk Modeling of Drought Impact on Water Supply in the Camastra Reservoir Mohamed-Amine Lahkim-Bennani1,2*, Brunella Bonaccorso2 1 R&DGéoAp, FSTT, Abdelmalek Essaadi University, Tetouan, Morocco 2 Department of Engineering, University of Messina, Villaggio S. Agata, Messina, Italy *e-mail: [email protected] Abstract Mediterranean water reservoirs are increasingly threatened by climate variability and increasing water demand. This study examines the impact of drought on the water supply system by linking drought predictors with unmet water demands using a multivariate statistical risk modelling approach. The proposed approach is applied to the Camastra Reservoir in the Basilicata region in Southern Italy. We develop and test a multivariate indicator of drought impact using monthly data for 1993–2023: gauge-based precipitation, Copernicus 0–40 cm soil-moisture anomaly (SMAI), and reservoir inflow derived from the water balance. Meteorological drought indices (SPI, SPEI) and hydrological indices (SMAI, SFI) were computed at 3and 6-month time scales. First, cross-correlation was used to quantify climate-to-impact lags. All driver–impact pairs peaked at 1-month lead with r = 0.59–0.73 (p < 0.05) and retained usable signal up to 5 months, indicating slow propagation and partial buffering by basin soil moisture. The hydrological response (SFI) correlated more strongly with meteorological indices (r = 0.72–0.73; 52–54% of variance explained) than soil-moisture anomalies (r = 0.59–0.61), with a small advantage of SPEI over SPI. Our results are in line with those of Baez-Villanueva et al. (2024), who found that across 100 near-natural Chilean catchments that meteorological indices generally outperform soil-moisture indices (ESSMI) as proxies for streamflow drought. Since SPI/SMAI (both 3 and 6 months) were close to N(0,1), we combined them into a Joint Drought Index (JDI) using a bivariate Gaussian copula (Genest and Favre, 2007), consistent with approaches that couple SPI and vegetation health (VHI) into a probabilistic precipitation–vegetation index (PPVI) (Monteleone et al. 2020). A Q–Q validation demonstrates that the model replicates the empirical joint probabilities with high accuracy (RMSE = 0.022, R² = 0.992, MAPE = 10.5%), with noticeable deviations only in the tails. For impact detection, Receiver Operating Characteristic (ROC) analysis against SFI thresholds (≤ −2, −1.5, −1, 0) demonstrated that JDI consistently outperformed single indices (SPI or SPEI), especially at 1–2 months lead, underscoring its value as an early-warning indicator for reservoir operations. The proposed framework affords a reproducible basis for establishing activation thresholds and can be applied without major modification across Mediterranean settings. Current developments focus on converting JDI skill into operational triggers and on testing probabilistic forecasts that link seasonal SPI/SMAI guidance to decision rules. 34 Figure 1. Comparison of SPI₆, SPEI₆ and JDI₆ performance in identifying reported drought events (6-month): ROC curves (POD vs POFD) with fixed detection thresholds Z_SPI = 0, Z_SPEI = 0, and Z_JDI = -0.099; AUC values shown in the legend. References Baez-Villanueva OM, Zambrano-Bigiarini M, Miralles DG, et al (2024) On the timescale of drought indices for monitoring streamflow drought considering catchment hydrological regimes. Hydrol Earth Syst Sci 28:1415–1439. https://doi.org/10.5194/hess-28-1415-2024 Genest C, Favre A-C (2007) Everything You Always Wanted to Know about Copula Modeling but Were Afraid to Ask. J Hydrol Eng 12:347–368. https://doi.org/10.1061/(ASCE)1084-0699(2007)12:4(347) Monteleone B, Bonaccorso B, Martina M (2020) A joint probabilistic index for objective drought identification: the case study of Haiti. Nat Hazards Earth Syst Sci 20:471–487. https://doi.org/10.5194/nhess-20-471-2020 35 Più energia per gli utenti, più acqua per l’ambiente: scenari di ibridizzazione idroelettrico-fotovoltaico per un invaso pre-alpino Domenico Micocci1*, Cristiana Bragalli1, Elena Toth1, Tobias Wechsler2, Massimiliano Zappa2 1 Dipartimento di Ingegneria Civile, Chimica, Ambientale e dei Materiali, Università di Bologna, Bologna, Italia 2 Istituto federale di ricerca per la foresta, la neve e il paesaggio WSL, Birmensdorf, Svizzera *e-mail: dome[email protected] Sommario Le fonti energetiche rinnovabili non programmabili (FRNP), fra cui il solare fotovoltaico (PV), pongono nuove sfide per la gestione del sistema elettrico (Lund et al., 2015), in quanto la loro disponibilità è essenzialmente guidata da variabili climatiche ed è perciò indipendente dal fabbisogno degli utenti. L’idroelettrico (HP) a serbatoio può contribuire in modo significativo a mitigare il disallineamento fra produzione e consumo: si tratta infatti di una delle poche fonti energetiche rinnovabili programmabili, la cui produzione può essere modulata secondo le esigenze. Un livello di flessibilità ancora maggiore è offerto dagli impianti idroelettrici con pompaggio, che consentono anche l’assorbimento dell’energia eccedente il fabbisogno degli utenti tramite il sollevamento dell’acqua da un invaso a quota inferiore ad uno superiore. L’integrazione idroelettrico-fotovoltaico (HP-PV) in un impianto ibrido, gestito quindi in maniera coordinata, può offrire profili di generazione più regolari rispetto ad un impianto PV autonomo, agevolando così la penetrazione della fonte solare nel mix energetico (Jurasz et al., 2020). D’altro canto, però, l’accoppiamento di un impianto PV ad un impianto idroelettrico a serbatoio esistente potrebbe modificare le modalità di gestione dell’invaso, con possibili conseguenze sull’evoluzione del volume utile e sulla disponibilità di risorsa idrica; tale aspetto, tuttavia, è raramente analizzato in letteratura, pur essendone riconosciuta l’importanza (e.g., François et al., 2014). Per contribuire a questo tema, abbiamo sviluppato un esperimento numerico, immaginando di integrare un impianto idroelettrico con pompaggio realmente esistente con un ipotetico impianto fotovoltaico galleggiante, installato direttamente sulla superficie dell’invaso superiore. Come caso di studio, si è considerato l’impianto di generazione e pompaggio Etzelwerk, gestito dalle Ferrovie Federali Svizzere (potenza nominale in generazione: 120 MW). L’impianto sfrutta un dislivello di circa 480 m fra il Lago di Zurigo (utilizzato come serbatoio inferiore) ed il Lago della Sihl, un importante invaso artificiale svizzero, situato nella regione pre-Alpina (superficie del bacino: 155.5 km2; quota media: 1243 m s.l.m.; regime idrologico: nivo-pluviale pre-alpino). Trattandosi di un impianto reversibile, l’acqua può essere turbinata dal Lago della Sihl verso il Lago di Zurigo o pompata nel verso opposto. È inoltre previsto il rilascio di un deflusso ecologico dal Lago della Sihl, legato alle condizioni di deflusso a valle dell’invaso artificiale. Si è immaginato di introdurre sul Lago della Sihl un ipotetico impianto fotovoltaico galleggiante, di estensione pari al 10 % dell’area dello specchio liquido alla quota di minima regolazione (pari a 3.15 km2) e di potenza nominale pari al 50 % circa della potenza installata dell’impianto idroelettrico. La gestione integrata dell’impianto ibrido è stata simulata attraverso un approccio multi-disciplinare appositamente sviluppato (Micocci et al., 2025), finalizzato a mettere in luce specialmente le implicazioni dell’ibridizzazione HP-PV sulla gestione della risorsa idrica. La metodologia adottata presenta una struttura modulare, facilmente adattabile a diversi casi di studio: i) un modello di produzione fotovoltaica converte le serie meteorologiche di input (temperatura e irradianza) nella serie della potenza fotovoltaica disponibile; ii) un modello di domanda elettrica (in questo caso, una semplice regressione lineare multipla guidata esclusivamente da variabili temporali) fornisce il profilo del fabbisogno dell’utenza (che nel presente studio è solo quella ferroviaria); iii) un modello di gestione dell’invaso definisce i rilasci dal serbatoio sulla base dei fabbisogni e del volume disponibile (in questo caso, si sono definite delle curve di gestione, dedotte da un’analisi statistica della serie storica dei livelli d’invaso dal 1981 al 2023). La simulazione è stata eseguita a risoluzione oraria (per cogliere possibili interazioni fra fluttuazioni intragiornaliere della domanda e della produzione solare) ed è stata estesa ad un arco temporale di 38 anni (19812018), per conseguire una robusta caratterizzazione delle condizioni idro-climatiche. Si sono utilizzati come input dati meteorologici storici disponibili per l’area di studio (fra cui precipitazione, temperatura e radiazione solare), mentre i necessari dati idrologici (afflussi al serbatoio, perdite per evaporazione e deflusso naturale dei tributari a valle dell’invaso) sono stati ottenuti tramite un modello idrologico calibrato per il bacino della Sihl, realizzato tramite il sistema di modellazione semi-distribuito PREVAH (Viviroli et al., 2009). 36 La simulazione dell’impianto ibrido HP-PV è stata ripetuta per tre scenari: i) NoPV è uno scenario di riferimento, senza impianto PV a supporto dell’idroelettrico; ii) nello scenario PV1 l’energia solare, se disponibile, contribuisce a soddisfare il fabbisogno degli utenti e l’eventuale surplus è accumulato tramite pompaggio; iii) nello scenario PV2 si aggiunge, rispetto allo scenario PV1, la possibilità di rilasciare a valle il 50 % dell’acqua che non è stato necessario utilizzare per l’uso idroelettrico grazie al contributo del solare, onde aumentare il deflusso ecologico nei mesi da maggio ad ottobre. I risultati mostrano che l’introduzione del fotovoltaico permette di incrementare la produzione annua complessiva di energia di circa il 20 % rispetto allo scenario di riferimento NoPV (cfr. Tabella 1). L’ibridizzazione (scenari PV1 e PV2) consente inoltre di migliorare l’affidabilità del sistema, come mostrato dall’indice di affidabilità (qui definito come il rapporto tra l’energia complessivamente erogata e quella richiesta nel periodo di simulazione), che aumenta dal 90 % circa nello scenario NoPV al 97 % circa negli scenari PV1 e PV2. In termini di prestazioni energetiche, si sono ottenuti risultati simili negli scenari PV1 e PV2, nonostante i rilasci aggiuntivi consentiti, fra maggio e ottobre, nello scenario PV2. Tali rilasci aggiuntivi, tuttavia, hanno un impatto significativo sul rilascio medio mensile, che aumenta tra il 14 % (in maggio) e il 50 % (da giugno ad agosto) nello scenario PV2 rispetto al PV1 (cfr. Tabella 1). In conclusione, i risultati ottenuti confermano alcuni benefici dell’ibridizzazione HP-PV già messi in luce in studi precedenti (quali l’aumento dell’energia producibile e dell’affidabilità); emerge inoltre un potenziale per il miglioramento delle condizioni di deflusso a valle senza che siano invalidati significativamente i benefici sul fronte energetico. Tabella 1. Energia annualmente producibile in ciascuno scenario e rilascio medio mensile a valle (medie calcolate sull’intero arco temporale di simulazione 1981-2018) Energia totale (GWh/y) Energia da HP (GWh/y) Energia da PV (GWh/y) Rilascio maggio (m3/s) Rilascio giugno (m3/s) Rilascio luglio (m3/s) Rilascio agosto (m3/s) Rilascio settem. (m3/s) Rilascio ottobre (m3/s) NoPV 256.6 256.6 – 0.47 0.55 0.60 0.63 0.59 0.71 PV1 319.1 257.7 61.4 0.47 0.55 0.60 0.63 0.59 0.71 PV2 315.2 254.1 61.1 0.54 0.85 0.91 0.93 0.81 0.91 Ringraziamenti Lo studio è condotto nell’ambito del Partenariato Esteso RETURN, finanziato dall’Unione Europea – NextGenerationEU (Piano Nazionale di Ripresa e Resilienza – PNRR, Missione 4 Componente 2, Investimento 1.3 - D.D. 1243 2/8/2022, PE0000005). Bibliografia François, B., Borga, M., Anquetin, S., Creutin, J.D., Engeland, K., Favre, A.C., Hingray, B., Ramos, M.H., Raynaud, D., Renard, B., Sauquet, E., Sauterleute, J.F., Vidal, J.P., Warland, G., 2014. Integrating hydropower and intermittent climate-related renewable energies: a call for hydrology. Hydrological Processes, 28(21), 5465–5468 Jurasz, J., Ciapała, B., 2017. Integrating photovoltaics into energy systems by using a run-off-river power plant with pondage to smooth energy exchange with the power gird. Applied Energy, 198, 21–35 Lund, P.D., Lindgren, J., Mikkola, J., Salpakari, J., 2015. Review of energy system flexibility measures to enable high levels of variable renewable electricity. Renewable and Sustainable Energy Reviews, 45, 785–807 Micocci, D., Bragalli, C., Toth, E., Wechsler T., Zappa M., 2025. Hybridization of an alpine pumped-storage hydropower plant with floating solar photovoltaics: a study from the water resource perspective. Renewable Energy, 253, 123530 Viviroli, D., Zappa, M., Gurtz, J., Weingartner, R., 2009. An introduction to the hydrological modelling system PREVAH and its preand post-processing-tools. Environmental Modelling & Software, 24(10), 1209–1222. 37 Assessment of regional-scale freshwater availability and sustainability of anthropogenic withdrawals: the GOV4WATER project Alessia Flammini1*, Jacopo Dari1, Carla Saltalippi1, Francesco Leopardi1,2, Martina Natali1,2, Renato Morbidelli1 1 Department. of Civil and Environmental Engineering, University of Perugia, Perugia, Italy 2 National Research Council, Research Institute for Geo-Hydrological Protection, Perugia, Italy *e-mail: [email protected] Abstract Freshwater availability is widely recognized as the litmus test of global warming, with certain areas of the worlds experiencing increasing water crisis, such as the Mediterranean basin (Giorgi, 2006). Such a circumstance determines the compelling need for sustainable water management policies, that, however, are often limited by scarcity and scattering of collected data about anthropogenic abstractions for civil, industrial, and agricultural purposes. The Umbria region, located in central Italy, is experiencing increasing air temperatures at rates higher than the planetary average, in association with decreasing yearly cumulated rainfall amounts (Dari et al., 2023) and reduced runoff volumes (Rahi et al., 2023). This situation contributes in making the Umbria region a perfect case study for the Interreg Europe Gov4Water project, aimed at enhancing water governance and planning in light of climate change. In this study, we exploit two well-established, spatially distributed data sets to compute net freshwater (𝑁𝐹) remaining available over the region as the difference between water input, i.e., precipitation, and natural consumption, namely evapotranspiration. To do this, total precipitation and actual evapotranspiration rates derived from BIGBANG v8 (Braca et al., 2024) and ERA5-Land (European ReAnalysis v5 – Land) (MuñozSabater et al., 2021) are considered. 𝑁𝐹 is calculated for both the data sets, which are characterized by different spatial samplings, i.e., 1 km and 9 km for BIGBANG and ERA5-Land, respectively. The considered period spans from 1951 to 2024. Trend analyses reveal decreasing 𝑁𝐹 over the Umbria region at rates equal to -2.03 mm/year and to -1.30 mm/year for BIGBANG and ERA5-Land, respectively. In addition, such a decrease is more pronounced over portions of the region where higher average 𝑁𝐹 are found (i.e., the Apennine ridge). Even though the two data sets provide 𝑁𝐹 estimates appreciably different in magnitudes, they are highly correlated in time (Pearson 𝑟 equal to 0.84 at the yearly scale). Such a result is reflected in the good skill of estimated 𝑁𝐹 through both data sets in showing clear track of documented drought events happened from 2000s onwards. After assessing current trends of 𝑁𝐹, the activities foreseen in the project aim at comparing amounts of freshwater potentially available against anthropogenic uses, in light of developing sustainable water management policies. To do this, data on measured water consumptions for different uses will be provided by water allocation managers operating in the Umbria region. Estimates coming from modeling platforms (e.g., Huang et al., 2018) will be considered as well. Finally, concerning agricultural water use (actually consisting in irrigation water use), satellite-derived estimates (Dari et al., 2024) will be considered. Along with this task, future projections of available 𝑁𝐹 will be developed. An overview on the rationale of the Gov4Water project is provided in Figure 1. 44 Modellazione idrologica avanzata del bacino dell’Arno per valutare la risposta delle risorse idriche agli scenari di cambiamento climatico Francesco Bressi1*, Stefano Tasin1, Marco Lompi2, Enrica Caporali2, Marco Borga3, Fabio Pilotti1, Anna Paola Lonati1, Matteo Dall’Amico1 1 Waterjade S.r.l., Trento, Italia 2 Dipartimento di Ingegneria Civile e Ambientale, Università degli Studi di Firenze, Firenze, Italia 3 Dipartimento di Territorio e Sistemi Agro-Forestali, Università degli Studi di Padova, Italia *e-mail: [email protected] Sommario La crescente pressione sulle risorse idriche, acuita dai cambiamenti climatici e territoriali, impone lo sviluppo di strumenti sempre più avanzati per la comprensione e la gestione dei sistemi idrologici. In questo contesto, è stato utilizzato un modello idrologico di nuova generazione che mira a una replica virtuale dinamica e integrata del bacino idrografico. In particolare, la sua implementazione informatica ottimizzata per le prestazioni e la sua architettura modulare consentono di creare e integrare componenti ad hoc, sia per rappresentare processi fisici specifici (es. carsismo, dinamiche glaciali), sia per la gestione avanzata del dato (es. moduli di machine learning per il pre-processing o il post-processing), permettendo di personalizzare le componenti del modello per ogni specifico progetto. Il presente contributo illustra l’applicazione del modello per il bacino idrografico del fiume Arno, un sistema fluviale di cruciale importanza nazionale. La metodologia adottata per la costruzione del modello del fiume Arno ha previsto le seguenti fasi: 1. Estrazione geomorfologica e creazione della griglia (mesh) di calcolo: a partire da Modelli Digitali del Terreno (DTM), è stata definita la discretizzazione spaziale del bacino, identificando la rete idrografica e suddividendo il territorio in Unità di Risposta Idrologica (HRU) elementari. 2. Raccolta e processamento dati meteo-idrologici: sono state acquisite, validate e processate serie storiche di dati di precipitazione, temperatura e portata ad alta risoluzione temporale. Particolare attenzione è stata posta alla pulizia dei dati, al riempimento dei dati mancanti e all’interpolazione spaziale delle forzanti meteorologiche. 3. Implementazione e calibrazione del modello idrologico semi-distribuito: è stato costruito un modello fisicamente basato, configurato sulla mesh di calcolo che sfruttando la sua modularità, permette di simulare le principali componenti del ciclo dell’acqua. La calibrazione dei parametri è stata effettuata confrontando le portate simulate con quelle osservate in diverse sezioni strumentate lungo l’asta principale e gli affluenti. Questo approccio multi-idrometrico ha permesso di calibrare progressivamente le diverse porzioni del bacino, propagando l’informazione da monte verso valle e affinando la rappresentazione dei processi idrologici nelle varie sottounità, fino alla sezione di chiusura di San Giovanni alla Vena. I risultati ottenuti dalla fase di calibrazione dimostrano l’elevata capacità del modello sviluppato nel riprodurre il regime idrologico del bacino dell’Arno. Per la sezione di chiusura di San Giovanni alla Vena, i valori di KGE (Tabella 1) superano 0,9 su diverse scale temporali (da oraria a annuale), indicando una buona corrispondenza tra deflussi simulati e osservati. L’analisi dei bilanci idrologici annuali ha inoltre permesso di validare la coerenza interna dei dati utilizzati e di quantificare le principali componenti del ciclo idrologico per il periodo storico di riferimento. I risultati così generati rappresentano una solida base conoscitiva per: • comprendere in dettaglio i pattern di ricarica e la risposta idrologica del sistema; • valutare la disponibilità idrica storica e identificare eventuali trend di stress idrico; • stabilire una baseline robusta per la valutazione degli impatti dei cambiamenti climatici. Inoltre, apre la strada a future implementazioni per sistemi di allerta delle piene (flood forecasting), per la 45 valutazione del rischio siccità e per una gestione operativa delle risorse idriche più reattiva e informata. Il lavoro presentato contribuisce quindi alla pianificazione e gestione sostenibile delle risorse idriche, fornendo una piattaforma modellistica avanzata per supportare i processi decisionali. Tabella 1. KGE calcolato tra la portata simulata e quella osservata per aggregazioni orarie, giornaliere, mensili e annuali all’idrometro di San Giovanni alla Vena (Arno) Metrica Orario Giornaliero Mensile Annuale KGE 0.90 0.91 0.95 0.9 Ringraziamenti Questa ricerca è stata sviluppata nell’ambito del progetto “RETURN - multi-Risk sciEnce for resilienT comUnities undeR a changiNg climatE” e ha ricevuto fondi dal Piano Nazionale di Ripresa e Resilienza (Mission 4, Component 2, Investement 1.3-D.D. 1243 2/8/2022, PE0000005). 46 Applicazione del sistema idrologico GEOframe per la stima di tendenze di siccità in un contesto mediterraneo: applicazione sul bacino del Crati nell'ambito del progetto Agriclima Riccardo Busti1*, Daniela Biondi2, Giuseppe Formetta1 1 DICAM, Università degli Studi di Trento, Trento 2 DIMES, Università della Calabria, Cosenza *e-mail: [email protected] Sommario La crescente vulnerabilità dei bacini mediterranei agli effetti del cambiamento climatico richiede strumenti modellistici avanzati per valutare la variazione della disponibilità idrica e le dinamiche degli eventi estremi. In questo studio viene applicato il modello idrologico semi-distribuito GEOframe al bacino del fiume Crati, in Calabria, per simulare e analizzare i processi idrologici chiave nel periodo storico 2000–2023. Vengono utilizzati diversi dataset meteorologici comprendenti stazioni a terra e/o rianalisi per rappresentare le precipitazioni e le temperature. Il modello riproduce processi come accumulo e fusione della neve, intercettazione, dinamiche dell’umidità del suolo, deflusso superficiale e sotterraneo, nonché la regolazione idrica dovuta alla presenza di dighe. La calibrazione e validazione sono condotte sulle portate osservate nelle stazioni di misura, mentre la validazione dell’umidità del suolo è effettuata con dati storici satellitari o in situ. Vengono stimati di conseguenza i bilanci idrici e gli indici di siccità idrologica e meteorologica di portate e precipitazioni al fine di identificare e analizzare gli eventi trascorsi di siccità. Figura 1. Inquadramento del bacino del fiume Crati 47 The GEOframe system deployment for the water budget quantification in the eastern mountain area of the Po river basin Gaia Roati1,2*, Marco Brian1, Giuseppe Formetta3, Shima Azimi2, Hossein Salehi2, Daniele Andreis2, John Mohd Wani2, Francesco Tornatore1, Riccardo Rigon2,3 1 Po River Basin District Authority, Parma, Italy 2 Center Agriculture Food Environment (C3A), University of Trento (Trento, Italy) 3 Department of Civil, Environmental and Mechanical Engineering (DICAM), University of Trento (Trento), Italy *e-mail: [email protected] Abstract As observed in recent years, extreme events, such as floods and droughts, have been reported to be more likely due to climate change and environmental modifications, and Italy in particular, experienced more frequent and intense drought events, with an exceptionally severe drought in 2022. In response to these hydrological phenomena and to advance the current numerical modeling framework for water resource management, the Po River Basin District Authority (ADBPO) initiated the deployment of the GEOframe modelling system across the entire district’s territory in 2021. This implementation aims to enhance the spatially and temporally distributed quantification and forecast of water availability within the district. The GEOframe modelling system (Formetta et al., 2014) is a completely open-source semi-distributed conceptual model, developed by a scientific international community led by the University of Trento, characterized by a high modularity and flexibility. Following the spatial interpolation of meteorological data and the geomorphological analysis, the model enables the simulation of all components of the hydrological balance, including evapotranspiration, snow accumulation, water storage, and runoff. The model uses the 1991–2020 period as its reference timeframe. All the simulations are carried out within the Hydrological Response Units (HRUs), namely the subbasins identified through geomorphological analysis. The selected level of detail was chosen as a good balance between simulation accuracy and the computational resources required, but can be increased on demand. Accordingly, model parameter calibration was performed using a “zonal calibration” approach, which means a calibration where the parameters were optimised separately for distinct hydrometric stations across the basin. As this represents the most computationally intensive phase of model implementation, a representative threeyear hydrological period was selected based on the availability and continuity of discharge data across different regions of the district. Calibration employed the Kling-Gupta Efficiency (KGE) metric, aiming to identify the optimal set of model parameters that best reproduces the observed discharge dynamics, minimising, then, the discrepancy between simulated water discharge evolution in time and measured hydrographs. The calibration process follows a hierarchical approach, where the parameters sets for upstream hydrometric stations, once calibrated, are held fixed during the subsequent calibration of the downstream stations to preserve upstream flow dynamics and ensure model consistency along the river network. The calibration of the model was initiated in the Valle d’Aosta region, corresponding to the uppermost section of the district, and systematically advanced downstream through the Piemonte, Emilia-Romagna, and Lombardia regions. Even if the GEOframe model will then be applied to the whole District, the actual study area is the only Po River Basin, namely the District up to Pontelagoscuro (FE) closure section. This area contains approximately 150 hydrometric stations; however, only a part of those hydrometers have been calibrated, according to data availability, reliability and representativeness of the water balance of each measuring station. More in detail, the calibrated hydrometers are 11 in Valle d’Aosta, 42 in Piemonte, 44 in Lombardia, and 30 in Emilia, trying to ensure a comprehensive representation of hydrological dynamics along the river continuum. Given the extensive area and the inherent complexity of the study domain, both in terms of hydrological variability and spatial distribution of hydrometric stations, various strategies and methodological adjustments were tested to enhance calibration performance and computational efficiency. In the Emilia-Romagna region, calibration efforts focused primarily on hydrometers located in the mountainous sectors, where streamflow is less influenced by anthropogenic factors such as withdrawals and reservoir regulation. However, the intermittent nature of the discharge regime in this area, characterized by alternating low-flow conditions and sharp flood peaks, posed significant challenges to achieving robust calibration results. Subsequently, the calibration process proceeded to the upstream portion of the Lombardia region, extending to some hydrometers in Piemonte, Trento Province, and Switzerland, since these stations are located on river 48 systems draining into three major regulated lakes in Lombardia: Lago Maggiore, Lago di Como, and Lago di Garda, to be eventually completed using the data of the outflow from the lakes down to the confluence with the main Po stream. Specifically, simulated and observed water volumes were computed and compared at key closure points, corresponding to the most downstream hydrometric stations within a subbasin. Following this step, calibration efforts were extended to include hydrometers located downstream of major lakes, focusing on those identified as the most reliable by the local regional environmental authority (ARPA Lombardia). Leveraging the full capabilities of the GEOframe modelling system, all components of the hydrological balance were simulated. This allowed for an assessment of the effects of water scarcity on agricultural systems across the basin. To enhance the model’s accuracy, the snow component will be further refined and integrated with improved parameterisation, namely, calibrating the GEOframe snow module with respect to the Snow Water Equivalent (SWE) maps produced by Dall’Amico et al. (2025), as part of the ADBPO's broader effort to characterise the hydrology of the district. After this first calibration process, the “regular” calibration, based on the comparison of simulated and measured water discharge, will take place, considering the snow parameters set as fixed. The first results of this process will be presented and analysed in this work. This ongoing development will further improve the GEOframe capability to simulate all the water balance components. Here we present, in particular, the enhancements in simulating the snow water equivalent and the snow cover. Snow is, in fact, a key component of the hydrological cycle in mountainous and high-latitude regions, functioning as a primary reservoir of precipitation and regulating downstream flow across a range of temporal scales—from daily fluctuations to seasonal trends. This regulatory function is important, especially during summer months and extended dry periods, notably in the Po River Basin, where alpine snowmelt plays a significant role in sustaining river discharge. In alpine environments, the spatial and temporal distribution of snow can strongly influence regional hydrology, with pronounced effects on both ecosystem dynamics and the management of water resources. The modular structure of the GEOframe model facilitated a collaborative approach during the calibration phase, reducing the time required and enhancing overall efficiency. This collaborative effort included the exchange of results, calibration strategies, and methodologies, which ultimately contributed to a more robust analysis of water availability, particularly in the mountainous regions of Emilia-Romagna and Lombardia. Overall, the model’s performance has been evaluated across various territories and basins within the Po River District, enabling a thorough assessment of the water cycle components and water volumes at selected verification closure points ranging from days to annual time scales. Additionally, the implementation of the snow component was further improved and validated by comparing the initial discharge estimates—derived from the preliminary calibration—with those obtained after the first calibration of snow parameters in the Valle d’Aosta basin. References Abera W., Formetta G., Borga M., Rigon R. (2017). Estimating the water budget components and their variability in a pre-alpine basin with JGrass-NewAGE. Advances in Water Resources, 104, pp. 37-54. Bancheri M., Rigon R., Manfreda S. (2020). The GEOframe-NewAge Modelling System Applied in a Data Scarce Environment. Water, 12(1), 86. Dall’Amico, M., Tasin, S., Di Paolo, F. (2025). 30-years (1991-2021) Snow Water Equivalent Dataset in the Po River District, Italy. Sci Data 12, 374. Formetta, G., Kampf, S. K., David, O., and Rigon, R. (2014). Snow water equivalent modeling components in NewAge-JGrass. Geoscientific Model Development, 7, pp. 725–736. Gupta H. V., Kling H., Yilmaz K. K., Martinez G. F. (2009). Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling, Journal of Hydrology, Volume 377, Issues 1–2, pp. 80-91. Pan X., Li X., Cheng G., Li H., & He X. (2015). Development and evaluation of a River-Basin-scale high Spatiotemporal precipitation data set using the WRF model: A case study of the Heihe River Basin. Remote Sensing, 7, pp. 9230-9252. Zanotti F., Endrizzi S., Bertoldi G., & Rigon R. (2004). The GEOTOP snow module. Hydrological Processes, 18, pp. 3667-3679. 49 RIVERTEMP classifier: un nuovo web tool per identificare e classificare i fiumi non perenni Paolo Vezza1, Isabelle Brichetto1, Carmela Cavallo2*, Christina Dolianidi3, Almudena González Costas4, Anastasios Karakostas5, Nikos Nikolaidis5, Maria Lilli5, Giammarco Manfreda1, Giovanni Negro1, Guillermo Palau-Salvador6, Maria Nicolina Papa2, Carles Sanchis-Ibor6, Spiros Tsalageorgos3 1 Politecnico di Torino, Torino, Italia 2 Università di Salerno, Salerno, Italia 3 DRAXIS Environmental SA, Thessaloniki, Grecia 4 FEMXA, Vigo, Spagna 5 Università Tecnica di Creta, Chania Creta, Grecia 6 Politecnico di Valencia, Valencia, Spagna *e-mail: [email protected] Sommario I fiumi non perenni (non-perennial rivers; NPRs) sono estremamente diffusi a livello globale e si stima che una quota sempre maggiore della rete fluviale smetterà di scorrere regolarmente nei prossimi decenni a causa dei cambiamenti climatici e dei prelievi idrici (Messager et al., 2021). Nonostante la loro ampia distribuzione, il monitoraggio e gli studi si sono tradizionalmente concentrati sui fiumi perenni, trascurando in gran parte quelli non perenni (Datry et al., 2014). Inoltre, le stazioni di misura idrometriche tradizionali non consentono di misurare in modo affidabile le portate prossime o uguali a zero in alvei con morfologia complessa e non permettono di rilevare l’eventuale presenza di pozze isolate, condizione significativa dal punto di vista ecologico (Zimmer et al., 2020; Cavallo et al., 2022). Questo ha determinato una forte carenza di informazioni sulla distribuzione dei tratti non perenni nel reticolo idrologico e sulla durata dei periodi di secca (ovvero assenza di acqua superficiale) o di stagnazione (presenza di pozze isolate) (Cavallo et al., 2022; Cavallo et al., 2025). Recentemente Cavallo et al. (2022) hanno sviluppato un metodo per la determinazione del regime di intermittenza dei NPRs basato sulla generazione di immagini in falsi colori, ottenute dalla combinazione delle bande SWIR, NIR e RED del dato multispettrale Sentinel-2 dell’Agenzia Spaziale Europea (ESA). Tali immagini permettono di distinguere con chiarezza le aree occupate dall’acqua rispetto alle altre tipologie di copertura del suolo presenti nel corridoio fluviale, consentendo così di identificare una delle tre possibili condizioni di flusso: “Flusso continuo” (o "Flowing"), caratterizzata da un deflusso superficiale continuo, (ii) “Stagnazione” (o “Ponding”), in cui la superficie bagnata è limitata a pozze d'acqua isolate non connesse tra loro, e (iii) Secca (o "Dry"), in cui il letto del fiume è caratterizzato da assenza di acqua superficiale. Alcuni esempi di immagini in falsi colori sono riportati nelle Figure a, b e c. Nell’ambito del progetto Erasmus+ RIVERTEMP è stata sviluppata una piattaforma web innovativa che implementa il metodo proposto da Cavallo et al. (2022), offrendo così uno strumento avanzato e user-friendly per l’osservazione e la caratterizzazione dei NPRs. Una volta effettuato l’accesso tramite login, l’utente è guidato attraverso una sequenza di fasi operative: (1) accede a una mappa globale della rete fluviale per individuare i tratti di NPRs di interesse; (2) disegna poligoni georeferenziati per delimitare l’area fluviale di interesse; (3) sceglie l’intervallo temporale da analizzare; (4) classifica le immagini satellitari Sentinel-2 disponibili, assegnando a ciascuna una delle tre condizioni di flusso (Flowing, Ponding o Dry). In caso di copertura nuvolosa che impedisce l’osservazione dell’alveo l’immagine verrà etichettata come “Cloudy” ed esclusa dalle analisi successive. Qualora l’utente disponga di dati raccolti durante attività di campo, come fotografie geolocalizzate o immagini acquisite da drone, questi materiali possono essere caricati direttamente nella piattaforma per confrontare le osservazioni satellitari con quelle raccolte in campo. Sulla base della classificazione supervisionata eseguita dall’utente, il sistema genera automaticamente la serie temporale delle condizioni di flusso, permettendo la caratterizzazione dei tratti fluviali in funzione della frequenza e della durata delle diverse condizioni osservate. Tra le funzionalità disponibili vi è anche la possibilità di estrarre grafici riepilogativi delle condizioni di flusso, come illustrato nella Figura g, e la classificazione degli idrotipi secondo Munné et al. (2021), riportata in Figura h. Questa metodologia è attualmente integrata nei corsi di idrologia e ingegneria fluviale offerti dalle quattro università partner del progetto RIVERTEMP. L’accesso alla piattaforma è gratuito e sono disponibili materiali didattici dedicati e un manuale d’uso. 50 Figure 1. In figura sono riportate le immagini (a), (b) e (c) in falsi colori, rappresentative delle condizioni di flusso Ponding, Dry e Flowing, confermate dalle foto geolocalizzate corrispondenti in (d), (e) e (f). L’immagine (g) mostra la serie temporale delle condizioni di flusso, mentre (h) mostra il grafico relativo agli idrotipi, entrambi prodotti dalla piattaforma web. La piattaforma web consente di archiviare le immagini satellitari classificate, insieme ai relativi metadati, all’interno di un sistema informativo geografico (GIS), basato sul database open-source PostgreSQL. L’intero dataset è facilmente consultabile attraverso la piattaforma stessa da parte di chiunque sia interessato. Ad oggi, grazie al contributo degli studenti e delle studentesse universitarie coinvolte nel progetto RIVERTEMP, sono stati analizzati e classificati oltre 300 km di fiumi non perenni distribuiti nell’Europa meridionale. L’approccio proposto permette di identificare e mappare i NPRs, consentendo di quantificare la dinamica dell’intermittenza del flusso all’interno delle reti fluviali e di supportare lo sviluppo di strategie efficaci per la gestione e la conservazione di questi ecosistemi. Bibliografia Cavallo, C., Papa, M. N., Negro, G., Gargiulo, M., Ruello, G., & Vezza, P. (2022). Exploiting Sentinel-2 dataset to assess flow intermittency in non-perennial rivers. Scientific Reports, 12(1), 21756. Cavallo, C., Sarno, L., Papa, M. N., Negro, G., Vezza, P., Ruello, G., & Gargiulo, M. (2025). Estimating dry bed periods in non-perennial rivers using Sentinel-2 satellite data. Journal of Hydrology, 133416. Datry, T., Larned, S. T., & Tockner, K. (2014). Intermittent rivers: a challenge for freshwater ecology. BioScience, 64(3), pp. 229235. Messager, M. L., Lehner, B., Cockburn, C., Lamouroux, N., Pella, H., Snelder, T., ... & Datry, T. (2021). Global prevalence of non-perennial rivers and streams. Nature, 594(7863), pp. 391-397. Munné, A., Bonada, N., Cid, N., Gallart, F., Solà, C., Bardina, M., ... Prat, N. (2021). A proposal to classify and assess ecological status in Mediterranean temporary Rivers: Research insights to solve management needs. Water, 13(6), 767. Zimmer, M. A., Kaiser, K. E., Blaszczak, J. R., Zipper, S. C., Hammond, J. C., Fritz, K. M., ... & Allen, D. C. (2020). Zero or not? Causes and consequences of zero‐flow stream gage readings. Wiley Interdisciplinary Reviews: Water, 7(3), e1436. 51 Comparing the performance of statistical models in seasonal streamflow forecasting: the case study of the Imera Meridionale River Basin, Sicily, Italy Shewandagn Lemma Tekle1,2*, Brunella Bonaccorso2, Mohamed Naim1,2 1 Advanced Studies, IUSS Pavia, Pavia, Italy 2 Department of Engineering, University of Messina, Messina, Italy *e-mail: [email protected] Abstract Seasonal streamflow forecasting is crucial for informed decision-making to achieve sustainable water management in a changing climate. In this study, we compared various statistical models applied for forecasting season-ahead inflows to the Olivo reservoir, which is mainly used for irrigation purposes. The performance of five statistical models, such as Principal Component Regression (PCR) (Li et al., 2010) Partial Least Squares regression (PLSR) (Liu et al., 2022; Thien & Yeo, 2022), Locally Weighted Principal Component Regression (LW-PCR) (Xu et al., 2022), Locally Weighted Partial Least Squares Regression (LW-PLSR) (Zhang et al., 2020), and Random Forest Regression (Sun et al., 2020) was assessed in terms of capability, simplicity, and flexibility. These models have been implemented to forecast streamflow data for three consecutive seasons (Fall SON, Winter DJF, and Spring MAM) from 1993 to 2023. The leave-one-out cross-validation (LOOCV) was applied in the tuning of the respective model parameters (Lumumba et al., 2024). In PCR and PLSR, LOOCV is applied to select the optimum number of components, while in LW-PCR and LW-PLSR, LOOCV is applied to both select the number of components and to tune the locality parameter or bandwidth. In the case of Random Forest regression, several parameters, such as the number of trees, are also tuned and optimized based on the model’s prediction error (Probst et al., 2019). Global predictors are selected based on the correlation analysis between in situ seasonal inflow data available at two gauges with global grided climatic variables such as Sea Surface temperature (SST), Sea level pressure (SLP), Geopotential height, near surface (850 hp) specific humidity, near surface (850 hp) air temperature, Meridional winds, and Zonal winds using the NOAA climate plotting and analysis tool (Table). Table 1. Selected global and local predictors for the three consecutive seasons SON Season DJF Season MAM Season Predictors Correlation Predictors Correlation Predictors Correlation Geopotential Height 0.39 Air Temperature -0.45 Zonal Wind -0.55 Specific Humidity -0.42 Meridional Wind r1 0.44 SST r1 0.42 Air Temperature r1 -0.53 Meridional Wind r2 -0.43 Meridional Wind 0.44 Air Temperature r2 -0.42 Meridional Wind r3 0.56 Air Temperature -0.50 Sea Level Pressure 0.40 SST -0.41 Specific Humidity 0.59 SST r1 0.43 Specific Humidity -0.43 SST r2 0.40 Soil Moisture 0.53 Zonal Winds r1 0.55 Streamflow 0.46 Streamflow 0.46 Zonal Winds r2 -0.41 Observed Precipitation 0.62 NAO 0.30 Observed Precipitation 0.32 Regions with statistically significant correlation values were identified in the vicinity of the selected case study, and area-weighted average time series of the variables were extracted to be used as predictors. Additionally, pre-season values of several teleconnection indices were correlated with the seasonal historical streamflow data, and indices with statistically significant correlation values were retained as potential predictors (Table 1). For local conditions, local variables such as soil moisture, both forecasted and observed rainfall, and pre-season 52 streamflow were correlated with the seasonal streamflow data to retain local variables with statistically significant correlation values. Finally, the variance inflation factor (VIF) was applied to reduce the number of predictors by removing highly collinear predictors. The results indicated that all five models are capable of adequately predicting the seasonal streamflow. However, regardless of their high computational costs, the LW-PCR and LW-PLSR are more flexible and have shown better performance and flexibility to handle non-linear relationships between the response and explanatory variables. Acknowledgements This study was partially funded by the Next Generation EU, Mission 4, Component 2, Investment 1.1 - Prin 2022 PNRR Call - DD n. 1409 14-09-2022. Project INnovative FOrecast-informed REServoir operations for sustainable use of water resources and climate change adaptation (INFORES) - CUP J53D23019320001. References Li, H. D., Liang, Y. Z. and Xu, Q. S., 2010. Uncover the path from PCR to PLS via elastic component regression. Chemometrics and Intelligent Laboratory Systems, 104(2), 341–346 Liu, C., Zhang, X., Nguyen, T. T., Liu, J., Wu, T., Lee, E. and Tu, X. M., 2022. Partial least squares regression and principal component analysis: Similarity and differences between two popular variable reduction approaches. In General Psychiatry (Vol. 35, Issue 1). BMJ Publishing Group Lumumba, V., Kiprotich, D., Mpaine, M., Makena, N. and Kavita, M., 2024. Comparative Analysis of CrossValidation Techniques: LOOCV, K-folds Cross-Validation, and Repeated K-folds Cross-Validation in Machine Learning Models. American Journal of Theoretical and Applied Statistics, 13(5), 127–137 Probst, P., Wright, M. N. and Boulesteix, A. L., 2019. Hyperparameters and tuning strategies for random forest. In Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery (Vol. 9, Issue 3). Wiley-Blackwell Sun, D., Wen, H., Wang, D. and Xu, J., 2020. A random forest model of landslide susceptibility mapping based on hyperparameter optimization using Bayes algorithm. Geomorphology, 362 Thien, T. F. and Yeo, W. S., 2022. A comparative study between PCR, PLSR, and LW-PLS on the predictive performance at different data splitting ratios. Chemical Engineering Communications, 209(11), 1439–1456 Xu, K., Fan, B., Yang, H., Hu, L. and Shen, W., 2022. Locally Weighted Principal Component Analysis-Based Multimode Modeling for Complex Distributed Parameter Systems. IEEE Transactions on Cybernetics, 52(10), 10504–10514 Zhang, X., Wei, C. and Song, Z., 2020. Fast Locally Weighted PLS Modeling for Large-Scale Industrial Processes. Industrial and Engineering Chemistry Research, 59(47), 20779–20786 53 An Open-Source Tool for Generating Hourly Synthetic Streamflow Series in Ungauged Basins Using Regional FlowDuration Curves Alan Spadoni1*, Rosanna Foraci2, Michele Di Lorenzo2, Tommaso Simonelli3, Attilio Castellarin2 1 Department of Civil, Chemical, Environmental and Materials Engineering, University of Bologna, Bologn, Italy 2 Hydro-Meteo-Climate Service of the Regional Agency for Prevention, Environment and Energy, Bologna, Italy 3 Hydraulic Risks Assessment and Management Service of the Po River Basin Authority, Parma, Italy *e-mail: [email protected] Abstract Statistical regionalization for streamflow prediction in ungauged catchments is a prolific research area. While numerous free and open-source software (FOSS) tools exist for predicting regional flow-duration curves (FDCs) at ungauged sites, a general FOSS tool specifically designed to generate continuous streamflow series from these FDCs is still lacking. This study introduces FDC2Qt, an R-package developed within a collaboration between the University of Bologna, the Po River Basin Authority (AdBPo), and the Emilia-Romagna Regional Authority (ARPAE). FDC2Qt may be used to generate long and hydrologically plausible synthetic daily and hourly streamflow series in ungauged or scarcely gauged catchments. FDC2Qt adopts a methodology consisting of three key steps outlined below and implemented as functions in the package. • First, a regional period-of-record FDC (POR-FDC) of daily streamflows is predicted for the ungauged site of interest (hereafter referred to as target site) through a regional index-flow approach; the approach estimates the curve as the product between the long-term average streamflow, 𝜇𝑄, and a regional dimensionless FDC (Castellarin et al., 2004). 𝜇𝑄 is estimated by a multi-regression model based on basin morphological and climatic descriptors, which in turn is identified using a Stepwise Regression Analysis (see panel a) (Efroymson, 1960). The regional dimensionless FDC is defined as a weighted average of dimensionless POR-FDC of gauged catchments with at least 5 complete years of daily streamflow observations (see panel b), where the weights are inversely proportional to the Euclidean distance between gauged catchments and target site in the morphoclimatic descriptors space (Burn, 1990); to reduce dimensionality of the descriptors’ space, a Principal Component Analysis is used. • Second, a daily streamflow series is generated at the target site using a non-linear interpolation method based on POR-FDCs (Smakhtin et al., 1997); after selecting a donor gauged catchment, for any given day in the donor site’s timeseries, the synthetic daily streamflow at the target site is defined as the streamflow associated with the same duration of the observed daily streamflow at the donor site (see panel c). The choice of the donor site depends on the user's purpose: if the aim is to generate hydrologically plausible synthetic timeseries, the longest daily streamflow series in the region should be selected, whereas, if one aims at reproducing historical runoff conditions, preference should be given to donor sites showing the highest synchronicity with the target site (e.g. spatial proximity). • Third, if hourly timeseries are needed, the synthetic daily series is downscaled to an hourly time step using a simple approach: a polynomial interpolation is used to describe flood events at hourly timescale, while linear interpolation is used for low flows. The rising and falling limbs of a flood event are approximated using two second-order polynomials honoring several constraints: hydrograph continuity (relative to previous and subsequent linearly interpolated values); given daily flood volumes, hourly flood peak and hydrograph shape. The hourly flood peak is obtained by regionalizing the hourly-to-daily peak ratio as a function of time of concentration Tc (see panel d), while the hydrograph shape combines the regionalization of the time-to-peak, rD, and of the flood volume reduction curve, 𝜀𝐷, for the mean duration of flood events in the target site 𝐷 (see panel e), respectively (Maione et al., 2003). Auxiliary functions included in the FDC2Qt package allow users to perform data quality assessment, by looking at precipitation elasticity of streamflow (Sankarasubramanian et al., 2001) and at the shape of empirical POR- 60 faster through injection wells and increased as porosity rose. Under transient conditions, higher porosity led to greater contaminant diffusion (Figure), although movement slowed when transitioning from the calcarenite to the limestone layer. These results highlight the influence of lithological properties on pollutant migration. Geological structures may act as natural barriers under certain conditions, helping to limit contaminant spread. The methodology applied in this study allowed for the development of a conceptual hydrogeological model of the karst aquifer, which was used to simulate mass transport under varying conditions. This approach offered a realistic representation of groundwater dynamics, essential for managing and mitigating pollution risks in karst systems, known for their high vulnerability and complex behaviours. Numerical modelling proved to be a valuable tool for systematically analysing complex hydrogeological systems, enabling the identification of vulnerable zones, such as areas affected by pollutant plumes or inverse conoids. The findings provide meaningful guidance for environmental regulations on pollutant discharge and can be applied to other karst aquifer contexts. Given the widespread distribution of karst environments, this study offers a practical framework for assessing contamination risks, supporting water resource protection and long-term environmental sustainability. Figure 1. Simulation of the transport in transient state. Scenarios of plume trend (50% - 70% - 90% limestone porosity) References Castiglioni, S., Valsecchi, S., Polesello, S., Rusconi, M., Melisa, M., Palmiotta, M., Manenti, A., Davoli, E., Zuccato, E., 2015. Sources and fate of perfluorinated compounds in the aqueous environment and in drinking water of a highly urbanized and industrialized area in Italy. https://doi.org/10.1016/j.jhazmat.2014.06.007 Chen, Z., Goldscheider, N., 2014. Modeling spatially and temporally varied hydraulic behavior of a folded karst system with dominant conduit drainage at catchment scale. Journal of Hydrology 514(6): 41-52. https://doi.org/10.1016/j.jhydrol.2014.04.005 Dillon, P., Stuyfzand, P.J., Grischek, T., Liuria, M., 2018. Sixty years of global progress in managed aquifer recharge. Hydrogeology Journal. https://doi.org/10.1007/s10040-018-1841-z Lazarova, V., Levine, B., Sack, J., Cirelli, G., Jeffrey, P., Muntau, H., 2001. Role of water reuse for enhancing integrated water management in Europe and Mediterranean countries. Water Sci. Technol 43: 25–33. https://doi.org/10.2166/wst.2001.0571 Li, F., Fang, X., Zhou, Z., Liao, X., Zou, J., Yua, B., Sun, W., 2019. Adsorption of perfluorinated acid onto soils: Kinetics, isotherms, and influences of soil properties. https://doi.org/10.1016/j.scitotenv.2018.08.209 Parise, M., Gunn, J., 2007. Natural and anthropogenic hazards in karst areas: an introduction. Geological Society, London 279(1): 1-3. https://doi.org/10.1144/sp279.1 Van Genuchten, M., Leij, F., Skaggs, T., Toride, N., Bradford, S., Pontedeiro, E., 2013. Exact analytical solutions for contaminant transport in rivers Journal of Hydrology and Hydromechanics 61(3):250-259. https://doi.org/10.2478/johh-2013-0032 61 Urban aquifer recharge and contamination by leaking sewers: a comparative analysis of different modeling approaches Andrea D’Aniello*, Dina Pirone, Luigi Cimorelli, Domenico Pianese Department of Civil, Architectural and Environmental Engineering, University of Naples Federico II, Napoli, Italy *e-mail: [email protected] Abstract Unintentional recharge by leaking sewers can have detrimental consequences on urban aquifers, like the diffusion of pathogens, soil and groundwater contamination, and the occurrence of dangerous interactions with shallow and deep anthropogenic alterations (Attard et al., 2016; Sridhar and Parimalarenganayaki, 2024). Therefore, although challenging, predicting the fate of sewer leaks and sewer-borne contaminants in the subsurface is a pressing need. Water balance calculations, indirect estimates, and a variety of physically based numerical models have been used to assess the fate of sewer leaks in the subsurface and to quantify their contribution to urban aquifer recharge. Fewer studies modeled the migration of sewer-borne contaminants in the subsurface as a result of pipe leakage. Due to the complexity of the urban subsurface, the scale of the problem, the high nonlinearity of the physical phenomena involved, and the associated computational burden, modeling efforts currently available in literature introduced one or more of the following simplifications: a 2D schematization of the problem; steady-state simulations; absence or simplification of unsaturated zone modeling; absence of anthropogenic alterations in the subsurface; fixed leak rates; fixed contaminant concentration in the leaked effluent or the adoption of a fixed contaminant mass flux, thus neglecting contaminant transport within the sewer network. Supported by a semi-distributed hydrological analysis of the combined sewer network and of the overlying catchment of an urban district, this work explored how different subsurface flow modeling approaches can predict the fate of sewer leaks and of sewer-borne contaminants in the subsurface. The US Environmental Protection Agency Storm Water Management Model - SWMM (Rossman and Simon, 2022) is used as surface flow model, sewer exfiltration at the bottom of sewer pipes is modeled using Darcy’s law (Ellis et al., 2009), whereas COMSOL Multiphysics® (COMSOL, 2024) is used as subsurface flow model. Three different high resolution physically based modeling approaches were compared. All of them provided a 3D transient and coupled simulation of the migration of sewer leaks and of related contaminants, but with a different level of detail of the physics reproduced. Carbamazepine (CBZ), a commonly prescribed drug used to treat trigeminal neuralgia and to control seizures in the treatment of epilepsy and bipolar disorders, is chosen as the reference sewer-borne contaminant, as it is widely used and approved as a stable indicator for sewer contamination of groundwater (Wolf et al., 2012). Results showed that physically consistent modeling of these phenomena is possible and feasible at large scales. A few hours were required to simulate the fate of sewer leaks and of CBZ in the subsurface over a time frame of 2 years, even including all aspects that are typically only accounted for separately in other modeling studies, like the presence of the unsaturated zone, the accurate representation of 3D pipe geometries, the presence of anthropogenic alterations that can alter sewage and contaminant migration, the computation of exfiltration rates and contaminant fluxes based on wastewater/stormwater flow within the pipes and on soil pressure heads over time, and the presence of a clogging/colmation layer within the modeled sewer defects. The three different physically based modeling approaches tested within this work showed how crucial it is to include a series of physical features when modeling sewer leaks and sewer-borne contaminants migration in the subsurface. Including unsaturated zone modeling is indeed the most important if the sewer network lies over the groundwater table. In the selected case study it takes 2 years to sewage and CBZ to barely reach the groundwater table. Conversely, the shallow aquifer is immediately affected by sewer leaks when the unsaturated zone is not accounted for, thus resulting in local groundwater table mounds and contaminant spreading directly through the aquifer. 62 Figure 1. Predicted exfiltration flow rates over time from selected sewer pipes and rainfall depth over the catchment. Although time consuming and sometimes complex at large scales, accurate representation of pipe geometry is also important. Indeed, simplifying pipe geometry to straight lines led to an overestimation of exfiltration flow rates from 13.7 to 27% on average than when pipes are represented as 3D objects, and of about 14.1% in terms of CBZ fluxes, thus resulting in higher contaminant concentrations in the subsurface. Furthermore, if pipes are represented just as lines, it is not possible to account for the potential alterations to the migration pathway of leaked sewage and contaminants induced by the presence of nearby pipes that may act as obstacles or accept sewage through infiltration, or by the presence of other anthropogenic features (D’Aniello et al., 2021, 2022). This preliminary work shows how a wise use of existing numerical models can allow coping with all the relevant phenomena involved in the prediction of the fate of sewer leaks and of sewer-borne contaminants in the subsurface, with very little simplification and with reasonable computational effort. High resolution physically based numerical models represent a viable option to help quantify the effects of sewer leaks on urban aquifer recharge and contamination and a powerful tool to support the definition of monitoring and remedial actions to preserve an endangered water resource. Acknowledgments This research was financially supported by the PRIN 2022 PNRR research project Aquifers Recharge and Contamination by Leaking Sewers – ARCLeakS (Project number P2022BMACR, CUP E53D23017100001), funded by the European Union – NextGenerationEU. References Attard, G., Winiarski, T., Rossier, Y. and Eisenlohr, L., 2016. Impact of underground structures on the flow of urban groundwater. Hydrogeology journal, 24, 5-19 COMSOL, 2024. Subsurface Flow Module User’s Guide – Version 6.3. COMSOL Multiphysics. D'Aniello, A., Cimorelli, L. and Pianese, D., 2021. Leaking pipes and the urban karst: a pipe scale numerical investigation on water leaks flow paths in the subsurface. Journal of Hydrology, 603, 126847 D'Aniello, A., Cimorelli, L. and Pianese, D., 2022. Utility trenches: sinks or barriers? Modeling the fate of leaked water in a crowded subsurface. Journal of Hydrology, 612, 128303 Ellis, J. B., Revitt, D. M., Vollertsen, J. and Blackwood, D. J., 2009. Sewer exfiltration and the colmation layer. Water Science and Technology, 59(11), 2273-2280 Rossman, L. A., and Simon, M. A., 2022. Storm Water Management Model User’s Manual Version 5.2. U.S. Environmental Protection Agency, 26 Martin Luther King Drive, Cincinnati, OH 45268. Sridhar, D., and Parimalarenganayaki, S., 2024. A comprehensive review on groundwater contamination due to sewer leakage: sources, detection techniques, health impacts, mitigation methods. Water, Air, & Soil Pollution, 235(1), 56 Wolf, L., Zwiener, C., and Zemann, M., 2012. Tracking artificial sweeteners and pharmaceuticals introduced into urban groundwater by leaking sewer networks. Science of the Total Environment, 430, 8-19 63 Individuazione delle perdite fognarie tramite l’analisi dei contaminanti emergenti negli acquiferi urbani: sviluppo di un approccio basato sull’intelligenza artificiale Dina Pirone*, Luigi Cimorelli, Daniele Martino, Domenico Pianese, Andrea D’Aniello Dipartimento di Ingegneria Civile, Edile e Ambientale, Università degli Studi di Napoli Federico II, Napoli. *e-mail: dina.[email protected] Sommario Fattori come l’usura delle condotte, i movimenti del terreno e la scarsa manutenzione delle infrastrutture fognarie possono provocare perdite non visibili che, nel tempo, possono condurre alla contaminazione degli acquiferi urbani. Pertanto, rilevare queste perdite è di fondamentale importanza per preservare la qualità delle risorse idriche sotterranee. Negli ultimi anni, numerosi studi hanno evidenziato come alcuni contaminanti emergenti, ad esempio residui di prodotti farmaceutici e i dolcificanti artificiali, possano essere impiegati come indicatori efficaci per consentire l’individuazione delle perdite fognarie (Liu et al., 2014). A differenza di altri inquinanti, come i nitrati, la presenza di contaminanti emergenti nelle falde acquifere è riconducibile inequivocabilmente alla contaminazione da acque reflue (Wolf et al., 2012). Questa caratteristica li rende validi strumenti per confermare la presenza di perdite fognarie e, dunque, per consentire di risalire alla sorgente di contaminazione sfruttando le analisi di qualità delle acque sotterranee. Tuttavia, la modellazione della migrazione dei contaminanti emergenti nelle falde acquifere presenta notevoli complessità. In primo luogo, risulta particolarmente difficile quantificare l’intensità della perdita fognaria, a causa dell’assenza di misure dirette e della variabilità spaziale e temporale dei fenomeni di esfiltrazione (Sridhar and Parimalarenganayaki, 2024). Un’ulteriore criticità è rappresentata dalla difficoltà di definire, in modo affidabile, il campo di moto delle acque sotterranee, che costituisce un elemento essenziale per prevedere il percorso e l’estensione del plume di contaminazione (D’Aniello et al., 2025). A ciò, si aggiunge che i modelli numerici dettagliati e fisicamente basati, pur essendo in grado di rappresentare con buona fedeltà i processi di migrazione dei contaminanti nel sottosuolo, richiedono tempi di calcolo elevati, rendendoli spesso poco pratici per applicazioni su larga scala o qualora sia necessario effettuare numerose simulazioni. Al contrario, modelli semplificati, pur essendo computazionalmente più rapidi, spesso non sono in grado di cogliere le dinamiche complesse tipiche della migrazione dei contaminanti nel sottosuolo (Rojas-Gómez et al., 2023). Per superare le attuali difficoltà nel rilevamento delle perdite fognarie a partire dall’analisi dei contaminanti emergenti negli acquiferi urbani, questo lavoro si propone di sviluppare un approccio innovativo basato sull’intelligenza artificiale (IA), capace di sfruttare le informazioni relative alla presenza di contaminanti emergenti nelle acque sotterranee per risalire alla sorgente della perdita, adottando un modello concettuale fisicamente basato. L’approccio proposto si articola in tre fasi principali: − Sviluppo del modello concettuale: viene proposto un modello fisicamente basato per simulare l’esfiltrazione del contaminante dalle condotte fognarie e la sua successiva migrazione nel sottosuolo, garantendo affidabilità e adattabilità a diversi contesti territoriali. Nel modello concettuale verranno fornite indicazioni sui dati necessari per stimare il tasso di perdita a partire da informazioni comunemente disponibili su scala urbana, quali numero di abitanti serviti, caratteristiche ed età delle condotte fognarie, profondità di scavo e soggiacenza della falda acquifera. − Costruzione del modello basato sull’IA: tramite algoritmi di Deep Learning, si affronterà il problema inverso, ossia la localizzazione della perdita sulla base delle informazioni di qualità rilevate nei punti di monitoraggio. Il modello IA verrà sfruttato anche per ottenere indicazioni su quante misure siano necessarie e con quale frequenza dovrebbero essere effettuate per massimizzare l’efficacia del rilevamento. − Analisi probabilistica: a ciascun tratto del sistema fognario verrà assegnata una probabilità di perdita, consentendo una valutazione quantitativa dello stato di integrità dello stesso. Tale informazione potrà essere utilizzata come proxy per la definizione di liste di priorità d’intervento, supportando decisioni strategiche in ottica di manutenzione predittiva, ottimizzazione delle risorse e riduzione del rischio di contaminazione dell’acquifero sottostante. L’approccio proposto si prefigge l’obiettivo di migliorare l’individuazione delle perdite nei sistemi fognari, così da ridurre tempi e costi di intervento, rendendo più efficiente la pianificazione delle attività di manutenzione. 64 Inoltre, questo lavoro mette ulteriormente in evidenza il valore strategico del monitoraggio dei contaminanti emergenti nelle acque sotterranee, promuovendone il potenziamento come strumento chiave per la protezione delle risorse idriche sotterranee dalla contaminazione di origine fognaria. Ringraziamenti Questa ricerca è stata finanziata dal progetto di ricerca PRIN 2022 PNRR Aquifers Recharge and Contamination by Leaking Sewers – ARCLeakS (numero progetto P2022BMACR, CUP E53D23017100001), finanziato dall’Unione Europea – NextGenerationEU. Bibliografia D’Aniello, A., Pirone, D., Cimorelli, L., Pianese, D., 2025. Urban aquifer recharge and contamination by leaking sewers: a comparative analysis of different modeling approaches. Le Giornate dell’Idrologia della Società Idrologica Italiana 2025, Bari, 8 – 10 settembre 2025 Liu, Y., Blowes, D. W., Groza, L., Sabourin, M. J., & Ptacek, C. J., 2014. Acesulfame-K and pharmaceuticals as co-tracers of municipal wastewater in a receiving river. Environmental Science: Processes & Impacts, 16(12), 2789-2795. Rojas-Gómez, K. L., Binder, M., Walther, M., & Engelmann, C., 2023. A parsimonious approach to predict regions affected by sewer-borne contaminants in urban aquifers. Environmental Monitoring and Assessment, 195(12), 1517. Sridhar, D., and Parimalarenganayaki, S., 2024. A comprehensive review on groundwater contamination due to sewer leakage: sources, detection techniques, health impacts, mitigation methods. Water, Air, & Soil Pollution, 235(1), 56. Wolf, L., Zwiener, C., and Zemann, M., 2012. Tracking artificial sweeteners and pharmaceuticals introduced into urban groundwater by leaking sewer networks. Science of the Total Environment, 430, 8-19 65 Irrigation volumes monitoring coupling an energy-water balance model with ground and satellite data Nicola Paciolla*, Chiara Corbari Dipartimento di Ingegneria Civile e Ambientale (DICA), Politecnico di Milano, Milano *e-mail: [email protected] Abstract Agriculture is the single major freshwater user worldwide, averaging 70% of the water resource consumption (Zucaro et al., 2016). Despite this heavy incidence, irrigation represents the most uncertain water flux, difficult to predict because of both anthropogenic and natural factors. Remote sensing observations of variables affecting the hydrological water cycle, like surface soil moisture (SSM), land surface temperature (LST) and vegetation indexes (e.g., NDVI, LAI) are available at increasingly higher spatial, temporal and spectral resolutions, allowing to integrate sparse observations from in-situ sampling networks. Coupling these data into a robust hydrological modelling framework via data assimilation is a promising way to shed light on poorly gauged processes like irrigation. In this presentation, different complementary approaches to estimate irrigation volumes will be discussed, built on the synergistic use of data (both in situ and from satellite) and an energy-water balance model. Two main approaches were tested at multiple spatial scales in the Capitanata Irrigation Consortium, a heterogeneous agricultural area in semi-arid Southern Italy, with a considerable dependency on irrigation for tomato, vegetables and tree crops. The first one is based on a probabilistic Montecarlo approach, whereas the second one focuses on the assimilation of satellite LST and SSM retrievals into the hydrological model. Both approaches employed the FEST-EWB (Corbari et al., 2011) energy-water balance model, which is a distributed model that closes, at every time step, both the surface energy and water balances for every pixel, computing local LST as an internal variable. This allows the model to run using only meteorological forcings and vegetation status and to use satellite observations of SSM and LST for other purposes, like calibration or data assimilation. The probabilistic approach is based on the generation of an ensemble of possible irrigation scenarios, adopting a Montecarlo approach to cover the most possible realizations, including the real ones. These irrigation scenarios are then used to run the FEST-EWB model. From the comparison of the model results and a number of observed variables, the irrigation scenarios closest to reality can be extracted. The model results are compared against different kinds of references, with increasing steps of uncertainty, from results of a benchmark simulation (fed with observed irrigation data), to in-situ measurements and satellite observations. This variety helps to provide a complete idea of the algorithm reliability, and the inclusion of satellite imagery allows to export the procedure to data-poor areas. Specifically, the variables tested in this work include surface soil moisture , deep soil moisture (SM2), evapotranspiration (ET) and land surface temperature (LST). Of these, SSM qualified as the most suited to the algorithm, as differences in irrigation timings caused little spread in the other variables ensembles. This approach has been tested for different irrigated fields. The second approach is based on the assimilation of remotely-sensed data to retrieve irrigation events over a wider area, with the hydrological model running on a mesh of 1 km spatial resolution over the whole Irrigation Consortium. Specifically, the FEST-EWB model was run in three main configurations, each accounting for irrigation in a different way: (a) a volume equivalent to the readily available water was budgeted for every vegetated pixel whenever water stress conditions were met, following FAO guidelines and providing a baseline for unstressed irrigation management (Allen et al., 2011); (b) LST data from MODIS at 1 km spatial resolution and daily revisit time were assimilated into the model, correcting the soil moisture status to match the thermodynamic conditions observed from satellite and budgeting irrigation water volumes for positive model LST biases (Corbari et al., 2025); (c) SSM data from Sentinel-1 at 1 km spatial resolution and roughly 6 days of revisit time were assimilated into the model, inducing a direct correction of the modelled soil moisture status and budgeting irrigation volumes for negative model SSM biases. These procedures were compared at field level, leading to the following main conclusions: (1) the probabilistic algorithm works best over fields with fewer irrigation events in a season (<10) and during high-ET periods, as very frequent events (>2-3 per week) crowd the signal and can make the procedure redundant; (2) for the satellite data assimilation approach, the spatial gap between satellite observations (1 datapoint per 1 km2) and average size of the local plots (on average, 7 ha, 14 times lower) is a major issue in capturing the field-scale differences, considering that the area is highly heterogeneous, with many neighboring fields running on different yearly cultivation schedules (both in terms of crop type and water management). 66 Figure 1. Study area and results for a monitoring station in a tomato field References Allen, R. G., et al. (1998). Crop evapotranspiration-Guidelines for computing crop water requirements-FAO Irrigation and drainage paper 56. Fao, Rome, 300(9), D05109. Corbari, C., et al. (2011). A distributed thermodynamic model for energy and mass balance computation: FEST– EWB. Hydrological Processes, 25(9), 1443-1452. Corbari, C., Paciolla, N., et al. (2025). Monitoring Anthropogenic Irrigation Water Use by assimilating satellite land surface temperature and soil moisture. In review. Zucaro, A., et al. (2016). Environmental loading of Italian semi-intensive snail farming system evaluated by means of life cycle assessment. Journal of Cleaner Production, 125, 56-67. 67 Bridging knowledge and practice in the WEFE Nexus: multilevel participatory modelling for sustainable irrigation in the mediterranean Virginia Rosa Coletta1*, Stefania Santoro1, Laura Selicato1, Claudia Panciera1,2, Ivan Portoghese1, Alessandro Pagano2, Raffaele Giordano1 1 Water Research Institute - National Research Council, Bari, Italy 2 Department of Civil, Environmental, Land, Construction and Chemistry, Polytechnic University of Bari, Bari, Italy *e-mail: [email protected] Abstract The Mediterranean region faces increasing water stress driven by climate change, agricultural intensification, and ecosystem degradation. While the Water-Energy-Food-Ecosystem (WEFE) Nexus approach offers a comprehensive framework to address these interdependencies, its implementation in practice remains fragmented and limited (Albrecht et al., 2018; Giordano et al., 2025). The EU-funded Project NexusLabs (PRIMA Project, GA 2422) seeks to close this gap by developing and applying integrated solutions in Living Labs across seven Mediterranean countries, including Italy. Within this framework a multi-level participatory modelling approach is proposed, to support decision-making in sustainable irrigation and water resource management under changing environmental and socio-economic conditions. The multi-level modelling addresses the complexity of the WEFE Nexus across physical, administrative, and social boundaries. It integrates different interactions between system components, the diversity of stakeholder behaviours, and the articulation of spatial and institutional scales (from local irrigation consortia to regional and Mediterranean-wide policy contexts). Participatory modelling refers to the active involvement of stakeholders throughout the modelling process, incorporating their perceptions, values, and experiential knowledge into the system representation, analysis and validation (Voinov et al., 2016). This is achieved through specific activities such as semi-structured interviews and workshops, which also facilitate shared problem understanding and social learning. The proposed modelling framework couples Participatory System Dynamics Modelling (PSDM) and AgentBased Modelling (ABM) to represent both system-level feedbacks and individual behavioural dynamics. PSDM enables simulation of macro-level dynamics, and specifically how both physical components (e.g., water flows, land use change) and intangible dimensions (such as governance structures and social norms) of the WEFE system interact over time, capturing feedback loops and unintended consequences under different scenarios (e.g., impacts of drought, water reuse policies, or land-use change) (Amorocho-Daza et al., 2024; Pagano et al., 2025). ABM, on the other hand, focuses on micro-level behavioural heterogeneity, modelling how agents (e.g., farmers, managers) make decisions, respond to incentives, and interact within social networks using the Theory of Planned Behavior (An, L., 2012; Pouladi et al., 2019; Aybuğa & Yücel Işıldar, 2022). This allows the exploration of how social structure and individual behaviours affect innovation uptake and policy effectiveness, which are factors often overlooked in conventional modelling (Mirzaei & Zibaei, 2021). The integration of PSDM and ABM is particularly valuable in the Nexus context, where macro-level dynamics and micro-level decisions are tightly coupled. Their joint use enables the co-design of scenarios of intervention that are both systemically coherent and behaviourally realistic, providing stakeholders and decision/policymakers with robust insights into the potential impacts of strategic choices. The modelling framework is being applied in different Mediterranean countries, including three Italian Living Labs, namely Torre Guaceto, Tarquinia, and the Veneto region, each with distinct challenges (e.g., nitrate contamination, over-abstraction, seasonal drought). These applications rely on inputs from hydrological, environmental, and socio-economic modelling developed in other WPs, as well as on stakeholder engagement activities and indicators datasets (e.g., the WEFE Index). A Serious Game will further support interactive exploration of Nexus dynamics under different scenarios. Beyond the NexusLabs Project, the proposed modelling framework is replicable and generalizable to other complex Nexus problems. More broadly, it represents a methodological advancement in the integrated management of coupled natural–human systems. By combining participatory approaches with hybrid modelling techniques, it allows for the joint consideration of physical processes, institutional settings, and behavioural 68 dynamics, thus supporting the design of adaptive and context-sensitive strategies for sustainable resource governance across a wide range of socio-environmental challenges. References Albrecht, T.R., Crootof, A., Scott, C.A., 2018. The Water-Energy-Food Nexus: a systematic review of methods for nexus assessment. Environmental Research Letters, 13 (4). Amorocho-Daza, H., Susnik, J., van der Zaag, P., Slinger, J.H., 2024. A model-based policy analysis framework for social-ecological systems: Integrating uncertainty and participation in system dynamics modelling. Ecological Modelling, 499, 110943. An, L., 2012. Modeling human decisions in coupled human and natural systems: Review of agent-based models. Ecological Modelling, 229, 25–36. Aybuğa, K., Yücel Işıldar, A. G., 2022. Agent-based approach on water resources management: A modified systematic review. Turkish Journal of Water Science and Management, 6(1), 1–15. Giordano, R., Osann, A., Enao, E., Lopez, L.M., Piquera, J.S., Nikolaidis, N.P., Lilli, M., Coletta, V.R., Pagano, A., 2025. Causal Loop Diagrams for bridging the gap between Nexus thinking and Nexus doing: evidence from two case studies. Journal of Hydrology, 650, 132571. Mirzaei, A., Zibaei, M., 2021. Water conflict management between agriculture and wetland under climate change: application of economic-hydrological-behavioral modelling. Water Resources Management, 35, 1– 21. Pagano, A., Coletta, V.R., Portoghese, I., Panagopoulos, A., Pisinaras, V., Chatzi, A., Malamataris, D., Babakos, K., Lilli, M.A., Nikolaidis, N.P., Giordano, R., 2025. On the use of Participatory System Dynamics Modelling for WEF Nexus management: Hints from two case studies in the Mediterranean region. Environmental Impact Assessment Review, 115, 108012. Pouladi, P., Afshar, A., Afshar, M. H., Molajou, A., and Farahmand, H., 2019. Agent-based socio-hydrological modeling for restoration of Urmia Lake: application of theory of planned behavior. Journal of Hydrology, 576, 736–748. Voinov, A., Kolagani, N., McCall, M. K., Glynn, P. D., Kragt, M. E., Ostermann, F. O., et al., 2016. Modelling with stakeholders—Next generation. Environmental Modelling & Software, 77, 196–220. 69 The quantification of energy for irrigation Davide Danilo Chiarelli1*, Paolo D’Odorico2, Aldo Fiori3, Akhil Unnikrishnan1, Ivan Lombardich1, Maria Cristina Rulli1 1 Department of Civil and Environmental Engineering, Politecnico di Milano, 20133, Milan, Italy 2 Department of Environmental Science, Policy, and Management, University of California, Berkeley, 94720, Berkeley, USA 3 Dipartimento di Ingegneria Civile, Informatica e delle Tecnologie Aeronautiche, Roma Tre University, Rome, Italy *e-mail: [email protected] Abstract Expanding irrigation presents a promising strategy to enhance agricultural productivity in response to rising global population pressures and evolving socio-political conditions (Foley et al., 2011, Mueller et al., 2012, D’Odorico et al., 2018, Godfray et al., 2010). However, effective irrigation not only depends on the availability of sufficient water resources but also requires adequate energy to distribute water efficiently across agricultural lands (Rosa et al., 2020). While numerous studies have evaluated the water needs for closing yield gaps through irrigation (Mueller et al., 2012), few have quantified the corresponding energy requirements. In this study, we introduce a globally applicable, spatially explicit, and physically based bottom-up method to estimate energy demands associated with irrigation. The analysis covers 42 crop types and three major irrigation technologies computing the need of water for irrigation using the WATNEEDS model (Chiarelli et al., 2020). Our results indicate that the current global energy demand for irrigation is approximately 1.1 × 10⁹ GJ per year. Of this, surface irrigation accounts for around 0.32 × 10⁹ GJ/year, while sprinkler systems use approximately 0.77 × 10⁹ GJ/year. Drip irrigation, though more efficient, currently represents only a minor share of total energy use. Cereal crops contribute the most to global energy consumption due to their extensive cultivation, whereas crops such as coffee, citrus, and sugarcane exhibit the highest energy use per unit area. Notably, many regions with high potential for irrigation expansion—particularly in developing countries across the Southern Hemisphere—are also challenged by energy access constraints. To explore this further, we assessed the feasibility of expanding irrigation into currently rainfed agricultural areas where crop water requirements can be sustainably met by local water resources, without depleting environmental flows or groundwater stocks. Our analysis identified approximately 115 million hectares suitable for sustainable irrigation expansion, with major opportunities in Africa (44.6 Mha), Eastern Europe (34.1 Mha), and Russia (9.8 Mha). Wheat, maize, and barley dominate these areas, collectively accounting for roughly 40 Mha. To evaluate feasibility from an energy perspective, we compared country-specific per capita energy supplies with the global average. We estimate that the additional global energy required to support sustainable irrigation expansion for the 42 crops is about 0.22 × 10⁹ GJ/year—an increase of roughly 20% over current irrigation energy demand. Strikingly, 29 of the top 50 countries with the highest projected energy demands fall below the global average energy supply per capita (IEA). The spatially explicit outputs from this study provide critical insights for the design of integrated water–energy– agriculture strategies. Such approaches can help guide sustainable agricultural development, enhance resilience, and support efforts to combat food insecurity and malnutrition. 76 Is Today Drier? A Metastatistical framework for drought frequency across Eras Maria Francesca Caruso1*, Gabriele Villarini2,3, Marco Marani1,4 1 Department of Civil, Architectural, and Environmental Engineering, University of Padova, Padova, Italy 2 Department of Civil and Environmental Engineering, Princeton University, Princeton, NJ, USA 3 High Meadows Environmental Institute, Princeton University, Princeton, NJ, USA 4 Research Center on Climate Change Impacts, University of Padova, Rovigo, Italy *e-mail: maria[email protected] Abstract As climate change alters precipitation and evapotranspiration patterns, drought events are shifting in severity, duration, and frequency across many regions worldwide. Unlike most other hydroclimatic phenomena, droughts extend over large spatial areas and persist over long timescales, requiring a deeper understanding of their spatio-temporal behavior to support effective monitoring, prediction, and adaptation strategies. Their prolonged duration poses significant challenges for historical observation, limiting our ability to trace the spatio-temporal patterns of these events. To address these challenges, this study examines the statistical properties of extreme droughts using climate model simulations from the fourth phase of the Paleoclimate Modelling Intercomparison Project (PMIP4; Jungclaus et al., 2017) and the sixth phase of the Coupled Model Intercomparison Project (CMIP6; Eyring et al., 2016). By integrating both historical and paleo-hydrological records, we assess the occurrence of extreme droughts across various geographic regions and time scales, with a particular focus on well-documented pre-industrial climate periods such as the Medieval Climate Anomaly (MCA; 900–1300 A.D.) and the Little Ice Age (LIA; 1500–1850 A.D.). An advanced non-asymptotic statistical framework (Marani and Ignaccolo, 2015; Zorzetto et al., 2016; Marra et al., 2019), which explicitly separates event intensity from frequency, is applied to more accurately capture the variability and recurrence of extreme droughts. Compared to the 100-year return period quantile derived from historical simulations, past drought events were generally less frequent and exhibited substantial variability across different climate states and historical epochs. References Eyring, V., Bony, S., Meehl, G. A., Senior, C. A., Stevens, B., Stouffer, R. J., and Taylor, K. E., 2016. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization, Geoscientific Model Development, 9, 1937–1958. DOI: 10.5194/gmd-9-1937-2016. Jungclaus, J. H., Bard, E., Baroni, M., Braconnot, P., Cao, J., Chini, L. P., Egorova, T., Evans, M., GonzálezRouco, J. F., Goosse, H., Hurtt, G. C., Joos, F., Kaplan, J. O., Khodri, M., Klein Goldewijk, K., Krivova, N., LeGrande, A. N., Lorenz, S. J., Luterbacher, J., Man, W., Maycock, A. C., Meinshausen, M., Moberg, A., Muscheler, R., Nehrbass-Ahles, C., Otto-Bliesner, B. I., Phipps, S. J., Pongratz, J., Rozanov, E., Schmidt, G. A., Schmidt, H., Schmutz, W., Schurer, A., Shapiro, A. I., Sigl, M., Smerdon, J. E., Solanki, S. K., Timmreck, C., Toohey, M., Usoskin, I. G., Wagner, S., Wu, C.-J., Yeo, K. L., Zanchettin, D., Zhang, Q., and Zorita, E., 2017: The PMIP4 contribution to CMIP6 – Part 3: The last millennium, scientific objective, and experimental design for the PMIP4 past1000 simulations, Geoscientific Model Development, 10, 4005– 4033, DOI: 10.5194/gmd-10-4005-2017. Marani, M., and Ignaccolo, M., 2015. A metastatistical approach to rainfall extremes. Advances in Water Resources, 79, 121–126, 10.1016/j.advwatres.2015.03.001. Marra, F., Zoccatelli, D., Armon, M., and Morin, E., 2019. A simplified MEV formulation to model extremes emerging from multiple nonstationary underlying processes. Advances in Water Resources, 127, 280–290, 10.1016/j.advwatres.2019.04.002. Zorzetto, E., Botter, G., and Marani, M., 2016. On the emergence of rainfall extremes from ordinary events. Geophysical Research Letters, 43 (15), 8076–8082, DOI: 101011%20.1002/2016GL069445. 77 Assessing snow water equivalent reconstruction using multisource high-resolution satellite data and hydrological models in alpine regions Michele Bozzoli1,4*, Valentina Premier5, Cristian Tonelli3,5, Giuseppe Formetta2, Giacomo Bertoldi4,6, Carlo Marin5, Mathias Bavay7 1 Center for Agriculture, Food and Environment (C3A), University of Trento, Trento, Italy 2 Department of Civil, Environmental and Mechanical Engineering, University of Trento, Trento, Italy 3 Department of Mathematics, Informatics and Geosciences, University of Trieste, Trieste, Italy 4 Institute for Alpine Environment, Eurac Research, Bolzano, Italy 5 Institute for Earth Observation, Eurac Research, Bolzano, Italy 6 Center for Climate Change and Transformation, Eurac Research, Bolzano, Italy 7 WSL Institute for Snow and Avalanche Research (SLF), Davos, Switzerland *e-mail: [email protected] Abstract Alpine regions are highly sensitive to the impacts of climate change, and snow melt dynamics plays a crucial role in their hydrological processes. A representative variable of snow melt is the snow water equivalent (SWE). However, SWE measurements are rare and limited to point scales, making it difficult to obtain accurate spatialized estimates. For this reason, remote sensing products offer a unique opportunity to provide spatialized observations. Recently, using optical remote sensing data from MODIS, Landsat and Sentinel-2, SAR data from Sentinel-1 and in situ observations, Premier et al. (2021) developed a multi-source data method to reconstruct daily snow cover area (SCA) maps at high spatial resolution (20 m). In this work, we investigate the effectiveness of combining this approach with a semi-distributed hydrological model (GEOframe) (Formetta et al., 2014) to reconstruct SWE at high spatial resolution (20 m) in the alpine catchment of Dischma, Kanton Graubünden, Switzerland (40 km2). The modeled results are compared against both observed discharge and high-resolution SWE maps reconstructed using snow depth data retrieved by airplane photogrammetry of Bührle et al. (2022) and then converted into SWE maps using the approach of Jonas et al. (2009). The GEOframe model can reproduce the observed discharge with high accuracy (KGE = 0.904, NSE = 0.823). However, being a semi-distributed model, modeled SWE spatial patterns are too coarse and less accurate. Therefore, multi-source high-resolution satellite snow products could provide valuable information to improve spatial accuracy. We find that the most effective SWE downscaling approach is based on the combination of topographic parameters and the snow cover duration estimated by the novel approach of Premier et al. (2021). Comparing SWE estimates based on this new combined approach with the observed ones, we find on average a mean bias error (MBE) of 51.42 mm and a Pearson correlation coefficient (r) of 0.740. The results suggest that our new method can reproduce the spatial patterns of the SWE quite well, but at the same time the SWE averaged by the catchment is bound to the water mass balance estimated by a hydrological model and therefore hydrologically consistent with the observed catchment averaged runoff. The presented approach could be seen from a two-fold perspective. Either a downscaling procedure to improve the capability of a semi-distributed hydrological model to estimate high-resolution SWE pattern in mountain regions, or a method to estimate SWE from multi-source satellite observations using the constraint on catchment-scale water budget coming from a hydrological model. 78 Figure 1. Comparison of observed and modelled SWE in the Dischma catchment (Kanton Graubünden, Switzerland) for the four available dates: [2019-03-16], [2020-04-06], [2021-04-16] and [2022-03-23] References Bührle, L., Ruttner, P., Marty, M., & Bu¨hler, Y. (2022). Snow depth mapping by airplane photogrammetry (2017 - ongoing). EnviDat. doi: 10.16904/envidat.418 Formetta, G., Antonello, A., Franceschi, S., David, O., & Rigon, R. (2014). Hydrological modelling with components: A gis-based open-source framework. Environmental Modelling & Software, 55, 190–200. doi: 10.5194/gmd-7-725-2014 Jonas, T., Marty, C., & Magnusson, J. (2009). Estimating the snow water equivalent from snow depth measurements in the swiss alps. Journal of Hydrology, 378 (12), 161–167. doi: 10.1016/j.jhydrol.2009.09.021 Premier, V., Marin, C., Steger, S., Notarnicola, C., & Bruzzone, L. (2021). A novel approach based on a hierarchical multiresolution analysis of optical time series to reconstruct the daily high-resolution snow cover area. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 9223–9240. doi: 10.1109/JSTARS.2021.3103585 79 Use of Snow Water Equivalent Products for Improving the Parameterisation of Hydrological Models in Northern Italy Gokhan Sarigil*, Mattia Neri, Elena Toth Department of Civil, Chemical, Environmental, and Materials Engineering (DICAM), University of Bologna, Bologna, Italy *e-mail: [email protected] Abstract Snow Water Equivalent (SWE) functions as a natural reservoir in mountainous regions, accumulating winter precipitation and releasing meltwater during warmer periods to sustain streamflow, recharge groundwater aquifers, support agricultural irrigation, and enable hydropower generation. Despite its critical importance for water resource management, direct SWE measurements are spatially and temporally limited, creating substantial challenges for hydrological assessments. Alternative approaches such as large-scale reanalysis products and rainfall-runoff models may face significant challenges including sparse data assimilation, coarse spatial resolution, simplified orographic precipitation, input data accuracy, precipitation-phase determination, and simplified snow thermodynamics. This study follows a two-phase approach to validate and improve snow monitoring across 96 catchments in Northern Italy's Alpine and Apennine catchments for the period 2010-2020. Catchments are categorized into five elevation bands to systematically analyse performance across varying orographic and climatic conditions. First, we evaluate SWE estimates from i) three reanalysis products derived from land-surface models (Table 1) and ii) the simulation of streamflow-calibrated rainfall-runoff models, against the high-resolution national reference dataset, IT-SNOW (Avanzi et al., 2023). Second, we explore strategies for integrating the IT-SNOW SWE data into model calibration, in order to improve snow process simulations. Table 1. Main characteristics of the reanalysis datasets used in this study Dataset Coverage Resolution Atmospheric model Land surface model Produced by IT-SNOW (reference) (Avanzi et al., 2023) Italy, 2011present 0.5 km, Daily Station data + Radar merge S3M CIMA ERA5-Land (Muñoz-Sabater et al., 2021) Global, 1950-present 9 km, Hourly IFS CHTESSEL ECMWF Cerra-Land (Verrelle et al., 2022) Europe, 1984-2020 5 km, Hourly HARMONIEAROME SURFEX COPERNICUS VHR-REA_IT (Raffa et al., 2021) Italy, 1981-2023 2.2 km, Hourly COSMO-CLM TERRA-URB CMCC Two conceptual hydrological models with different structural formulations are employed: the GR6JCemaneige model (Coron et al., 2017; Valéry et al., 2014) and the TUW model (semi-distributed version of HBV, Viglione and Parajka, 2019). For integrating the IT-SNOW SWE data into model calibration, three distinct multi-objective calibration strategies are implemented attempting to fit both observed flow and SWE series with the Kling-Gupta Efficiency (KGE): i) sequential optimization, where snow parameters are first calibrated with SWE data, followed by calibration of the remaining parameters using streamflow observations, ii) NSGA-II multiobjective optimization to simultaneously optimize both objectives, identifying Pareto front solutions, and iii) weighted objective functions combining streamflow and SWE fitting with varying weights. The three reanalysis products revealed significant spatial variations in SWE estimates (Figure 1). CERRALand underestimated SWE across lower elevations, but showed significant overestimation in the highest catchments. ERA5-Land significantly overestimated in high mountains, especially above 1200 m, but demonstrated the best correlation among reanalyses. VHR-REA consistently underestimated SWE at all elevations and exhibited the weakest temporal patterns. 80 Figure 1. Spatial patterns of SWE in Northern Italy (2010–2020). (a) Mean annual SWE distribution of gridded reanalysis products; (b) Bias in mean annual SWE (IT-SNOW – other estimates). Multi-objective calibration significantly improved SWE simulation performance (KGE-SWE), while maintaining good streamflow performance (KGE-Q) for both models. GR6J-Cemaneige achieved better performance metrics than TUW across most elevation bands, though TUW remained competitive for streamflow at high elevations. The calibration strategy choice had greater impact on TUW, where sequential calibration reduced streamflow performance in snow-dominated catchments. NSGA-II optimization revealed strong trade-offs between SWE and streamflow objectives in high-elevation catchments, with concave Pareto fronts indicating that improving one objective sacrificed the other. Low-elevation catchments displayed instead very flat Pareto fronts corresponding to high streamflow performance, showing that SWE simulation might be substantially improved without compromising streamflow accuracy. This research contributes to understanding performance characteristics of different reanalysis products and calibration strategies in snow-influenced hydrological regimes, providing practical guidance for snow monitoring and hydrological modeling in Mediterranean mountain regions. Future research will focus on uncertainty quantification to assess the differences of the calibration strategies for more model reliability. References Avanzi F, Gabellani S, Delogu F, Silvestro F, Pignone F, Bruno G, ... , Ferraris L (2023) IT-SNOW: a snow reanalysis for Italy blending modeling, in-situ data, and satellite observations (2010–2021). Earth System Science Data Discussions, 2022, 1-30. https://doi.org/10.5194/essd-15-639-2023 Coron L, Thirel G, Delaigue O, Perrin C & Andréassian V (2017) The suite of lumped GR hydrological models in an R package. Environmental modelling & software, 94, 166-171. https://doi.org/10.1016/j.envsoft.2017.05.002 Muñoz-Sabater J, Dutra E, Agustí-Panareda A, Albergel C, Arduini G, Balsamo G, ... & Thépaut J N (2021) ERA5-Land: A state-of-the-art global reanalysis dataset for land applications. Earth system science data, 13(9), 4349-4383. https://doi.org/10.5194/essd-13-4349-2021 Raffa, M., Reder, A., Marras, G. F., Mancini, M., Scipione, G., Santini, M., & Mercogliano, P. (2021). VHRREA_IT dataset: very high-resolution dynamical downscaling of ERA5 reanalysis over Italy by COSMOCLM. Data, 6(8), 88. https://doi.org/10.3390/data6080088 Valéry, A., Andréassian, V., & Perrin, C. (2014). 'As simple as possible but not simpler': What is useful in a temperature-based snow-accounting routine? Part 2–Sensitivity analysis of the Cemaneige snow accounting routine on 380 catchments. Journal of Hydrology, 517, 1176-1187. https://doi.org/10.1016/j.jhydrol.2014.04.058 Verrelle, A., Glinton, M., Bazile, E., Le Moigne, P., Randriamampianina, R., Ridal, M., ... & Mladek, R. (2022). CERRA-Land sub-daily regional reanalysis data for Europe from 1984 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS): Reading, UK. DOI: 10.24381/cds.a7f3cd0b Viglione, A., & Parajka, J. (2019). TUWmodel: Lumped/Semi-Distributed Hydrological Model for Education Purposes, R package version 1.1-1. 81 Previsione della resa mensile di tre sorgenti mediante un modello ibrido CNN-LSTM e SF Francesco Castaldo1,2*, Claudio Arena1, Antonio Francipane1, Leonardo Valerio Noto1 1 Dipartimento di Ingegneria, Università degli studi di Palermo, Palermo, Italy 2 Istituto Universitario di Studi Superiori, IUSS Pavia, Pavia, Italy *e-mail: [email protected] Sommario Il presente studio si concentra sulla previsione della resa mensile, R (in Mm³), di tre sorgenti situate nella provincia di Palermo, Gabriele, Risalaimi e Scillato, accoppiando tecniche di machine learning e dati di seasonal forecast (SF) di precipitazioni (P) e temperature (T). Sebbene spesso trascurate nelle analisi idrologiche, le sorgenti rivestono talvolta un ruolo strategico nell’approvvigionamento idrico, come nel caso della città di Palermo, in cui contribuiscono in maniera significativa all’approvvigionamento per usi civili. Per ottimizzare la gestione di tali risorse, è stato sviluppato in Matlab un modello ibrido CNN-LSTM (Convolutional Neural Network - Long Short-Term Memory), in grado di catturare pattern spaziali e temporali complessi dei dati idrologici. I dati meteorologici utilizzati per la calibrazione del modello sono stati ottenuti dal database BigBang di ISPRA e comprendono serie mensili di precipitazioni e temperature dal 1951 al 2023 che sono state spazialmente mediate sull’area di ricarica delle sorgenti. Le rese osservate, in milioni di metri cubi (Mm3), sono state fornite da AMAP, l’ente gestore del ciclo idrico integrato della città di Palermo, e sono disponibili per i periodi 1958–2007 e 2009–2018, con alcuni gap (es., nei primi sei mesi del 2009 per la sorgente Scillato). Per ogni sorgente, è stato creato un modello, successivamente calibrato sul periodo 1958–2007 e validato su quello 2009–2018, utilizzando in input le variabili meteorologiche di ISPRA e come target le rese osservate fornite da AMAP. I modelli addestrati sono stati utilizzati in fase previsionale, utilizzando in input i dati di SF del sistema SEAS5 dell’ECMWF, disponibili dal gennaio 1995. Tali previsioni coprono un orizzonte temporale (lead time - LT) da zero (LT0) a sei mesi (LT6), e si riferiscono alla cella spaziale compresa tra 13° e 14° E e tra 37° e 38° N, in cui ricadono le tre sorgenti e le loro aree di ricarica. Le previsioni di P e T delle SF sono state preventivamente corrette tramite una rete neurale, utilizzando come target le osservazioni di P e T di ISPRA, e quindi utilizzate come input dei modelli calibrati per prevedere le rese delle singole sorgenti per il periodo 1995–2018. Le performance dei modelli sono state valutate sulla base dell’R² e dell’RMSE, sia per la singola sorgente, sia per il sistema aggregato, considerando sia il periodo 1995– 2007, che coincide con il periodo di calibrazione dei modelli, sia il periodo 2009–2018, che coincide con quello di validazione. Come atteso, dalle metriche riportate in Tabella 1 si osservano prestazioni migliori nel periodo di calibrazione, coerentemente con il periodo di addestramento del modello, anche se le prestazioni risultano essere soddisfacenti anche nel periodo di validazione. Dall’analisi dei risultati si riscontra un decremento delle performance delle SF all’aumentare del lead time, confermando che previsioni a breve termine (LT0–LT1) risultano più affidabili rispetto a quelle a medio termine (LT5–LT6), che si riflette anche nella previsione della resa. In Figura 1 si mostra un esempio della capacità del modello CNN-LSTM nel prevedere le rese mensili delle sorgenti, costituendo un tool di previsione della disponibilità idrica che può supportare in maniera efficace decisioni strategiche nella pianificazione e gestione delle risorse per la città di Palermo. 82 Tabella 1. Metriche di performance (R² e RMSE %) del modello CNN-LSTM per la previsione delle rese mensili (R) delle tre sorgenti (GAB = Gabriele, RIS = Risalaimi, SCI = Scillato) e del sistema aggregato (TOT = somma delle tre sorgenti), calcolate per due intervalli temporali distinti: periodo coincidente fase di calibrazione (periodo C, 1995–2007) e di validazione (periodo V, 2009–2018). Per ciascun lead time (LT0–LT6) delle previsioni stagionali (SF). L’RMSE è rapportato al valore massimo osservato nella serie. LT0 LT1 LT2 LT3 LT4 LT5 LT6 GAB RMSE (periodo C) [%] 4,43% 4,61% 4,91% 5,28% 5,47% 6,03% 6,19% RMSE (periodo V) [%] 7,22% 7,49% 7,99% 8% 7,87% 7,93% 9,49% R2 (periodo C) [-] 0,72 0,7 0,66 0,6 0,57 0,46 0,43 R2 (periodo V) [-] 0,34 0,3 0,22 0,24 0,28 0,26 0,21 RIS RMSE (periodo C) [%] 5,68% 7,25% 9,26% 10,74% 12,72% 14,09% 13,03% RMSE (periodo V) [%] 8,71% 8,98% 9,88% 10,22% 9,07% 9,31% 9,03% R2 (periodo C) [-] 0,78 0,64 0,42 0,22 -0,08 -0,32 -0,13 R2 (periodo V) [-] 0,5 0,47 0,37 0,33 0,48 0,44 0,47 SCI RMSE (periodo C) [%] 3,61% 4,04% 4,42% 5,09% 5,54% 5,78% 6,49% RMSE (periodo V) [%] 7,03% 7,40% 8,47% 9,77% 10,08% 10,18% 10,20% R2 (periodo C) [-] 0,82 0,77 0,73 0,64 0,58 0,54 0,42 R2 (periodo V) [-] 0,66 0,62 0,52 0,36 0,33 0,31 0,3 TOT RMSE (periodo C) [%] 3,48% 4,24% 5,06% 5,72% 6,82% 7,32% 7,27% RMSE (periodo V) [%] 4,76% 5,21% 6,61% 7,46% 7,16% 7,12% 7,33% R2 (periodo C) [-] 0,85 0,78 0,69 0,61 0,45 0,37 0,37 R2 (periodo V) [-] 0,82 0,78 0,65 0,57 0,61 0,61 0,59 Figura 1. a) Scatterplot tra valori osservati e previsti per la sorgente Scillato nei periodi di calibrazione e validazione del modello; b) Scatterplot tra piogge osservate (BigBang) e piogge grezze SF LT0 per Scillato; c) Scatterplot tra piogge osservate (BigBang) e piogge delle SF al LT0 corrette con ANN per Scillato; d) Rese osservate e rese previste con SF di P e T corrette con ANN nei periodi corrispondenti a calibrazione e validazione per il sistema delle tre sorgenti (Gabriele, Risalaimi e Scillato) al LT0. Bibliografia Hochreiter, S., & Schmidhuber, J. (1997). Long Short-Term Memory. Neural Computation, 9(8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735 Khorram, S., & Jehbez, N. (2023). A Hybrid CNN-LSTM Approach for Monthly Reservoir Inflow Forecasting. Water Resources Management, 37(10), 4097–4121. https://doi.org/10.1007/s11269-023-03541-w 83 Exploring Validation Methodologies for Modelled Soil Moisture Using Ground Measurements and Remote Sensing Based Data Fabio Delogu1, Francesco Silvestro1, Fabio Gardella2, Giorgio Boni3, Luca Repetto2* 1 CIMA Research Foundation, Savona, Italy 2 ARPAL - Agenzia Regionale per l’Ambiente, Genova, Italy 3 Dipartimento DICCA, Università Degli Studi di Genova, Genova, Italy *e-mail: luca.[email protected] Abstract This work aims to compare different Soil Moisture products against the in-situ measurements of Volumetric Water Content (VWC). The datasets analyzed include distributed satellite-based products, specifically: the SMAP (Soil Moisture Active Passive) SPL2SMP_E dataset (ONeill et al., 2021) and the HSAF-ASCAT Soil Moisture products (e.g., H16, H103). In contrast, the model-based datasets comprise the three top-layer Soil Moisture estimates from the ECMWF and the Root Zone Soil Moisture derived from a distributed physically based hydrological model (Laiolo et al., 2016). Some datasets, such as the ECMWF soil moisture product, required the merging of two different data sources: the near-real-time H26 product and the offline product H142, whose data record extension ends in 2019. This merging was necessary to cover the full period of analysis, which begins in July 2021 with the start of the VWC ground-based observations. The in-situ measurements, instead, include nine stations with sensors at three different depths, located across the Ligurian Region as part of the OMIRL regional observatory network of the Regional Environmental Agency (ARPAL). Modelled and satellite-derived soil moisture values were extracted from grids by averaging the pixel values within a defined spatial window centred around each in-situ observation point location. Then, the measurements were synchronized with the in–situ ground observations by resampling the data to a daily frequency, after which an exponential filter was applied, if necessary (Brocca et al., 2013). The point-scale measurements were aggregated into a single VWC value per point via a weighted average between the values at different depths. Since satellite and model data represent soil moisture differently, rescaling techniques are applied to make them comparable to ground-based VWC. Commonly used strategies include min–max normalization, linear regression scaling, and various forms of cumulative distribution function (CDF) matching (Brocca et al., 2013). This study evaluates and identifies the most effective combination of aggregation and rescaling methods by assessing the accuracy of each rescaled product with different metrics (e.g. NSE, RMSE, Bias, Pearson r). The objective of this study is to lay the groundwork for the future development of an operational near real-time comparison of modeled soil moisture with the ground-based observations, to detect possible overor underestimations of the modeled soil moisture itself. 84 Figura 1. Time series of rescaled soil moisture products (ECMWF, SMAP, HMC) and observed VWC at 10 cm depth from Vara Superiore station (April 1–May 31, 2025), including daily heatmaps and accuracy metrics of the 2 months period. References Laiolo P. et al., “Impact of different satellite soil moisture products on the predictions of a continuous distributed hydrological model,” Int. J. Appl. Earth Obs. Geoinformation, vol. 48, pp. 131–145, Jun. 2016, doi: 10.1016/j.jag.2015.06.002. “Omirl Online.” Accessed: Jun. 11, 2025. [Online]. Available: https://omirl.regione.liguria.it/Omirl/#/animations Brocca, L., Melone, F., Moramarco, T., Wagner, W., & Albergel, C. (2013). Scaling and filtering approaches for the use of satellite soil moisture observations. Remote Sensing of Energy Fluxes and Soil Moisture Content, 411, 426. O’Neill P. et al., “SMAP Enhanced L2 Radiometer Half-Orbit 9 km EASE-Grid Soil Moisture, Version 5.” NASA National Snow and Ice Data Center Distributed Active Archive Center, 2021. doi: 10.5067/LOT311EZHH8S. “Products list - H SAF Official web site.” Accessed: Jun. 11, 2025. [Online]. Available: https://hsaf.meteoam.it/Products/ProductsList?type=soil_moisture 85 Dynamic Soil-Atmosphere Interactions in Mediterranean Forests: Insights from Wavelet Analysis Ilenia Murgia1*, Christian Massari2, Matteo Verdone1, Konstantinos Kaffas3, David Labat4, Daniele Penna1,2 1 Department of Agriculture, Food, Environment and Forestry, University of Florence, Florence, Italy 2 Research Institute for Geo-Hydrological Protection, National Research Council, Perugia, Italy 3 Department of Science, Roma Tre University, Rome, Italy 4 Géosciences Environnement Toulouse (GET), Université Toulouse, Toulouse, France *e-mail: [email protected] Abstract Understanding eco-hydro-meteorological (EHM) interactions and their feedback mechanisms is crucial for comprehending hydrological dynamics in forested catchments, particularly in Mediterranean regions characterized by strong climatic seasonality and complex topography. While traditional time series analyses often oversimplify these complex, non-stationary dynamics, our study employs wavelet-based analysis to investigate the multi-scale interactions and crucial lead-lag relationships among precipitation, vapor pressure deficit, soil moisture, and sap flow. We hypothesize that EHM interactions display threshold-dependent switching behavior, where the direction and intensity of feedback—and consequently, lead-lag dynamics—shift in response to critical environmental conditions such as soil moisture availability and atmospheric demand, varying across topographic positions within a Mediterranean forested catchment. Utilizing continuous wavelet transform and wavelet coherence on data from the Lecciona experimental hillslope in central Italy (Tuscany), we captured distinct temporal and periodic dynamics of EHM variables. Our findings reveal that hillslope soil moisture exhibits rapid, event-driven responses to precipitation, with recharge mechanisms varying between wet and dry spells, showing greater sensitivity to climatic forcing compared to riparian soil moisture, which demonstrates smoother, more buffered responses. Notably, sap flow displays intense daily cycles, and hillslope sap flow experiences a significant reduction during summer drought. Crucially, our wavelet coherence analysis illuminates dynamic shifts in lead-lag relationships. For instance, during wet periods, precipitation often leads to rapid increases in soil moisture on hillslopes. However, under dry conditions - especially in hillslope locations - we observe instances where soil moisture fluctuations can precede precipitation events or where VPD increasingly influences transpiration dynamics, with soil moisture lagging. In contrast, riparian positions, sustained by subsurface water, maintain soil moisture as a more consistent leading driver of sap flow and VPD during dry periods. This research demonstrates the efficacy of wavelet-based analysis in identifying temporal shifts in lead-lag dynamics and feedback mechanisms within forest soil-atmosphere interactions across various time and frequency scales. Such insights are crucial for enhancing our understanding of forest ecosystem resilience and for informing more effective water and forest resource management strategies in Mediterranean climates, where water scarcity and the effects of climate change are becoming increasingly pronounced. 92 Modeling Streamflow under Data Scarcity: Regionalization of IHACRES Using Catchment Attributes in a Mediterranean Setting Francesco Alongi*, Caterina Alonzo, Antonio Francipane, Leonardo Valerio Noto Department of Engineering, University of Palermo, Palermo, Italy *e-mail: [email protected] Abstract Rainfall-runoff (RR) models are fundamental tools in hydrology, enabling the simulation and prediction of watersheds responses to precipitation and other climatic inputs. These models represent key hydrological processes (e.g., infiltration, surface runoff, evapotranspiration) and are widely applied in water resource management, flood forecasting, drought assessment, and climate change impact analysis. Regardless of whether they are conceptual, empirical, or physically based, RR models rely on a set of parameters that regulate the representation of hydrological processes within the model. Parameters must be calibrated to ensure models can reproduce streamflow dynamics accurately, minimizing discrepancies between simulated and observed discharge. However, the calibration process is highly dependent on the availability, continuity and quality of long-term hydrometeorological records, including precipitation, temperature, and streamflow data. In gauged catchments, where discharge observations are available, calibration can be performed directly. In contrast, ungauged basins lack such data, making conventional calibration unfeasible. In these cases, model parameters must be estimated through alternative strategies, such as regionalization techniques, empirical relationships, or analogies with hydrologically similar gauged basins (Samuel et al., 2011; Arsenault et al., 2019). These approaches assume that catchments sharing similar climatic, topographic, and land-use characteristics also exhibit similar hydrological behavior, allowing parameter values to be transferred using similarity metrics. Although such indirect estimation methods extend the applicability of RR models to data-scarce regions, they inherently introduce uncertainty due to lack of direct observations. Among these approaches, regionalization techniques specifically aim to mitigate this uncertainty by establishing statistical or conceptual relationships between model parameters and measurable catchment attributes (e.g., basin area, slope, soil and vegetation types, climatic conditions). The analysis presented in this study was carried out using the IHACRES model (Identification of Hydrographs And Components from Rainfall, Evaporation, and Streamflow; Jakeman & Hornberger, 1993), a lumped conceptual rainfall-runoff model specifically designed to simulate streamflow responses using limited climatic input data. IHACRES is based on the transformation of rainfall into streamflow through a simplified yet effective representation of catchment-scale hydrological processes. Its structure is particularly suited for applications in data-scarce environments, due to its parsimony, flexibility, and low computational demand (Ye et al., 1997; Lotfirad et al., 2019). The model consists of two main modules: (i) a non-linear loss module, which converts raw rainfall into effective rainfall by accounting for soil moisture dynamics and temperature-dependent evapotranspiration losses, and (ii) a linear routing module, based on unit hydrographs, which translates the effective rainfall into streamflow at the catchment outlet. IHACRES has demonstrated strong adaptability across a wide range of climatic and hydrological settings, and it is especially effective in semi-arid and arid regions, including catchments with intermittent or ephemeral streamflow (Croke et al., 2007). These characteristics make it a particularly relevant choice for modeling hydrological processes in Mediterranean environments. In this study, the IHACRES model was applied to a selection of gauged catchments distributed across Sicily (Italy), chosen to represent the diversity of climatic, morphological, and hydrological conditions characterizing the region. For the purposes of calibration and validation, a comprehensive and long-term hydrometeorological dataset was employed. This dataset includes daily time series of precipitation, air temperature, and streamflow spanning the period 1951-1997, and was made available by the River Basin Authority of the Hydrographic District of the Sicily Region. The calibration of the IHACRES model was performed using a multi-objective approach designed to achieve a balanced performance across multiple hydrological evaluation criteria. To determine the optimal set of model parameters, a ranking-based strategy was employed: for each simulation, model results were ranked according to each performance metric, and the final solution was selected as the best compromise among the top-performing configurations. The adopted approach yielded reliable results during the calibration phase, with accurate reproduction of observed hydrographs, flow volumes, and FDCs. Furthermore, the robustness of the calibrated parameter sets was confirmed during the validation phase, demonstrating the model’s ability to generalize to independent data. 93 Following the successful calibration and validation of the IHACRES model, the aim was to extend its applicability to ungauged Sicilian basins (i.e., model regionalization), where direct streamflow observations are unavailable (Figure 1). As a fundamental preliminary step, a correlation analysis was conducted to explore relationships both among the calibrated model parameters themselves and between parameters and key physiographic and climatic characteristics of the catchments (i.e., catchment area, mean slope, elevation, soil texture, land cover types, and climatic indicators, such as mean annual precipitation and temperature). Building upon these insights, the study considered some regionalization approaches. The performance of each technique was assessed using statistical metrics (Mihret et al., 2025). This comparative analysis provided useful insights into the relative strengths of the different methods in terms of predictive reliability and robustness, with the goal to identify the most effective regionalization technique for a reliable model parameter estimation in the ungauged Sicilian catchments. Figure 1. Flowchart of the IHACRES model regionalization strategy. References Arsenault, R., Breton-Dufour, M., Poulin, A., Dallaire, G., & Romero-Lopez, R. (2019). Streamflow prediction in ungauged basins: Analysis of regionalization methods in a hydrologically heterogeneous region of Mexico. Hydrological Sciences Journal, 64(11), 1297–1311. https://doi.org/10.1080/02626667.2019.1639716 Croke, B. F. W., & Jakeman, A. J. (2007). Use of the IHACRES rainfall-runoff model in arid and semi-arid regions. In H. Wheater, S. Sorooshian, & K. D. Sharma (A c. Di), Hydrological Modelling in Arid and SemiArid Areas (1a ed., pp. 41–48). Cambridge University Press. https://doi.org/10.1017/CBO9780511535734.005 Jakeman, A. J., & Hornberger, G. M. (1993). How much complexity is warranted in a rainfall‐runoff model? Water Resources Research, 29(8), 2637–2649. https://doi.org/10.1029/93WR00877 Lotfirad, M., Salehpoor Laghani, J., & Ashrafzadeh, A. (2019). Using the IHACRES model to investigate the impacts of changing climate on streamflow in a semi-arid basin in north-central Iran. Journal of Hydraulic Structures, 5(1). https://doi.org/10.22055/jhs.2019.27816.1090 Mihret, T. T., Zemale, F. A., Worqlul, A. W., Ayalew, A. D., & Fohrer, N. (2025). Unlocking watershed mysteries: Innovative regionalization of hydrological model parameters in data-scarce regions. Journal of Hydrology: Regional Studies, 57, 102163. https://doi.org/10.1016/j.ejrh.2024.102163 Samuel, J., Coulibaly, P., & Metcalfe, R. A. (2011). Estimation of Continuous Streamflow in Ontario Ungauged Basins: Comparison of Regionalization Methods. Journal of Hydrologic Engineering, 16(5), 447–459. https://doi.org/10.1061/(ASCE)HE.1943-5584.0000338 Ye, W., Bates, B. C., Viney, N. R., Sivapalan, M., & Jakeman, A. J. (1997). Performance of conceptual rainfall‐ runoff models in low‐yielding ephemeral catchments. Water Resources Research, 33(1), 153–166. https://doi.org/10.1029/96WR02840 94 Modellazione idrologica-idrodinamica integrata per la simulazione della risposta idrologica di un bacino di testata Pierfranco Costabile*, Carmelina Costanzo, Luca Furnari, Francesco Greco, Margherita Lombardo, Alfonso Senatore, Giuseppe Mendicino Dipartimento di Ingegneria dell’Ambiente, Università della Calabria, Rende (CS), Italia *e-mail: [email protected] Sommario Questo lavoro presenta un’analisi comparativa tra due approcci modellistici idrologici applicati alla simulazione afflussi-deflussi in un piccolo bacino di testata in ambiente mediterraneo. In particolare, sono stati messi a confronto un modello idrologico integrato, Integrated Surface-Subsurface Hydrological Model (ISSHM) e un modello Rain-on-Grid (RoG) in cui le perdite per infiltrazione sono trattate direttamente nel bilancio superficiale. Il primo, HydroCAL è un ISSHM innovativo basato sul paradigma degli Automi Cellulari (Mendicino et al., 2006; Furnari et al., 2021, 2024) mentre il secondo, UniCal (Costabile & Costanzo, 2021; Costabile et al., 2024) è un modello idrodinamico completo che risolve le equazioni di Saint-Venant bidimensionali e tiene conto delle perdite per infiltrazione mediante il modello di Green-Ampt (Costabile et al., 2023). Entrambi i modelli sono stati implementati su architetture di calcolo parallelo ad alte prestazioni, ma adottano formulazioni fisiche e numeriche differenti per rappresentare i processi idrologici superficiali e sotterranei. Lo scopo principale dello studio è valutare in che misura un modello idrologico-idrodinamico integrato di tipo RoG, che non include esplicitamente i feedback dell’umidità del suolo, possa approssimare la risposta idrologica prodotta da un ISSHM fisicamente basato. Le simulazioni hanno interessato un evento di piena verificatosi nel dicembre 2022 nel bacino del torrente Turbolo (7 km2), affluente del fiume Crati (Calabria), dotato di sensoristica attiva che fornisce in continuo dati di pioggia e di portata. La parametrizzazione e le strategie di calibrazione dei modelli sono commisurate alle finalità per le quali i modelli sono concepiti: HydroCal distingue 19 classi di uso del suolo e 7 tipi di suolo con una stratificazione verticale multi-layer. Mantenendo la stessa parametrizzazione descritta da Furnari et al. (2024), il modello ha richiesto una fase di warm-up per approssimare il deflusso di base osservato ed una calibrazione del solo parametro krgo, funzione della copertura del suolo. Il modello RoG, per il caso specifico, distingue due soli valori di scabrezza (canale e versante) e due classi di suolo (sabbie e argille). Il modello RoG è stato calibrato attraverso un approccio semplificato di tipo trial-and-error, agendo principalmente sulla conducibilità idraulica a saturazione, mantenendo costanti gli altri parametri. I risultati mostrano che, nonostante le differenze strutturali e concettuali, entrambi i modelli riproducono con buona accuratezza l’idrogramma osservato (Figura 1). Figura 1. Confronto tra portate osservate e simulate Il modello RoG è particolarmente efficace nel rappresentare la fase iniziale della piena, caratterizzata da deflussi rapidi e da un picco in prossimità dell’evento piovoso principale. La stima del tempo di picco risulta ben 95 allineata alle osservazioni, indicando che, in condizioni di risposta rapida dovuta a meccanismi di tipo hortoniano, il modello RoG può fornire una rappresentazione adeguata. Tuttavia, nella fase successiva al colmo, quando dominano processi di tipo dunniano e il contributo sotterraneo diventa più rilevante, il modello ISSHM offre prestazioni superiori, grazie alla rappresentazione esplicita della dinamica dell’umidità del suolo. Dal punto di vista spaziale, i due modelli mostrano differenze nella rappresentazione della rete di drenaggio. Il modello RoG, grazie a una discretizzazione variabile fino a 2 metri nelle aree canalizzate, evidenzia una rete più densa e continua, capace di cogliere le variazioni locali di larghezza del canale e l’eterogeneità idrodinamica su scala fine. Al contrario, la simulazione ISSHM, eseguita con una risoluzione uniforme di 20 metri, mostra una struttura della rete idrografica più frammentata, soprattutto nei corsi d’acqua secondari, ma comunque coerente nella geometria generale. Le caratteristiche delle reti di drenaggio superficiale generate rivelano un buon accordo tra i due approcci, evidenziando, coerentemente, una maggiore eterogeneità nelle altezze d’acqua simulate dal modello RoG (Tabella 1). Tabella 1. Indicatori statistici di base relativi alla rete di drenaggio superficiale generata dai due modelli (hmax > 0.05 m) Indicatore ISSHM-HydroCAL RoG-UniCal Altezza idrica massima (m) 0.41 0.58 Altezza idrica media (m) 0.12 0.13 Deviazione standard (m) 0.07 0.11 Nel complesso, i risultati indicano che il modello RoG può rappresentare in modo soddisfacente la dinamica di eventi di piena rapida, risultando particolarmente utile in contesti applicativi orientati alla valutazione del rischio. Tuttavia, per analisi più approfondite che richiedano una descrizione fisicamente fondata dei processi idrologici, in particolare nei regimi controllati dalla saturazione, il modello ISSHM risulta più adatto anche utilizzando risoluzioni spaziali non particolarmente accurate. Lo studio suggerisce una possibile complementarità tra i due approcci, aprendo la strada allo sviluppo futuro di strategie ibride capaci di integrare l’efficienza spaziale dei modelli RoG con la rappresentazione fisica completa offerta dagli ISSHM. Questa analisi preliminare fornisce un contributo originale nel colmare il divario metodologico tra modelli orientati alla previsione del rischio idraulico e modelli idrologici integrati, promuovendo una visione che superi la logica di semplice accoppiamento idrologico-idraulico. Le prospettive future includono l’estensione della sperimentazione a condizioni idrometeorologiche diverse e il miglioramento della rappresentazione dei processi idrologici nei modelli RoG per supportare sia la previsione di impatti ad alta risoluzione sia l’analisi dei processi idrologici su scala di bacino. Bibliografia Costabile, P., & Costanzo, C. (2021). A 2D-SWEs framework for efficient catchment-scale simulations: Hydrodynamic scaling properties of river networks and implications for non-uniform grids generation. Journal of Hydrology, 599, 126306. Costabile, P., Costanzo, C., Kalogiros, J., & Bellos, V. (2023). Toward street‐level nowcasting of flash floods impacts based on HPC hydrodynamic modeling at the watershed scale and high‐resolution weather radar data. Water Resources Research, 59(10), e2023WR034599. Costabile, P., Costanzo, C., Lombardo, M., Shavers, E., & Stanislawski, L. V. (2024). Unravelling spatial heterogeneity of inundation pattern domains for 2D analysis of fluvial landscapes and drainage networks. Journal of Hydrology, 632, 130728. Furnari, L., De Rango, A., Senatore, A., Mendicino, G. (2024). HydroCAL: A novel integrated surface– subsurface hydrological model based on the Cellular Automata paradigm. Advances in Water Resources, 185, 104623. Furnari, L., Senatore, A., De Rango, A., De Biase, M., Straface, S., Mendicino, G. (2021). Asynchronous cellular automata subsurface flow simulations in twoand three-dimensional heterogeneous soils. Advances in Water Resources, 153, 103952. Mendicino, G., A. Senatore, G. Spezzano, and S. Straface (2006), Three-dimensional unsaturated flow modeling using cellular automata, Water Resources Research, 42, W11419. 96 A cost-effective approach for estimating flood hazardscenarios induced by climate change Leonardo Mancusi1, Arianna Trevisiol1, Andrea Abbate1, Paola Faggian1, Raffaele Albano2* 1 Sustainable Development and Energy Sources Department, Ricerca sul Sistema Energetico RSE SpA, Milano, Italy 2 Department of Health Science, University of Basilicata, Potenza, Italy *e-mail: [email protected] Abstract Floods are a matter of concern as they are a common natural hazard in many parts of the world. For this reason, in compliance with the EU Flood Directive, implemented in Italy by Legislative Decree 49/2010 in the Flood Risk Management Plan (FRMP), the District Basin Authorities have developed river flood hazard maps to provide a knowledge tool for critical infrastructure planning. These maps, obtained by analyzing the frequency of floods in the past, characterize the national territory (at least of the main hydrographic network) for three distinct hazard scenarios depending on the return period (RP): scenario P1 (rare events typically with RP = 500 years), scenario P2 (infrequent events, RP = 100-200 years), scenario P3 (frequent events, RP = 20-50 years). Short and heavy rainfalls are the main causes of floods. Since they are expected to increase in both frequency and intensity over Italy (Faggian and Trevisiol, 2024), some studies are carried out in RSE to estimate future changes in hazards for the National Electricity Transmission Grid (NTG), due to climate change, by using precipitation data from 1971 to 2100 provided by 11 regional Euro-CORDEX climate models. Because NTG spans the entire country, it is necessary to proceed with multilevel approaches, having increasing degrees of depth and complexity, to generate hazard maps covering the entire domain of interest. This study is a first level of analysis with the aim to provide in a relatively short time a description of flood hazards at the national scale, which could be followed by more detailed studies focused on areas of highest risk, developed by means of hydrologic-hydraulic models in hydrological downscaling. The process, presented below for the sample area of the Italian region Emilia-Romagna, consists of the following points: • on the basis of climate models, for each homogeneous rainfall sub-area defined by the VAPI project (VAlutazione delle PIene project, http://www.idrologia.polito.it/gndci/Vapi.htm, last access: 6 June 2025), extreme daily rainfall levels (RL) are evaluated for the historical period (REF = 1971-2000) with RP 20, 200, and 500 years. Then, for the same RL, the RP are computed in the three future periods 2021-2050, 2041-2070, and 2071-2100; • considering that heavy rainfalls are the main cause of flooding, the change in flood frequency is assumed to be equal to the change in frequency of extreme precipitations: RP changes for P1, P2, and P3 are inferred from each climate model, applying the Extreme Value Analysis technique, and from the range of ensemble model variation the uncertainty of the results is estimated; • starting from the EU-DEM digital terrain model at 25 m of resolution, hydrologically conditioned to make the runoff paths consistent with the FRMP maps, the Geomorphic Flood Index (GFI) (Albano et al., 2025) with unit threshold value (Fig. 1a) is evaluated, hence, for each homogeneous zones, three threshold values (Fig. 1b) are calibrated to optimally reproduce the FRMP maps of P1, P2 and P3; • to evaluate hazard maps in the three future periods at RP= 20, 200, and 500 years, the so-called “lookup method” (Kimura et al., 2023) is used to find the correspondence between flood levels and RP in historical and future periods. In Tab. 1, in addition to the map of homogeneous sub-areas in Emilia-Romagna, the RP values obtained from the median of the ensemble of climate models are reported. Despite modeling dispersion (Fig. 1c), the analysis indicates a general trend toward reduced RP in the future, i.e., an increase in flooding frequency in comparison with REF. For each future 30-year period, by interpolating the data in Tab. 1, for RP 20, 200 and 500 years, the corresponding historical RPs are calculated. Then, from the GFI vs. RP relationship in the historical period, threshold values of GFI in the future are inferred by using the RPs calculated in the previous step. Finally, from the new GFI values, future flood areas P1, P2 and P3 are deduced (Fig. 1d). 97 Table 1. Map of homogeneous sub-areas in Emilia-Romagna and related values of RPs, calculated as the median of the model ensemble, in the periods 2021-2050, 2041-2070, 2071-2100 and corresponding to RP 20, 200, 500 years in REF. Future Periods 2021-2050 2041-2070 2071-2100 Zones / RPs 20 200 500 20 200 500 20 200 500 BoAnPi_C 12 64 104 11 40 69 7 55 118 BoAnPi_B_Reno 15 81 141 12 44 71 20 69 126 BoAnPi_A 11 82 85 10 39 63 8 65 59 BoAnPi_D 10 39 56 9 69 110 6 25 37 BoAnPi_E 17 115 213 9 40 56 6 36 91 BoAnPi_D_Reno 15 77 145 9 45 38 7 67 181 BoAnPi_B 19 83 188 11 95 186 12 107 133 BoAnPi_C_Reno 15 92 168 12 41 68 12 83 211 BoAnPi_E_Reno 17 117 134 12 57 91 9 42 96 a) b) c) d) Figure 1. a) Map of GFI with unitary threshold; b) Map of GFI threshold values calibrated against the P2 scenario of the FRMP; c) Boxplot of modeled RPs for four homogeneous area in the period 1971-2000 (blue), 2001-2030 (green), 20212050 (yellow), 2041-2070 (orange ), and 2071-2100 (red) corresponding to RP=200 year in REF; d) Map of floodable areas obtained through GFI for the historical P2 scenario and for 2021-2050 and 2041-2070 projections. Acknowledgements This work has been financed by the Research Fund for the Italian Electrical System under the Three-Year Research Plan 2025-2027 (MASE, Decree n.388 of November 6th, 2024), in compliance with the Decree of April 12th, 2024”. References Albano, R. and Adamowsky, J. (2025) Use of digital elevation models for flood susceptibility assessment via a hydrogeomorphic approach: A case study of the Basento River in Italy, Nat. Haz., 121, 8, 104758; Faggian, P., Trevisiol, A. (2024) Climate extreme scenarios affecting the Italian energy system with a multihazard approach. Bull. of Atmos. Sci.& Technol., 5, 4; Kimura, Y., Hirabayashi, Y., Kita, Y., Zhou, X., and Yamazaki, D. (2023) Methodology for constructing a floodhazard map for a future climate, Hydrol. Earth Syst. Sci., 27, 1627–1644. 98 An index-based approach to identify areas with low flood risk sustainability Daniela Biondi*, Danilo Spina, Giovanna Capparelli, Francesco Cruscomagno Department of Computer, Modelling, Electronics and Systems Engineering (DIMES), University of Calabria, Rende (CS), Italy *e-mail: [email protected] Abstract The increasing frequency and intensity of extreme weather events, in the context of global changes, highlights the urgency of understanding and mitigating hydrogeological risks. Accurate identification of areas exposed to potential flooding plays a key role in formulating land management strategies and defining adaptation policies. Within the framework of the PON Governance and Institutional Capacity 2014-2020 program, a definition of Areas with Lower Risk Sustainability (ALRS) was introduced: "Areas where critical hydrogeological and hydraulic phenomena could occur, mostly characterized by high kinetic energy, and where conditions of exposure and vulnerability exist that make it difficult to promptly and effectively implement measures to safeguard public and private safety and assets" (Versace et al., 2021a). As part of the “Tech4You - Technologies for climate change adaptation and quality of life improvement” project, an Individual Flood Risk (IFR) index was developed using an empirical, index-based approach. This methodology was tested in the urban area of Cosenza, a municipality intersected by a dense hydrographic network, primarily the Crati and Busento rivers. The IFR index represents the risk to which an individual is exposed, providing a comprehensive measure of flood-related threats. It can take values between zero and one, with higher values indicating higher risk. The quantitative evaluation of IFR is framed within a broader framework, called QUEST (QUantitative ESTimator) (Versace et al., 2021b). Flood risk is conceptually structured into several components that align with the main terms of the risk equation—hazard, vulnerability, exposure, and coping capacity—building upon the formulation of UNISDR (2009), which expands on that of Varnes (1984): • Hazard refers to the likelihood of flood related processes that may result in loss of life, injury, or other health impacts, as well as property damage. • Exposure indicates the presence of people, infrastructure, housing, production capacities and other tangible assets located in hazard-prone areas. • Vulnerability refers to the susceptibility of communities, assets, or systems to the impacts of hazards, depending on conditions shaped by physical, social, economic, and environmental factors or processes. • Coping Capacity is a behavioural capacity of people, institutions, organizations and systems, using available skills, values, beliefs, resources, and opportunities, to address, manage and overcome adverse conditions in the short to medium term. The scores for hazard, exposure, vulnerability, and coping capacity range from 0 to 1. It should be noted that, in the calculation of the index, coping capacity incorporates “positive” indicators into the analysis, helping to mitigate the impact of the event and effectively reducing the risk value. The way in which risk is calculated, is therefore based on the following formulation: Risk = Hazard x Exposure x Vulnerability x (1 – Coping Capacity) (1) The estimation of the IFR index follows a systematic, step-by-step process involving the cascading analysis of approximately 50 indicators (including direct measures, proxies, and composite indices), grouped into 24 attributes and 8 domains that characterize the key components of the risk equation. The construction of the index applies various aggregation rules, such as selecting the best or worst indicator, as well as using weighted arithmetic and geometric averages. Indicators are first estimated and normalized, then aggregated hierarchically in a bottom-up manner through successive levels. The critical scenario considered was a potential river overflow caused by a 500-year return period flood event in the study area. A 2D modeling framework was adopted, using HEC-RAS 6.6 to simulate floodable areas and event intensity (e.g., water depth and flow velocity). The evaluation area for the index also includes zones historically affected by flooding (see Fig. 1a), as documented 99 in official sources (e.g. PAI and PGRA). Figure 1b shows the IFR index results at a 5×5 m grid resolution for the urban area of Cosenza. Data originates from various sources, including existing databases (e.g., ISTAT), thematic maps, technical reports, and ad hoc surveys and studies. Additionally, satellite-derived datasets (such as the Copernicus Urban Atlas) were used to obtain detailed and up-to-date information on land use and urban morphology. Figure 1c shows the detail of the confluence between the two rivers. The IFR index can serve multiple purposes beyond risk assessment. It can be used to prioritize areas for emergency response, guide the allocation of resources, and support the evaluation of non-structural measures (such as early warning systems or land-use planning policies). Moreover, the index allows for the comparison of different risk scenarios, including the effects of climate change or future urban development. Figure 1. a) Floodable areas, including historically flooded zones (various sources); b)IFR index at a 5×5m grid scale; c) IFR index detail at the confluence of the Crati and its left-bank tributary, the Busento River. Acknowledgments This work was funded by the Next Generation EU - Italian NRRP, Mission 4, Component 2, Investment 1.5, call for the creation and strengthening of ’Innovation Ecosystems’, building ’Territorial R&D Leaders’ (Directorial Decree n. 2021/3277) - project Tech4You - Technologies for climate change adaptation and quality of life improvement, n. ECS0000009. This work reflects only the authors’ views and opinions, neither the Ministry for University and Research nor the European Commission can be considered responsible for them. References Varnes, D.J., 1984. Landslide Hazard Zonation: A Review of Principles and Practice. Natural Hazards. UNESCO, Paris. Versace, P., De Luca, D. L., De Santis, D., Penna, G. and Spina, D., 2021a. Linee guida per l’identificazione delle aree a minore sostenibilità del rischio per gli scenari di evento tipici identificati. Report tecnico del “Programma per il supporto al rafforzamento della governance in materia di riduzione del rischio idrogeologico e idraulico ai fini di protezione civile” a valere sul PON Governance e Capacità istituzionale 2014-2020. Versace, P., De Santis, D., Penna, G., Politanò, L. and Presta, D., 2021b. Definizione di linee guida per un modello di valutazione standard della pianificazione di Protezione Civile comunale e intercomunale. Report tecnico del “Programma per il supporto al rafforzamento della governance in materia di riduzione del rischio idrogeologico e idraulico ai fini di protezione civile” a valere sul PON Governance e Capacità istituzionale 2014-2020. UNISDR (United Nations International Strategy for Disaster Reduction), 2009. UNISDR Terminology on Disaster Risk Reduction. United Nations International Strategy for Disaster Reduction (UNISDR). https://www.undrr.org/publication/2009-unisdr-terminology-disaster-risk-reduction a ) b c 100 L’interazione tra la dinamica fluviale e del mare in aree antropizzate prone al rischio idrogeologico Antonia Longobardi1,2*, Fabio Dentale1, Albina Cuomo2, Roberta D’Ambrosio1, Angela Di Leo1, Giacomo Nicoletti1, Michele Pisani2, Anna Sansanelli1, Domenico Guida1,2 1 Dipartimento di Ingegneria Civile, Università di Salerno, Via Giovanni Paolo II, 132 Fisciano (SA), Italia 2 C.U.G.RI, Consorzio inter-Universitario per la previsione e prevenzione dei Grandi Rischi, Via Giovanni Paolo II, 132, Fisciano (SA), Italia *e-mail: [email protected] Sommario Lo studio della interazione tra le dinamiche fluviali e quelle della linea di costa rappresenta nella sua complessità una chiave di lettura indispensabile nella tutela degli ecosistemi degli ambienti di transizione e della funzione di filtro che essi svolgono. Gli ambienti di transizione sono particolarmente vulnerabili agli impatti antropici quali la realizzazione di infrastrutture, le sistemazioni idrauliche e dei versanti, elementi che agiscono anche sul rischio di allagamento in tali ambiti. Con riferimento a quest’ultimo aspetto la letteratura insegna come gli elementi chiave nella determinazione degli allagamenti possano appartenere a tre diversi e principali ambiti: fluviale (alluvioni in ambito naturale), urbano (allagamenti per eccesso di carico idraulico) e costiero (innalzamento livello del mare, maree, mareggiate) (Green et al., 2025). Nella realtà, i tre processi menzionati posso agire insieme a fornire un quarto caso, che ne rappresenta di fatto la combinazione. Ad esempio, i cicloni tropicali/extratropicali intensi possono generare forti precipitazioni che aumentano la portata dei fiumi, mentre allo stesso tempo i forti venti e le basse pressioni causano grandi mareggiate e onde. A vantaggio di semplificazione e per la complessità che implicitamente introduce, si tende a guardare al rischio di accadimento nei diversi ambiti dal punto di vista univariato, ma è nota la circostanza per la quale questa operatività può portare ad una notevole sottostima del periodo di ritorno con cui gli eventi combinati possono verificarsi (Mitu et al., 2023). L’area corrispondente alla Costa d’Amalfi (Salerno) è caratterizzata da coste alte localmente interrotte da aree litorali sabbiose sviluppate alla foce di torrenti maggiori. Eventi alluvionali e franosi hanno frequentemente interessato quest’area fin dal Medioevo, sottolineando la rapida morfodinamica dei versanti (Diodato et al., 2014). Tra questi fenomeni, il 25 ottobre 1954 la pioggia incessante di una perturbazione, dai caratteri di un ciclone, colpì l’abitato di Salerno ed i rilievi circostanti. Il valore della piovosità fu di 504 mm con una durata complessiva di 16 ore. L’area interessata dall’alluvione fu di oltre 500 km2 ed il bilancio fu gravissimo in termini di vite umane. L'evento costituisce un esempio di estremo idrologico in quanto presenta, per quasi tutte le durate, valori del periodo di ritorno superiori a 1000 anni (Braca et al., 2007). In particolare, l’abitato di Maiori, attraversato dal Torrente Regina Maior, riportò danni ingenti in seguito allo sfondamento di due tratti tombati ed al crollo delle abitazioni ed opifici ai margini del canale (Esposito et al., 2004). Il bacino del Regina Maior (30 Kmq), il cui sbocco a mare è caratterizzato da una foce armata, è per la presenza di più tratti di tombinatura che rivestono l’alveo nell’area urbanizzata fino al raggiungimento della foce, rappresentativo della complessità nella gestione di sistemi di allertamento come misure non strumentali di mitigazione del rischio in presenza di eventi estremi combinati. Il C.U.G.RI, Centro inter-Universitario per la previsione e prevenzione dei Grandi Rischi, si è fatto promotore della messa a punta di un sistema sperimentale di allertamento per il tratto tombato dell’alveo che si basa su un approccio modellistico validato da un sistematico monitoraggio sperimentale. Relativamente al tratto tombato, le portate di evento con assegnato periodo di ritorno sono state determinate attraverso l’applicazione della procedura VAPI Campania e sono state utilizzate in una modellazione monodimensionale mediante il software HECRAS per determinare i riempimenti del canale, utilizzando come condizione al contorno di valle le altezze alla foce armata determinate dalla presenza del mare. Allo scopo è stata effettuata una valutazione delle caratteristiche ondametriche, in termini di valori medio-annui (annuali/stagionali) ed estremi, attraverso i quali è stato possibile stimare, mediante il codice numerico SWAN, le condizioni d’onda sottocosta che determinano i fenomeni di innalzamento del livello medio mare. Le risultanze modellistiche verranno validate mediante una rete di sensori pluviometrici ed idro-geo-chimici, coadiuvati dalla recente installazione di una stazione LSPIV (Large Scale Particle Image Velocimetry) che mediante l'analisi del movimento tra frame consecutivi, determina il campo di velocità superficiale attraverso il quale sarà possibile determinare la portata ed anche il livello di interazione tra i livelli di acqua fiume-mare. 101 Figura 1. Il contesto territoriale (A), la stazione per il monitoraggio di livelli e velocità (B), esempio profilo di velocità della corrente in superficie (C) Bibliografia Braca G., Tranfaglia G., Esposito E., Porfido S., Violante C., Mazzarella A., 2007. Analisi meteorologica e idrogeologica dell'alluvione di Salerno del 25-26 ottobre 1954 - Atti del Convegno “Le alluvioni in Italia” (Roma, 7-8 maggio 2007). L’acqua (rivista dell’Associazione Idrotecnica Italiana), 3, pp. 51-64 51. Diodato N., Bellocchi G., Fiorillo F., Longobardi A., 2014. Historical reconstruction of erosive storms driving damaging hydrological events in Southern Italy (Bonea basin). In: Nazzareno Diodato and Gianni Bellocchi (eds), Storminess and Environmental Changes: Climate Forcing and Responses in the Mediterranean Region, Springer, pp. 179-192. Esposito E., Porfido S., Violante C., 2004. Il nubifragio dell'ottobre 1954 a Vietri sul Mare. Costa di Amalfi, Salerno" (a cura di). Pubblicazione GNDCI n.2870 - Consiglio Nazionale delle Ricerche, pp.1-382. Green J., Haigh I.D., Quinn N., Neal J., Wahl T., Wood M., Eilander D., de Ruiter M., Ward P., Camus P., 2025. Review article: A comprehensive review of compound flooding literature with a focus on coastal and estuarine regions. Natural Hazard and earth System Science 25, 747–816. Mitu M.F., Sofia G., Shen X., Anagnostou E.N., 2023. Assessing the compound flood risk in coastal areas: Framework formulation and demonstration, Journal of Hydrology, 626, 130278. 108 Analisi morfo-idrologica per lo studio degli allagamenti pluviali urbani Giulio Paradiso*, Daniele Ganora Dipartimento di Ingegneria dell'Ambiente, del Territorio e delle Infrastrutture (DIATI), Politecnico di Torino, Torino, Italia *e-mail: giuli[email protected] Sommario La gestione del rischio idraulico in ambiente urbano richiede approcci innovativi per affrontare la crescente variabilità degli eventi meteorologici estremi, accentuata dai cambiamenti climatici e dall'espansione urbana (Cea et al., 2025). L’ambiente urbano in particolare è caratterizzato da reticoli idrografici complessi ed effimeri (rii, tratti tombati, strade) che alterano la naturale direzione dei flussi specialmente in presenza di ostacoli artificiali (edifici, veicoli) e che vengono spesso trascurati nelle analisi convenzionali per la definizione del rischio idraulico. Questo lavoro propone, pertanto, una nuova metodologia, replicabile, per classificare eventi di pioggia e mappare le aree a rischio, integrando analisi topografiche, idrologiche e idrauliche. L’area di studio scelta è la l’area collinare del comune di Torino, caratterizzata da una combinazione complessa di pendenze elevate, urbanizzazione diffusa e una fitta rete rii parte in alveo e parte tombati, non rappresentati nelle classiche mappe di rischio. La metodologia si sviluppa in tre fasi principali. La prima prevede un’analisi topografica, basata su dati topografici (Modello digitale del terreno), per individuare le direzioni preferenziali di flusso e le potenziali zone di accumulo idrico. La seconda fase è focalizzata sull’analisi delle serie storiche di pioggia ad alta risoluzione temporale, utilizzate per estrarre parametri chiave quali intensità massima e media, durata, volume. Sulla base dell’analisi conoscitiva degli eventi pluviometrici che caratterizzano l’area di studio viene implementato un metodo di classificazione degli eventi di pioggia mediante l’uso di copule. La terza fase, infine, integra modelli idrologici semplificati e strumenti idraulici per simulare la propagazione dei flussi a scala di micro-bacino, valutando l’influenza di elementi urbani quali strade, muri, edifici, rete di drenaggio, canali irrigui, sulle dinamiche di deflusso. Questa metodologia consente di mappare nuove zone di pericolosità non incluse nelle mappe convenzionali, con un focus alla scala di edificio, così creando uno strumento più efficace per la pianificazione urbana e le attività di protezione civile. Infine, le analisi sopra descritte vengono valutate con diverse risoluzioni spaziali dei dati di input per definire il miglior compromesso tra accuratezza del modello, disponibilità dei dati e tempistiche di calcolo. I risultati costituiscono un passo verso un approccio integrato, basato su dati reali e osservazioni, per migliorare la gestione del rischio idraulico in ambienti urbani complessi. 109 Figura 1. Mappatura dell'allagamento pluviale determinata tramite modellazione bidimensionale in HEC-RAS con metodo Rain on Grid utilizzando un evento di pioggia osservato e rappresentazione delle direzioni di drenaggio calcolate tramite analisi topografica del DEM (Digital Elevation Model) mediante i software GRASS e R Bibliografia Cea, L., Sañudo, E., Montalvo, C., Farfán, J., Puertas, J., & Tamagnone, P. (2025). Recent advances and future challenges in urban pluvial flood modelling. Urban Water Journal, 22(2), 149–173. https://doi.org/10.1080/1573062X.2024.2446528 110 Assessment of damage scenarios due to pluvial flooding using remote sensing and object detection models Arianna Cauteruccio1, Giorgio Boni1*, Roozbeh Rajabi2, Gabriele Moser3 1 Department of Civil, Chemical and Environmental Engineering (DICCA), University of Genova, Genoa, Italy. 2 Geophysical Institute, University of Alaska Fairbanks, Fairbanks, USA 3 Department of Naval, Electrical, Electronic and Telecommunications Engineering - DITEN, University of Genova, Genoa, Italy *e-mail: [email protected] Abstract In this work, damage scenarios due to pluvial flooding are obtained by modelling the propagation of the stormwater volumes using the HEC-RAS 2D software. The case study was selected within the activities of the Return – NRRP project in the city of Genoa (Italy). It is a densely built urban catchment that was recently (September 24th, 2022) affected by pluvial flooding associated with a rainfall event characterized by a low return period (between 1.5 and 3 years). The study area is equipped with a traditional tipping-bucket rain gauge and one Smart Rainfall System (SRS), see Colli et al. (2019). Two further SRSs and two rain gauges are available close to the investigated area. This configuration allows to investigate the role of the spatial resolution of the forcing input on the flood hazard and the related impact on damage scenarios, here related to the economic losses of flooded vehicles. Figure 1. On the left overview of the investigated urban area (red portion) with indicated the position of the three rain gauges (white circles) and the three SRSs (red circles) with the associated atmospheric links (red lines). Sample of the detected vehicles using the YOLO deep learning model (central panel). On the right underestimation of the number of flooded vehicles when using different sensors and assuming the rain gauge located within the study area as the reference. The vehicles exposed to flood hazard are detected using the You Only Look Once (YOLO) deep learning model applied to aerial images at a spatial resolution of 5 cm. A total of 11340 vehicles were detected with an accuracy of 95%. Literature vulnerability curves for city cars and minivans, as a function of the water depth, were applied (Martínez-Gomariz et al., 2019). The rainfall measurements provided by the rain gauge located within the study area is assumed as the reference. Results in terms of flood hazard maps reveals that the point nature of measurements taken at rain gauge stations outside of the study area reduces to a half the flooded volume and by 10% the maximum water depth. In all cases a reduction of the extension of the flooded area is obtained but this is lower when SRSs are adopted instead of rain gauges positioned outside of the study area. The underestimation of the number of flooded vehicles varies between 6% and 14% with an underestimation of the total monetary damage of about 15% and 60%, respectively. The results reveal that SRSs thanks to the atmospheric links exploited can provide an aerial measurement of rain by capturing events characterized by high rain rate and low spatial extension, despite being characterized by a lower accuracy than traditional rain gauges, and being calibrated with rain gauges themselves. These rainfall events are the most prone to produce pluvial flooding in the study area due to the insufficient capacity 111 of the stormwater drainage network and the fast hydrological response of the urban basin (see e.g. Loglisci et al., 2024). Acknowledgements This study was carried out within the RETURN Extended Partnership and received funding from the European Union Next-GenerationEU (National Recovery and Resilience Plan – NRRP, Mission 4, Component 2, Investment 1.3 – D.D. 1243 2/8/2022, PE0000005). References Colli, M., Stagnaro, M., Caridi, A., Lanza, L.G., Randazzo, A., Pastorino, M., Caviglia, D.D., Delucchi, A. A, 2019. Field Assessment of a Rain Estimation System Based on Satellite-to-Earth Microwave Links. IEEE Trans. Geosci. Remote. Sens., 57, pp.2864–2875. Loglisci, N., Boni, G., Cauteruccio, A., Faccini, F., Milelli, M., Paliaga, G., and Parodi, A., 2024. The role of citizen science in assessing the spatiotemporal pattern of rainfall events in urban areas: a case study in the city of Genoa, Italy, Nat. Hazards Earth Syst. Sci., 24, 2495–2510. Martínez-Gomariz, E., Gómez, M., Russo, B., Sánchez, P., Montes, J-A, 2019. Methodology for the damage assessment of vehicles exposed to flooding in urban areas. J Flood Risk Management, 12:e12475. 112 Vehicles instability criterion for flood risk assessment in urban areas Omayma Amellah*, Raffaele Albano, Aurelia Sole Department of Civil Engineering, University of Basilicata, Potenza, Italy *e-mail: [email protected] Abstract Flooding are capable of destabilizing most of the objects they encounter, including vehicles. which potentially worsens the impacts of flooding especially when those cars are carried away by water flow, causing significant loss in terms of economic material, and fatalities. vehicles can lose their stability even under very shallow water depths and velocity and can turn into fatal traps for occupants. Effective flood management demands serious assessment of the instability mechanism faced by vehicles in areas prone to floods. The aim of this work is to introduce a new physic-based approach that relies on the interaction between the dynamic of water flow hydraulics and the physical body characteristics of submerged cars in flood waters. The methodology adopted is based on an integrated approach that combine complementary methods to develop a model capable of predicting vehicle stability in a given flood conditions. In order to understand and anticipate when cars lose their stability due to hydrodynamic forces and investigate the thresholds of destabilization for certain conditions of flow depths and velocities. To achieve this goal a physics based approach to analyse the forces acting on vehicles is carried out and data of experiments gives the water depth and flow velocity of incipient cars instability were integrated. A model based on the hydrodynamic forces exerted by flood water on a submerged body is developed to asses vehicle stability. Key parameters were introduced to facilitate adjustments and improve model flexibility. In order to make the computation efficient a compact criterion was formulated. The Initial estimation of the parameter were derived from physical analysis of cars then were refined based on the experimental data for accuracy purpose. with this approach specific instability limits were defined for each group category of cars. The resulting curves obtained from model developed in this work were compared with other different studies that defined stability thresholds for various categories of vehicles which validated and confirmed that the developed model is useful and accurate for flood risk assessment with regard to vehicles instability in urban areas. 113 Analisi multi-rischio e multi-livello dello stato dei Ponti Fluviali della Rete Stradale nel Bacino dell’Agri gestita dalla Provincia di Potenza: La Classe di Attenzione per Rischio Idraulico Aurelia Sole1,2, Raffaele Albano3*, Carmine Limongi1, Pietro Vuono1, Beniamino Onorati1, Giuseppe Francesco Cesare Lama1, Ruggero Ermini4, Annamaria De Vincenzo1, Domenica Mirauda3 1 DiING, Università degli Studi della Basilicata, Potenza, Italy 2 CNR-IRPI, Italy 3 DISS, Università degli Studi della Basilicata, Potenza, Italy 4 DIUSS, Università degli Studi della Basilicata, Matera, Italy *e-mail: raf[email protected] Sommario La crescente richiesta di attenzione legata alla valutazione dei rischi del patrimonio infrastrutturale italiano rappresenta una sfida cruciale per la salvaguardia delle persone e delle opere soggette al naturale deterioramento del tempo e degli agenti esterni. In tale contesto è necessario definire approcci multi-rischio, che permettano valutazioni integrate (sismico, strutturale, idraulico e geologico), e ben organizzate secondo livelli a diverso grado di analisi e di intervento. Nell’ambito del Progetto di Ricerca Tech4You, si prevede la realizzazione di un sistema di supporto multi-livello e multi-rischio per la gestione delle priorità di intervento e di manutenzione straordinarie. In questo senso il presente lavoro propone una procedura innovativa che permette di analizzare in maniera speditiva e puntuale lo stato di fatto degli attraversamenti fluviali delle strade gestite dalla Provincia di Potenza (Basilicata), indicando le azioni da mettere in campo al fine di ridurre il rischio idraulico ad essi associato. In particolare, il lavoro svolto analizza alcuni ponti su strade provinciali che ricadono nel bacino del fiume Agri (poco monitorata e di rilevanza socio-economica nel territorio) sulla base di quanto indicato dalle Linee Guida ministeriali, per definire le Classi di Attenzione e le priorità di intervento e adottando tre tipologie di rischio idraulico: sormonto, erosione localizzata ed erosione generalizzata (Brath e Montanari, 2000). Tali analisi adottano un approccio gerarchizzato che parte dal censimento delle opere (Livello 0), integrate da ispezioni in campo (Livello 1), delle definisce le specifiche Classi di Attenzione (Livello 2), per definire le tipologie di monitoraggio ed intervento in funzione della gravità dei rischi riscontrati (Livelli 3-4-5). Il metodo proposto adotta modelli di semplice applicazione ma al contempo di ottima affidabilità, come il Metodo VAPI (VAlutazione delle Piene, http://www.idrologia.polito.it/gndci/Vapi.htm) o la Formula Razionale (Kuichling 1889), per le valutazioni idrologiche e idrauliche e fa uso di un sistema semi-automatizzato che permette di definire la Classe di Attenzione in maniera rapida basandosi su una scelta rapida e intuitiva delle opzioni rilevate in fase di ispezione; successivamente l’applicazione di simulazioni idrauliche permetterà valutazioni specifiche, delle integrate dalle risultanze delle campagne di misurazione (Livello 3). La valutazione della pericolosità, della vulnerabilità e della esposizione sono stati implementati basandosi sulle LL.GG. Il Livello 3, si basa poi su simulazioni idrauliche mono-dimensionali in condizioni di moto vario per valutare gli indicatori utili a stimare la classe di pericolosità da sormonto e da fenomeni di erosione localizzata - il tirante idrico, il franco idraulico, la massima escavazione da contrazione e l’indice di erosione localizzata IEL - Con riferimento al caso di studio considerato, i risultati relativi alle indagini di Livello 1 e 2 permettono di accertare come il tratto fluviale del Torrente Sauro sia interessato da sostanziali fenomeni di movimentazione dei sedimenti che hanno determinato, in corrispondenza del ponte ubicato sulla strada provinciale S.P. Saurina al km 22+400, significativi fenomeni di erosione localizzata con la venuta a giorno dei plinti di fondazione di entrambe le pile. Il comportamento idraulico del medesimo ponte è stato analizzato con un modello idraulico-computazionale (HEC – RAS, https://www.hec.usace.army.mil/software/hec-ras/) esaminando un tratto di circa 8.5 km, caratterizzato da una pendenza media dell’1.3% e la cui conducibilità idraulica è stata assunta pari a 35.7 m1/3/s basandosi su curve granulometriche di dettaglio dedotte mediante image processing di fotografie ad alta definizione del materiale presente in alveo. Mentre l’idrogramma di piena con periodo di ritorno decennale è stato dedotto dalle analisi condotte da Fiorentino e Margiotta (1998). La profondità di scavo, in corrispondenza di ogni pila, è stata stimata attraverso la formulazione proposta dalla Colorado State University, “CSU” (Richardson et al., 1990). Le simulazioni idrodinamiche condotte hanno considerato due casi: il primo ipotizza la presenza delle sole pile aal’epoca di rilevazione del DTM utilizzato, mentre il secondo considera la condizione rilevata ad oggi ovvero il sistema plinto-pila. Gli angoli di attacco della corrente sulla pila n° 2 è stato assunto pari a 10°, mentre per la pila n°1 l’incidenza della corrente avviene con direzione quasi perpendicolare. I risultati ottenuti sono stati 114 confrontati con quelli che sono deducibili adottando le formule empiriche o semi-empiriche presenti nella letteratura scientifica (Melville, 1998; Yanmaz, 1989; Kothiary et al., 1992; Johnson, 1995; Ettema et al., 1998; Yanmaz, 2001; Arneson et al. 2012; Pandey et al., 2018). In particolare, è evidente che i fenomeni di erosione localizzata siano fortemente dipendenti dall’angolo di incidenza della corrente, dalle dimensioni della struttura di fondazione, oltreché dai parametri idraulici maggiormente influenti sui fenomeni di escavazione (granulometria, velocità, tirante, ecc.). In Fig. 1a-b, per l’opera presa in esame le profondità di erosione si attestano intorno a poco meno di 1.00 m per la pila n°1 e circa 1.50 m per la pila n°2. Le simulazioni computazionali permettono di definire con maggior precisione la profondità di scavo e ciò garantisce una più attendibile valutazione della Classe di pericolosità associata al fenomeno di erosione localizzata. Infatti, a differenza dell’approccio tradizionale, la stima dell’indice (IEL) può essere condotta considerando l’esatto valore della profondità di scavo, invece del valore pari a due volte il diametro della pila ipotizzato in fase di prima approssimazione. I risultati evidenziano, inoltre, la necessità di implementare un sistema di monitoraggio che permetta di controllare lo stato di evoluzione del fenomeno. Per il caso studio si rende inoltre necessario, in corrispondenza della sezione investigata, l’implementazione di adeguati interventi di manutenzione straordinaria finalizzati al consolidamento strutturale dell’opera di fondazione e al ripristino di un’opera di difesa idraulica. Figura 1. a) Andamento planimetrico della corrente idrica osservata durante in campo (Aprile, 2025); b) Modellazione idraulica del ponte e stima della profondità di scavo in presenza di una portata decennale. Ringraziamenti Il presente lavoro è stato svolto nell’ambito del progetto “Tech4You” - Ecosistemi dell’Innovazione, Avviso n. 3277 del 28.12.2021 – PNRR – codice progetto MUR: ECS0000009 - CUP H23C22000370006Tec4you. Bibliografia Arneson, L. A., Zevenbergen, L. W., Lagasse, P. F. & Clopper, P. E. (2012), Evaluating scour at bridges, Hydraulic Engineering Circular No. 18, 5th edition Brath, A., Montanari, A. (2000), The effects of the spatial variability of soil infiltration capacity in distributed flood modeling, Hydrological Processes 14 (15) Claps, P., Fiorentino, M. (2005). Valutazione delle Piene in Italia, Rapporto di sintesi per la regione Basilicata (bacini del versante ionico). GNDCI-CNR. Dipartimento di Ingegneria e Fisica dell’Ambiente, Università della Basilicata-Potenza, 2005. Ettema, R., Mostafa, E. A., Melville, B. W. & Yassin, A. A. (1998), Local scour at skewed piers, Journal of Hydraulic Engineering 124 (7), https://doi.org/10.1061/(ASCE)0733-9429(1998)124:7(756) Johnson, P. A. (1995), Comparison of pier-scour equations using field data, Journal of Hydraulic Engineering 121 (8), 626–629. https://doi.org/10.1061/(ASCE)0733-9429(1995)121:8(626) Kothyary, U. C., Garde, R. C. J. & Ranga Raju, K. G. (1992), Live-bed scour around cylindrical bridge piers, Journal of Hydraulic Research 30 (5), 701–715. https://doi.org/10.1080/00221689209498889 Fiorentino, M. e M.R. Margiotta. (1999) La valutazione dei volumi di piena e il calcolo semplificato dell'effetto di laminazione di grandi invasi. In G. Frega (ed.), Tecniche per la difesa dall'inquinamento, Editoriale Bios, Cosenza, pp.203-222, 1999. Linee Guida per la classificazione e gestione del rischio, la valutazione della sicurezza ed il monitoraggio dei ponti esistenti. Melville, B.W., Sutherland, A.J., (1988). Design method for local scour at bridge piers, Journal of Hydraulic Engineering, ASCE, Vol. 114, October 1988. Pandey, M., Sharma, P. K., Ahmad, Z. & Karna, N. (2018), Maximum scour depth around bridge pie in gravel bed streams, Natural Hazards 91, 819–836. https://doi.org/10.1007/s11069-017-3157-z 115 Richardson, E.V., D.B. Simons and P. Julien. (1990). Highways in the River Environment, FHWAHI-90-016, Federal Highway Administration, U.S. Department of Transportation, Washington, D.C. Rossi, F., Fiorentino, M. and Versace, P. (1984). Two‐component extreme value distribution for flood frequency analysis. Water Resources Research, 20(7), pp.847-856. Yanmaz, M. A. (1989), Time Dependant Analysis of Clear Water Scour Around Bridge Piers, Doctorate Thesis, METU, Ankora, Turkey Yanmaz, M. A. (2001), Uncertainty of local scour parameters around bridge piers, Journal Engineering and Environmental Science 25 (4), 127–137. 116 Advanced Multi-Risk Analysis of the River Bridges Status in the Provincial Road Network of Potenza (Basilicata): SegmentationGuided Machine Learning for Bridge Defect Detection in WideAngle Surface Images Aurelia Sole1, Raffaele Albano2*, Gilda Manfredi1, Giuseppe Santarsiero1, Valentina Picciano1 1 Department of Engineering, University of Basilicata, Potenza, Italy 2 Department of Health Science, University of Basilicata, Potenza, Italy *e-mail: raf[email protected] Abstract Ensuring the structural integrity of exiting bridge infrastructures across the roads network is critical for public safety and resilience to hydrological hazards. This multidisciplinary challenge is explicitly addressed in the Guidelines for the classification and management of risk, safety assessment, and monitoring of existing bridges, issued in Italy by the Ministry of Infrastructure and Transport (Ministry of Infrastructure, CSLP 2020). These Guidelines adopt a comprehensive risk-based approach, considering both structural vulnerabilities—such as those related to structure-foundation systems and seismic events—and environmental threats, including landslides and flooding. To facilitate a comprehensive multi-risk assessment, the Guidelines introduce a five-level methodology (levels 0 through 4), each representing an increasing depth of analysis. Level 0 involves the collection and review of all available documentation about the bridge, including design documents and outcomes of periodic inspections. Level 1 comprises targeted field inspections aimed at assessing the current condition of the bridge, identifying the presence and extent of any degradation or defects affecting structural and non-structural components. At Level 2, the information gathered in the previous stages is used to evaluate the Class of Attention (CoA) for each of the four main risk categories—structural, seismic, hydrological, and geological—based on three key parameters: hazard, vulnerability, and exposure. The partial CoA for each risk is then integrated into a global CoA, categorized on a five-level scale: low, medium-low, medium, medium-high, and high. Depending on the resulting classification, infrastructure managers are required to implement appropriate mitigation measures to ensure the asset's safety and functionality. In evaluating the structure-foundation and seismic Classes of Attention, the vulnerability factor assumes a critical role. This factor is heavily affected by the structural health of the bridge, as determined by the degree of degradation observed in individual structural elements during inspection. In this context, the adoption of advanced methodologies, such as computer vision and deep learning techniques for automated damage detection, can significantly enhance the efficiency and objectivity of the assessment process. These technologies offer considerable advantages, including the reduction of inspection time, the minimization of subjective biases, and the overall improvement and streamlining of the decision-making workflow. Traditional image-based defect detection approaches have primarily relied on close-up images of cracks, spalling, or material degradation. While these methods are effective in controlled inspection settings, they often struggle when applied to wide-angle images that capture larger portions of the structure, such as those obtained during field surveys or drone inspections. In recent years, the YOLO (Redmon et al. 2016) family of object detection algorithms has gained traction in this field due to its high detection speed and real-time performance. For instance, Zhang et al. (2019) demonstrated the effectiveness of an improved YOLOv3 model in detecting various types of concrete bridge damage, including fractures, spalling, and exposed tendons. However, when YOLO models are applied to wide-angle images of entire bridge structures, their performance tends to degrade. One of the main limitations is the tendency to misclassify background elements, such as vegetation, ground cover, or sky, as structural defects. This issue arises because these models are often trained on close-up datasets and therefore lack the contextual awareness needed to distinguish between relevant structural features and irrelevant surroundings. Earlier studies have also emphasized similar challenges. Chen et al. (2010) highlighted the impact of environmental noise, such as discoloration, occlusion, and lighting variability, on automated crack detection. Likewise, Narazaki et al. (2018) noted that when multiple structural elements are present in an image, traditional detection methods are prone to high false-positive rates due to limited scene understanding. This work presents a segmentation-guided machine learning framework designed to overcome the limitations of current defect detection systems when operating on wide-angle surface images of masonry and concrete 117 bridges. A common failure mode observed in standard models trained on highly zoomed-in datasets is the misclassification of occlusions and background elements, such as vegetation, ground cover, or the sky, as structural anomalies. These false positives significantly reduce the reliability of automated inspection tools in operational settings. To address this, we propose a two-stage vision pipeline. In the first stage, a zero-shot segmentation model is employed to isolate the relevant structural components of the bridge from the surrounding scene. This step leverages the model’s ability to generalize across contexts without requiring task-specific training. In the second stage, a dedicated defect detection algorithm is applied, but only within the spatial boundaries defined by the segmentation mask. This guided filtering mechanism effectively reduces noise in the detection process and improves both precision and interpretability. The approach is particularly suited for masonry and concrete structures, where textural and geometric cues can vary significantly across different viewing scales. By focusing detection on semantically meaningful regions, the system achieves better robustness to occlusions and environmental variability. Figure 1. Graphical Architecture of the proposed system Acknowledgements This work has been carried out within the scope of the “Tech4You” Innovation Ecosystem project, Notice no. 3277 of 28.12.2021 – Intervention proposals for the creation and strengthening of “innovation ecosystems” PNRR – MUR project code: ECS0000009 - CUP H23C22000370006Tec4you. References Chen, J., Hutchinson, T. C., & Aghazadeh, B. S. (2010). Image‐Based Framework for Concrete Surface Crack Monitoring and Quantification. Advances in Civil Engineering, 2010. Ministry of Infrastructure, CSLP, 2020. Guidelines on Risk Classification and Management, Safety Assessment and Monitoring of Existing Bridges. Ministry of Infrastructure: Rome, Italy, 2020 (in Italian). Narazaki, Y., Hoskere, V., Hoang, T. A., & Spencer, B. F. (2018). Vision-based Automated Bridge Component Recognition Integrated With High-level Scene Understanding. arXiv preprint arXiv:1805.06041. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You Only Look Once: Unified, Real-Time Object Detection. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 779–788). Zhang, C., Chang, C. C., & Jamshidi, M. (2019). Bridge Damage Detection using a Single-Stage Detector and Field Inspection Images. arXiv preprint arXiv:1812.10590. 124 Flow charateristics evaluation in gravel bed rivers Donatella Termini1,2*, Peyman Peykani1 1 Department of Engineering, University of Palermo, Palermo, Italy 2 NBFC, National Biodiversity Future Center, Palermo, Italy *e-mail: [email protected] Abstract River flow modeling under different conditions is of paramount importance for river engineering. However, still today, there is no generic model that can adequately capture different aspects of flow hydrodynamics over a variety of situations. In flow modeling, river roughness is generally defined with different parameter sets often being able to obtain model predictions approximating well to the observed data. In mountainous and alluvial plain environments gravel-bed rivers are prevalent, exhibiting properties and bed roughness characteristics that profoundly govern flow resistance distribution and sediment transport regimes. The evaluation of hydrodynamics in gravel bed rivers is a complex subject, requiring both theoretical and experimental analyses. Numerical modeling has also emerged as a pivotal tool, as cost-effective alternatives to physical experiments [Grimaldi et al., 2018]. However, despite the effort of several researchers, hydrodynamics in gravel bed rivers is still poorly understood and the ability to rationalize the characteristics of a given channel and predict how it will respond to a change in conditions is still limited. Gravel bed rivers are characterized by a significant heterogeneous distribution of the grains in terms of size and shape. The realistic characterization of the gravel bed surface is very complex, but it is crucial in order to capture the flow mechanisms at different scales constituting the core components of total resistance coefficients to analyze fluvial processes [Navaratnam et al., 2018; Termini et al., 2022]. Different approaches have been proposed in the literature for identifying the gravel-bed surface to characterize the bed roughness [e.g., Aberle and Nikora, 2006] and to adequately simulate the near-bed flow characteristics considering the spatial variability of the grains [e.g., Alfonsi et al., 2019]. But how to reasonably quantify gravel bed surface roughness is still an open question. In this context, the aim of the present work is to give a contribution to fill this gap by combining experimental and numerical analyses. A laboratory study has been conducted over the non-uniform gravel bed (Figure 1) realized in a straight laboratory flume constructed at the Hydraulic Laboratory of the Department of Engineering – University of Palermo (Italy) to explore the flow structure, especially in the near-bed region. Figure 1. a) Photo of the gravel bed surface; b) scanned reproduction of the channel reach considered for simulations The collected data has been used to define a procedure to extrapolate the gravel bed surface. The imagebased technique has been applied to identify the gravel bed structure and the spatial distribution of the grains. Statistical analysis has been conducted to evaluate the gravel surface and the variogram has been determined to quantify the bed roughness proprieties. The results are used for numerical simulations carried out by using available codes (Ansys CFX, OpenFOAM), and the comparison between the numerical results and the experimental data is used to assess the goodness of the procedure. L l 125 Acknowledgements This study was supported partially by RETURN Extended Partnership and received funding from the European Union Next-GenerationEU (National Recovery and Resilience Plan – NRRP, Mission 4, Component 2, Investment 1.3 – D.D. 1243 2/8/2022, PE0000005) and partially by Project code CN_00000033, Concession Decree No. 1034 of 17 June 2022 adopted by the Italian Ministry of University and Research, CUP UNIPA B73C22000790001, Project title “National Biodiversity Future Center - NBFC”. The authors wish to thank Ing. Federica Lavignani for her contribution in laboratory data analysis. References AberleJ., NikoraV. 2006. Statistical properties of armored gravel bed surfaces. Water Resour. Res. 42, W11414. Alfonsi, G., Ferraro, D., Lauria, A., Gaudio R. 2019. Large-eddy simulation of turbulent natural-bed flow. Physics of Fluids 31, 085105. Grimaldi, S.; Li, Y.; Walker, J.P.; Pauwels, V.R.N. 2018. Effective Representation of River Geometry in Hydraulic Flood Forecast Models. Water Resour. Res. 54, 1031–1057 Navaratnam, C. U., Aberle, J., Qin, J., Henry, P. 2018. Influence of Gravel-Bed Porosity and Grain Orientation on Bulk Flow Resistance’. HENRY. Hydraulic Engineering Repository. Water 2018, 10, 561. Termini D., Lavignani F., Benistati N. 2022. Investigation of turbulence characteristics and mixing layer thickness in gravel bed flows. 11th International Conference on Fluvial Hydraulics - RiverFlow2022, Kingston and Ottawa (Canada), November 8-10. 126 Un approccio basato sui dati per il miglioramento della previsione delle piene: il caso del progetto TECH4You nel bacino del Crati Marco Dionigi1*, Stefania Camici1, Melissa Sessa1, Domenico De Santis2, Christian Massari1, Silvia Barbetta1, Tommaso Moramarco1 1 Istituto di Ricerca per la Protezione Idrogeologica, Consiglio Nazionale delle Ricerche, Perugia, Italia 2 Istituto di Ricerca per la Protezione Idrogeologica, Consiglio Nazionale delle Ricerche, Rende, Italia * e-mail: [email protected] Sommario L’esposizione ai rischi naturali è in costante aumento a livello globale, a causa dell’espansione urbana e dei cambiamenti climatici in atto. Tra i pericoli più rilevanti, le alluvioni e le piene improvvise mostrano un incremento significativo, causando impatti sempre più gravi sulle comunità e sulle economie locali. In questo scenario, risulta fondamentale sviluppare strumenti efficaci per la mitigazione del rischio, integrando conoscenze scientifiche, tecnologie innovative e un quadro istituzionale solido e coordinato. Il rafforzamento del dialogo tra ricerca, governance e società civile è cruciale per tradurre le soluzioni tecniche in pratiche concrete, orientate alla resilienza territoriale. In Italia, paese particolarmente vulnerabile agli eventi estremi, sono stati avviati numerosi progetti finalizzati al monitoraggio, alla previsione e alla riduzione degli impatti delle piene. Tra questi si inserisce il progetto TECH4You, ecosistema dell’innovazione delle regioni Calabria e Basilicata, finanziato dal Ministero dell’Università e della Ricerca nell’ambito del PNRR. Il progetto promuove il trasferimento tecnologico a supporto della transizione ecologica e digitale, della tutela della biodiversità e della valorizzazione del patrimonio naturale e culturale, contribuendo al raggiungimento degli obiettivi del Quadro di Sendai delle Nazioni Unite, che sottolinea l’importanza di investire nella riduzione del rischio di disastri. All’interno di questo contesto, il presente contributo descrive la costruzione di un database integrato pioggiadeflusso per il bacino del fiume Crati, finalizzato al supporto operativo delle attività di previsione delle piene. Il database è sviluppato attraverso l’accoppiamento di un generatore stocastico di pioggia e temperatura con un modello idrologico in continuo, in grado di simulare le condizioni di umidità del suolo e i relativi deflussi. Grazie ad un’ampia gamma di scenari generati, il sistema consente di tenere conto delle incertezze legate sia alla previsione meteorologica sia allo stato idrologico del bacino. Una caratteristica innovativa consiste nella possibilità di accedere rapidamente a idrogrammi di piena precalcolati, evitando l’esecuzione in tempo reale dei modelli accoppiati, con un notevole risparmio in termini di tempo operativo nelle fasi di allerta. Ringraziamenti Questo lavoro è stato finanziato dall’iniziativa Next Generation EU – Piano Nazionale di Ripresa e Resilienza (PNRR) italiano, Missione 4, Componente 2, Investimento 1.5, nell’ambito dell’avviso per la creazione e il rafforzamento degli “Ecosistemi dell’Innovazione”, con l’obiettivo di costruire “Leader territoriali di R&S” (Decreto Direttoriale n. 2021/3277) – progetto TECH4You – Tecnologie per l’adattamento ai cambiamenti climatici e il miglioramento della qualità della vita, n. ECS0000009. 127 Enhancing flood forecasting through multi-model hydrological simulation and Bayesian uncertainty estimation with remote and modelled precipitation forcings Domenico De Santis1*, Elenio Avolio2, Silvia Barbetta3, Stefania Camici3, Marco Dionigi3, Sara Modanesi3, Daniela Biondi4, Tommaso Moramarco3, Christian Massari3 1 Research Institute for Geo-Hydrological Protection, National Research Council, Rende, Italy. 2 Research Institute of Atmospheric Sciences and Climate, National Research Council, Lamezia Terme, Italy 3 Research Institute for Geo-Hydrological Protection, National Research Council, Perugia, Italy 4 Department of Computer, Modeling, Electronics and Systems Engineering, University of Calabria, Rende, Italy *e-mail: dome[email protected] Abstract In river basins with short response times, accurate prediction of streamflow in the channel network is essential for effective flood risk management. In this perception, numerous hydrological models are proposed to simulate the rainfall-runoff transformation, which could be coupled with shortand medium-term operational weather forecasts provided at kilometer-scale by convection-permitting models. Another valuable source for forcing hydrological models is remotely sensed, near real-time precipitation estimates, with several operational products available based on observations from passive microwave and infrared sensors. A quantification of the uncertainty of predicted streamflow then constitutes essential information in flood emergency management, with probabilistic forecasts that can be used as efficient support for risk-based decisions. Several methods for estimating predictive uncertainty can be found in the literature, also involving an effective combination of different models based on their ability to predict observed hydrological response, to provide more skillful forecasts than a single model does over the range of possible hydrological situations. The proper way to exploit information from advanced remote sensors and weather forecasting, available with improved accuracy and increasingly finer resolutions, is still an open question within a framework which includes predictive uncertainty estimation. In addition to the evaluation of the reliability of the forcing data, the suitability of the different processing methods is of interest, with a view to efficient support for the real-time early warning activities. As part of the “Tech4You - Technologies for climate change adaptation and quality of life improvement” project, a flood forecasting system is being developed, aimed to usefully support the stakeholders in risk assessment. The system is based on a multi-model application of the Model Conditional Processor (MCP) for predictive uncertainty estimate (Coccia & Todini, 2011; Barbetta et al., 2017). MCP is a Bayesian methodology which uses two joint distributions to improve the adaptation to low and high flows. The Crati River basin, the largest in Calabria in terms of discharge and drainage area, was chosen as case study for the period 2011-2024. A selection of hydrological models was considered with the aim of overall representing the actual streamflow response of the basin. The following conceptual models were used: MISDc2L, HBV-96, GR6H, and NAM. Hydrological modelling was performed on an hourly time resolution and at subbasin scale. An assessment of the selected models was carried out using gauge-based precipitation data. The latter were derived from measurements provided by the Italian Civil Protection Department spatially interpolated on a 1 km grid (Bruno et al., 2021; Avanzi et al., 2023). The calibration objective function was defined as a weighted sum of the well-known Kling-Gupta Efficiency index with an application-specific score for high flow behaviors, and specifically the annual peak flow bias (APFB) proposed by Mizukami et al. (2019). The set of hydrological models was calibrated and validated, providing acceptable performance overall, also considering specific flood events. The KGE index varies between 0.90 and 0.93 and between 0.81 and 0.84 in calibration and validation, respectively, whereas the efficiency on APFB is close to the optimum in calibration and ranges between 0.84 and 0.94 in validation depending on the model. These first analyses highlighted the capability of the selected hydrological models to simulate the observed streamflow when forced with the gauge-based precipitation. The accuracy of the forecast provided by the modelling system further improved when single models’ predictions were effectively combined by using MCP to obtain probabilistic discharge forecasting. Subsequent investigations are focused on the use of alternative precipitation forcings, and specifically: • simulations of Weather Research and Forecasting (WRF) model at convection-permitting scale (2-km) 128 • NASA’s Integrated Multi-satellitE Retrievals for GPM (IMERG) Early Run product (0.1° resolution) in combination, or not, with bias correction approaches to remove systematic differences with ground-based data. The alternative forcings (with or without bias correction) are used to drive the four calibrated hydrological models, whose forecasts are then synergistically integrated during the predictive uncertainty estimation phase. Although a benefit from bias correction is generally expected for the simulation of the single, already calibrated model, the added value of this processing step could still be reduced by the adjustment of the reproduced streamflow due to the MCP application. This analysis therefore provides insights into the assessment and intercomparison of different hydrological models (driven by several forcings), the added value of MCP application (in different scenarios), the role of bias correction approaches in conjunction with MCP (for different forcings), the potential of state-of-the-art satellitebased and modelled precipitation data for hydrological applications and the effectiveness of different workflows in increasing the efficiency of streamflow simulations. Acknowledgements This work was funded by the Next Generation EU - Italian NRRP, Mission 4, Component 2, Investment 1.5, call for the creation and strengthening of ‘Innovation Ecosystems’, building ‘Territorial R&D Leaders’ (Directorial Decree n. 2021/3277) - project Tech4You - Technologies for climate change adaptation and quality of life improvement, n. ECS0000009. This work reflects only the authors’ views and opinions, neither the Ministry for University and Research nor the European Commission can be considered responsible for them. Spatially interpolated precipitation data were provided by CIMA Research Foundation and are based on in-situ precipitation and radar data from the database of the Italian Civil Protection Department. References Avanzi, F., Gabellani, S., Delogu, F., Silvestro, F., Pignone, F., Bruno, G., Pulvirenti, L. et al. IT-SNOW: a snow reanalysis for Italy blending modeling, in situ data, and satellite observations (2010-2021), Earth System Science Data, 2023, 15(2), 639-660. Barbetta, S., Coccia, G., Moramarco, T., Brocca, L. & Todini, E. The multi temporal/multi-model approach to predictive uncertainty assessment in real-time flood forecasting, Journal of Hydrology, 2017, 551, 555-576. Bruno, G., Pignone, F., Silvestro, F., Gabellani, S., Schiavi, F., Rebora, N., Giordano, P. & Falzacappa, M. Performing Hydrological Monitoring at a National Scale by Exploiting Rain-Gauge and Radar Networks: The Italian Case, Atmosphere, 2021, 12(6), 771. Coccia, G. & Todini, E. Recent developments in predictive uncertainty assessment based on the model conditional processor approach, Hydrology and Earth System Sciences, 2011, 15, 3253-3274. Mizukami, N., Rakovec, O., Newman, A. J., Clark, M. P., Wood, A. W., Gupta, H. V., & Kumar, R. On the choice of calibration metrics for “high-flow” estimation using hydrologic models. Hydrology and Earth System Sciences, 2019, 23(6), 2601–2614. 129 L’importanza delle scale di deflusso per la gestione degli eventi di piena: applicazioni in Sardegna Sara Frongia, Enrica Perra, Giaime Tocco, Rossana Bussa, Luigi Perra, Massimo Melis, Saverio Liberatore, Claudio Arras, Domenico Caracciolo* Dipartimento Geologico, idrogeologico e idrografico. Agenzia Regionale per la Protezione dell’Ambiente della Sardegna (ARPAS), Italia *e-mail: [email protected] Sommario Il Dipartimento Geologico, idrogeologico e idrografico (DGII) di ARPA Sardegna redige le scale di deflusso in adempimento dei compiti istituzionali finalizzati alla pubblicazione degli Annali Idrologici parte II, e per favorire la gestione degli eventi meteorologici da parte della Protezione Civile Regionale attraverso il monitoraggio idrometrico con la rete fiduciaria. Per il costante miglioramento delle attività suddette, esso opera in attuazione della convenzione stipulata il 29/12/2020 con l’Agenzia del Distretto Idrografico della Sardegna (ADIS) e ISPRA, finalizzata all’esecuzione di una campagna straordinaria di misure di portata da eseguire nelle sezioni di monitoraggio idrometrico della Sardegna. La conoscenza sistematica e capillare della portata che defluisce nei corsi d’acqua naturali costituisce uno dei presupposti fondamentali per qualunque politica ambientale e di protezione civile: di difesa e di previsione di fenomeni di piena e siccità, di gestione della risorsa idrica, di qualità dell’acqua, di protezione degli ecosistemi fluviali, di difesa dall’inquinamento, di caratterizzazione dei corpi idrici, di gestione delle operazioni di rilascio dalle dighe, di valutazione dell’impatto dei cambiamenti climatici sul regime idrologico ecc. La portata liquida in alveo costituisce pertanto una delle variabili ambientali più importanti ed è al contempo una delle grandezze fisiche più difficili, oltre che economicamente più onerose, da misurare. Il monitoraggio continuo della portata nei corsi d’acqua è legato all’esistenza di un legame tra la portata che attraversa una sezione e il livello idrometrico corrispondente, noto come ‘scala di deflusso’ o ‘scala delle portate’. La conoscenza della portata in una specifica sezione idrometrica, in continuo e in tempo reale, si realizza attraverso l’installazione di un sensore per la misura in continuo del livello idrometrico e la definizione e aggiornamento della scala di deflusso mediante l’esecuzione continua di misure puntuali di portata. A partire dal 2022, ARPA Sardegna condivide sul proprio sito istituzionale le scale di deflusso elaborate in corrispondenza delle sezioni idrometriche nelle quali siano disponibili, in funzione della geometria della sezione, almeno 12-15 misure di portata validate; laddove in base all’analisi dei dati, della geometria e delle proprietà idrauliche della sezione, la scala di deflusso sia suddivisa in più rami (di magra, di morbida, e di piena), ciascun ramo è stato definito con non meno di 5 misure validate, in modo da interpretare in modo corretto l’intero range dei livelli idrometrici. A causa della variabilità idraulica e morfologica dell’alveo, nonché di ulteriori modifiche, anche temporanee, è indispensabile e di fondamentale importanza la continua verifica e aggiornamento della scala di deflusso e dei limiti entro cui essa è valida. Il DGII sta lavorando all’aggiornamento della scala di deflusso, in particolare del ramo di piena, della sezione idrometrica ‘Flumineddu ad Allai’, localizzata all’interno del bacino idrografico del Fiume Tirso, il maggiore fiume della Sardegna per lunghezza e superficie del bacino; tale sezione è di fondamentale importanza per il monitoraggio degli apporti al Fiume Tirso da uno dei suoi principali affluenti, il Flumineddu. In questa sezione idrometrica il ramo di piena della scala di deflusso è stato costruito utilizzando misure di portata relative ad un evento di piena del 2021; durante l’evento di piena del 16-18 aprile 2025, a causa di modifiche della sezione idrometrica, le portate transitanti sono risultate inferiori a quelle previste dalla scala di deflusso. Ciò ha fatto sì che l’Ente gestore della diga di Pranu Antoni, sita a valle della sezione idrometrica, stimando un afflusso da monte superiore a quello effettivo, abbia effettuato un rilascio dalla diga superiore a quello che sarebbe stato necessario. Inoltre, in assenza di misure dirette di portate di piena, spesso a causa della difficoltà nell’esecuzione e dell’intensificarsi dei fenomeni siccitosi, le scale di deflusso non possono essere definite empiricamente mediante l’interpolazione di punti livello-portata. Un metodo per l’estrapolazione del ramo di piena per le sezioni dove non siano disponibili misure dirette di portata consiste nell’utilizzo della modellistica idraulica e la conseguente predisposizione di scale di deflusso teoriche. Nell’ambito di questo lavoro viene riportato l’aggiornamento della scala di deflusso per la sezione idrometrica Flumendosa a Ballao, appartenente al bacino idrografico del Fiume Flumendosa, il secondo fiume della Sardegna per lunghezza e di importanza strategica 130 insieme al Fiume Tirso per il sistema idrico della regione, con l’elaborazione del ramo di piena ottenuto attraverso modellazione idraulica. Bibliografia ISO 18320, 2020. Hydrometry – Measurement of liquid flow in open channels – Determination of the stagedischarge relationship, pp. 1-54. ISPRA, 2017. Programma Nazionale di misure di portata in corsi d’acqua finalizzate alla definizione della scala di deflusso. Valutazione tecnico-economica. Tavolo Nazionale per i Servizi di Idrologia Operativa, 20(7), pp.1-13. 131 Un modello di machine learning per la valutazione della suscettibilità alle alluvioni nella fascia costiera del Basento Marica Rondinone1*, Silvano Fortunato Dal Sasso2, Htay Htay Aung2, Mauro Fiorentino1, Vito Telesca1 1 Dipartimento di Ingegneria, Università della Basilicata, Potenza, Italia 2 Dipartimento per l'Innovazione Umanistica, Scientifica e Sociale, Università della Basilicata, Potenza e Matera, Italia *e-mail: [email protected] Sommario Le alluvioni costituiscono una delle principali minacce naturali per il territorio italiano, con impatti rilevanti su insediamenti urbani, infrastrutture e ambiente. Per affrontare la crescente complessità dell'analisi del rischio idraulico, questo studio propone una metodologia basata sull'algoritmo di apprendimento automatico XGBoost per la valutazione della suscettibilità alle alluvioni. L’approccio integra dati geospaziali ad alta risoluzione e tecniche di interpretazione dei modelli (eXplainable AI - XAI) per produrre mappe predittive a supporto della pianificazione e mitigazione del rischio (Rondinone et al., 2025). Il bacino del fiume Basento, in Basilicata, è storicamente soggetto a eventi alluvionali ricorrenti e impattanti (Albano et al., 2024; Dal Sasso et al., 2017). Eventi di piena significativi si sono verificati nel marzo 2011, ottobre e dicembre 2013, causando danni estesi. L'inventario delle aree alluvionate è stato costruito utilizzando dati da Copernicus EMS, immagini radar COSMO-SkyMed e archivi regionali. I dati satellitari ad alta risoluzione (2.5 m) hanno consentito una mappatura dettagliata dell'estensione degli eventi. Sono stati selezionati otto fattori condizionanti di natura spaziale: elevazione, elevazione relativa, pendenza, esposizione, distanza dal reticolo idrografico, densità di drenaggio, uso del suolo e litologia. I dati raster sono stati elaborati in ambiente GIS con risoluzione 20 m x 20 m. Il modello XGBoost è stato addestrato su una base dati suddivisa in training set (70%) e test set (30%). Per evitare fenomeni di overfitting legati alla complessità geomorfologica e all’eterogeneità del territorio, l’analisi è stata applicata a una porzione del bacino con altitudine inferiore ai 350 m s.l.m., maggiormente soggetta a eventi alluvionali e caratterizzata da maggiore omogeneità geomorfologica. Per compensare lo sbilanciamento delle classi (aree inondate vs. non inondate), si è applicato il metodo SMOTE (Pradipta et al., 2021). La valutazione del modello è stata condotta mediante cross-validation a 10 fold e test su set indipendenti, confermandone la robustezza e l'affidabilità. L’interpretazione dei risultati è stata supportata dall’uso della tecnica SHAP (SHapley Additive exPlanations), in linea con quanto mostrato da Cappelli et al. (2024). Il modello ha raggiunto un'accuratezza pari a 0.983. La classe "inondata" ha ottenuto una precisione del 68.7% e un recall del 97.5%. La Figura 1 mostra i risultati della classificazione ottenuta dal modello XGBoost nella zona costiera del fiume Basento. Le aree in ciano indicano le zone correttamente identificate come soggette a inondazione, mentre le aree in blu e rosso rappresentano, rispettivamente, i falsi positivi (FP) e i falsi negativi (FN). Il 62% delle aree classificate come alluvionate ricade in corrispondenza del reticolo idrografico ufficiale (CTR), a conferma dell'affidabilità spaziale del modello. Le analisi di errore mostrano una bassa percentuale di falsi positivi (10.55%) e falsi negativi (1.69%). I tre fattori più influenti nella classificazione, secondo l'analisi SHAP, sono: l’elevazione (53%), l’elevazione relativa (14%) e la densità di drenaggio (13%). Le aree con bassa altitudine e alta densità di drenaggio sono risultate le più esposte al rischio. L’approccio si è dimostrato efficace per la mappatura della suscettibilità alle alluvioni nel bacino del Basento. L’integrazione con tecniche di interpretazione del modello ha consentito inoltre di individuare le principali variabili geospaziali che influenzano maggiormente il fenomeno alluvionale, supportando decisioni più consapevoli nell’ambito della pianificazione e mitigazione. La metodologia si conferma scalabile e adattabile ad altri contesti esposti a pericolosità idraulica. Sviluppi futuri potrebbero includere l’applicazione di tecniche di interpretabilità globale e locale, per comprendere non solo quali fattori influenzano maggiormente il comportamento complessivo del modello, ma anche per analizzare come e per quali valori specifici le singole variabili incidano sui risultati del modello. 132 Figura 1. a) Mappa di classificazione delle aree alluvionate; b) istogramma della global SHAP Feature Importance Bibliografia Albano, R., Limongi, C.,Dal Sasso, S.F., Mancusi, L., Adamowski, J., 2024. Flood scenario spatio-temporal mapping via hydrological and hydrodynamic modelling and a remote sensing dataset: A case study of the Basento river (Southern Italy). Int. J. Disaster Risk Reduct. 2024, 111, 104758. Cappelli F, Castronuovo G, Grimaldi S, Telesca V. 2024. Random Forest and Feature Importance Measures for Discriminating the Most Influential Environmental Factors in Predicting Cardiovascular and Respiratory Diseases. Int J Environ Res Public Health. 2024 Jul 2;21(7):867. doi: 10.3390/ijerph21070867. PMID: 39063444; PMCID: PMC11276884. Dal Sasso, S.F., Manfreda, S., Capparelli, G., Versace, P., Samela, C.; Spilotro, G., Fiorentino, M., 2017. Hydrological and geological hazards in Basilicata. L’Acqua 2017, 3, 77–85. Pradipta, G.A., Wardoyo, R., Musdholifah, A., Sanjaya, I.N.H., Ismail, M. 2021. SMOTE for Handling Imbalanced Data Problem: A Review. In 2021 Sixth International Conference on Informatics and Computing (ICIC), Jakarta, Indonesia, 2021; pp. 1-8. doi: 10.1109/ICIC54025.2021.9632912. Rondinone, M., Dal Sasso, S.F., Aung, H.H., Contillo, L., Dimola, G., Schiattarella, M., Fiorentino, M., Telesca, V., 2025. Assessing Flood and Landslide Susceptibility Using XGBoost: Case Study of the Basento River in Southern Italy. Appl. Sci. 2025, 15, 5290. 133 QT-DREAM: un approccio idrologico distribuito per la simulazione delle piene di progetto in contesti scarsamente strumentati Federica Mesto1*, Andrea Gioia2, Rocco Bonelli3, Silvano Fortunato Dal Sasso4, Luciana Giuzio3, Margherita Lombardo2,5, Salvatore Manfreda6, Maria Rosaria Margiotta7, Beniamino Onorati7, Biagio Sileo4, Pasquale Perrini1, Vincenzo Totaro2, Vito Iacobellis2, Mauro Fiorentino7, Vera Corbelli3 1 DISSPA, Università degli Studi di Bari Aldo Moro, Bari 2 DICATECh, Politecnico di Bari, Bari 3 Autorità di Bacino Distrettuale dell'Appennino Meridionale, Caserta. 4 DiUSS, Università della Basilicata, Matera. 5 DIAm, Università della Calabria, Rende 6 DICEA, Università di Napoli Federico II, Napoli 7 DiING, Università della Basilicata, Potenza *e-mail: [email protected] Sommario La stima delle piene di progetto (T-floods) rappresenta una delle sfide principali dell’idrologia, soprattutto in bacini scarsamente strumentati. In questo studio si propone un approccio basato su un modello idrologico distribuito, QT-DREAM, evoluzione del modello DREAM (Distributed model for Runoff, Evapotranspiration, and Antecedent soil Moisture simulation, Manfreda et al., 2005), per simulare gli idrogrammi di piena associati a tempi di ritorno pari a 30, 200 e 500 anni. Il modello è stato applicato a 20 bacini strumentati situati nelle regioni Puglia e Basilicata, caratterizzati da condizioni climatiche e morfologiche eterogenee. Una delle principali innovazioni metodologiche introdotte riguarda la generazione di griglie tridimensionali (spazio-tempo) di runoff, ottenute tramite un’unica simulazione sull’intero bacino principale. Queste griglie vengono successivamente applicate direttamente ai sottobacini strumentati annidati, evitando ricalcoli locali del bilancio idrologico e garantendo la coerenza spaziale tra le diverse sezioni. L’approccio, concepito anche in vista di applicazioni Rain-On-Grid all’interno di modelli idrodinamici 2D, permette di estendere le valutazioni idrologiche a sezioni non monitorate, offrendo così un valido supporto alla gestione del rischio idraulico su scala di bacino. Il confronto tra le simulazioni effettuate sull’intero bacino principale e quelle condotte sui singoli sottobacini ha evidenziato una buona coerenza interna del metodo, con differenze medie nei colmi di piena inferiori al 10%. Inoltre, il modello ha dimostrato una buona capacità di riprodurre la distribuzione di frequenza dei colmi di piena (Iacobellis and Fiorentino, 2000; Gioia et al., 2008) con uno scarto medio rispetto ai valori osservati pari a circa il 15% nei bacini della Puglia e al 6% in quelli della Basilicata. L’affidabilità del modello è ulteriormente rafforzata dall’impiego di parametri fisicamente coerenti e dalla consistenza spaziale dei dati di input (uso del suolo, tipo di suolo, topografia), derivati da fonti ad alta risoluzione. Il presente lavoro propone, pertanto, un framework operativo per la valutazione del rischio alluvionale in contesti caratterizzati da una bassa densità di osservazioni, aprendo la strada all’integrazione con modelli idraulici per la produzione di mappe di pericolosità. 140 Conceptual hydrological modelling in complex and data-scarce mountainous basins using additional information sources: a case study in the Indian Himalayas Domenico De Santis1*, Silvia Barbetta2, Farhad Bahmanpouri2, Sumit Sen3, Ashutosh Sharma3, Ankit Agarwal3, Sagar Gupta3, Viviana Maggioni4, Francesco Avanzi5, Christian Massari2 1 Research Institute for Geo-Hydrological Protection, National Research Council, Rende, Italy 2 Research Institute for Geo-Hydrological Protection, National Research Council, Perugia, Italy 3 Department of Hydrology, Indian Institute of Technology Roorkee, Roorkee, India 4 Department of Civil, Environmental & Infrastructure Engineering, George Mason University, Fairfax, USA 5 CIMA Research Foundation, Savona, Italy *e-mail: dome[email protected] Abstract The Indian Himalayan Region holds significant hydrological interest, with streamflow response dynamics that continue to warrant in-depth investigation. Basins characterized by high mountains and densely populated valleys are prone to flash floods and other river flow-related disasters, due to heavy rains triggered by the interaction between the complex orography and the Indian Summer Monsoon (ISM). A complex interplay of meteorological, topographical, and runoff generation factors controls streamflow variability, whose accurate forecast is pivotal for effective flood risk management. However, analysing and modelling main hydrological processes, especially during floods, present substantial challenges, given the data scarcity (typical of highaltitude areas) and high temporal/spatial heterogeneity of hydrological variables. Rain gauges often lack accuracy and representativeness, whereas meteorological models and remote sensing observations generally suffer from large biases and uncertainties in complex terrain. Accurately quantifying basin-scale water budget components remains a major challenge in these regions, making water balance closure particularly difficult and hindering a precise representation of hydrological dynamics. Hydrological modelling can bridge this gap, ensuring consistency across multiple data sources, by identifying and mitigating some systematic errors. Conceptual and semi-distributed models can be particularly well-suited for basins characterized by complex dynamics and limited data availability; however, accurately capturing the hydrological processes that drive streamflow variability remains essential for reliable modelling. A ‘fit-forpurpose’ conceptualization can well represent how rain, snow, and glacier melt contribute to runoff generation, especially during high flows. The integration of additional information has the potential to provide robust hydrological simulations. However, a more realistic representation of internal processes does not necessarily lead to better streamflow simulations. This study focused on the partially glacierized, monsoon-dominated Alaknanda River basin (~8600 km²), one of the two headstreams of the Ganges. Data scarcity was addressed through a dual approach: by employing a specifically tailored, simplified, and parsimonious hydrological modelling framework, and by constraining calibration using satellite-based data. The conceptual and semi-distributed MISDc-2L hydrological model, previously applied to a multitude of basins and proven effective for flood prediction, was modified here with a tailor-made snow module and the addition of a static and a dynamic glacier module. The modelling framework was enhanced by using multiple reference data. Specifically, information on glacier stored water loss and actual evapotranspiration (AET) was used to close the water balance and improve the model ability to reproduce magnitude and patterns of hydrological fluxes. This was tackled by adjusting the reanalysis precipitation, which underestimated the observed streamflow at the outlet. Multiple scenarios were explored, differing in the data used to constrain model calibration, the methods applied to correct systematic precipitation errors, and the treatment of glacier melt – whether explicitly modelled or not. The analysis showed that: • Despite its simplified and parsimonious conceptualization, the model demonstrated a strong ability to reproduce observed streamflow during both the calibration and validation periods with a KGE of 0.88 and 0.83, respectively, for the baseline scenario. These increased to 0.93 and 0.92 when using a two-parameter precipitation adjustment formula. The model proved to be reliable under average flow conditions and effectively captured high-flow regimes. However, its performance in simulating specific flash flood events was limited, primarily due to localized 141 inaccuracies in the rainfall data. • The model accurately reproduced reference estimates of glacier storage changes and AET when additional data were embedded in the calibration framework. Using additional reference datasets enhanced the model’s ability to represent physical processes but introduced trade-offs in the estimation of the seasonal hydrograph. This can be attributed to limitations in the conceptualization of hydrological processes and errors in different datasets leading to incorrect parameter estimation. • Parsimonious precipitation adjustments can significantly improve streamflow simulation, by handling biases in the precipitation dataset. Model simulations enabled the quantification of meteorological forcing underestimation for hydrological applications, with the extent of such underestimation varying by season and precipitation intensity. • Glacier melt contributes marginally to overall streamflow, but its inclusion improves internal model consistency. Simple conceptualizations, such as temperature-driven melting combined with static or V-A scaling approaches for glacier evolution, may be appropriate for hydrological simulations in monsoon-dominated basins. Here, reference data available for calibration were well reproduced both in terms of magnitude and spatiotemporal patterns, but during the validation period the model did not capture the expected increase in glacier stored water loss. The 20-year modelling analysis (2001-2020) yielded hydrologically consistent estimates of the main water fluxes, further refined through the integration of independent multi-source data. This analysis enabled a deeper exploration of various aspects of the water cycle within the study basin – capturing seasonal and spatial patterns as well as interannual variability – and contributed to advancing process understanding in hydrologically heterogeneous, monsoon-dominated basins of the Indian Himalayas. Overall, the framework offers a practical modelling strategy for data-scarce Himalayan basins with similarly complex hydrological processes, offering valuable insights for regional flood forecasting and water balance assessments. Nonetheless, limitations persist due to coarse input data, simplified representations of snow and glacier dynamics, and weak parameter constraints at higher elevations. Future work should focus on improving the downscaling of meteorological inputs, enhancing the physical modelling of snow and ice processes, and applying multi-objective calibration techniques using more robust, uncertainty-characterized datasets – including reference data for snow dynamics. The availability of accurate high-resolution precipitation data is essential to improve the predictability of high flows during the ISM season, regardless of the complexity of the hydrological model. The inability to capture localized heavy rainfall events limits the effectiveness of streamflow post-processing, constrains the use of hydrological models as predictive tools for flood forecasting, and hinders progress in understanding hydrological responses to extreme precipitation in the Indian Himalayas. Acknowledgements This study was carried out in the framework of the FLOSET Project “Probabilistic floods and sediment transport forecasting in the Himalayas during the extreme events”, funded in the context of the ‘ITALY-INDIA JOINT SCIENCE AND TECHNOLOGY COOPERATION CALL FOR JOINT PROJECT PROPOSALS FOR THE YEARS 2021 2023’. 142 Residual Dynamics in Hydrological Models: Insights from a large sample of catchments and models Luca Lombardo1*, Simon Michael Papalexiou2,3, Cyril Thébault2, Martyn P. Clark2, Richard M. Vogel4, Alberto Viglione1 1 Department of Environmental Engineering, Politecnico di Torino, Italy 2 Department of Civil Engineering, University of Calgary, Canada 3 University of Saskatchewan, Canada 4 Department of Civil & Environmental Engineering, Tufts University, Medford, MA, USA *e-mail: luca.[email protected] Abstract Residuals from hydrological models are critical for evaluating model performance, improving predictive accuracy, and deepening the understanding of hydrological processes. Enhancing predictive methods is especially crucial for capturing extreme events, which have significant implications for risk management and planning. These residuals, however, are influenced by model structures, preprocessing methods, and catchment characteristics. This study addresses these complexities by systematically analyzing the statistical properties of residuals under various transformations and preprocessing treatments. The analysis spans a diverse dataset of catchments across a broad range of hydroclimatic conditions, with residuals generated from simulations of multiple hydrological models, ensuring both the generality and robustness of the findings. Key aspects of the research include the evaluation of residual properties under transformations, such as logtransformation, and the role of preprocessing steps. Through this approach, the study provides a more consistent framework for assessing variability, skewness, kurtosis, autocorrelation, and dependency structures in residuals. Additionally, the analysis encompasses heteroskedasticity and tail dependencies, capturing the nuances of residual behavior across different contexts. The dataset’s extent is a defining strength of this study. By involving simulations from a wide range of hydrological models (78 configurations) and including catchments with varying climatic and physical characteristics (more than 400 basins in the United States, ranging from dry to wet climates), the research delivers insights that are widely applicable to diverse hydrological conditions. This breadth ensures that findings are relevant for both theoretical advancements and practical applications, offering guidance to researchers and practitioners working with different modeling systems and catchment types. A central result highlights the transformative impact of removing seasonality from residuals. Deseasonalization not only stabilizes key residual properties but also reduces variability across models, facilitating a clearer evaluation of model performance and error structures, underscoring the importance of standardizing preprocessing techniques in hydrological modeling, as it enables more robust and interpretable diagnostic frameworks. A direct consequence of residual analysis is then the development of stochastic models for uncertainty estimation as well as the possible development of alternative objective metrics to calibrate rainfallrunoff models. 143 GETAFLOOD: un tool integrato di modellazione e analisi idrologica per la derivazione automatica delle piene di progetto nei bacini idrografici siciliani Antonio Francipane*, Dario Treppiedi, Leonardo Valerio Noto Dipartimento di Ingegneria, Università degli Studi di Palermo, Palermo, Italia *e-mail: [email protected] Sommario La stima delle piene di progetto rappresenta un elemento di fondamentale importanza in molti settori dell’idrologia e dell’idraulica, dalla pianificazione e gestione delle infrastrutture fino alla difesa del territorio, grazie alla quale è possibile progettare opere in grado di resistere a eventi di piena estremi e minimizzare i rischi per le comunità e per l’ambiente. Tale processo, spesso, vede l’utilizzo di modelli idrologici, anche a struttura semplificata, che forzati con dati meteo-climatici e con alcune caratteristiche del bacino idrografico in studio, ne modellano la risposta idrologica. Con riferimento ai modelli idrologici utilizzati, questi possono svolgere un ruolo cruciale nella comprensione delle interazioni che intercorrono tra precipitazioni e deflusso all’interno del bacino, consentendo di analizzare l’intensità e la frequenza con cui accadono eventi estremi, come piene e siccità, di valutare l’impatto di scenari climatici futuri sui cambiamenti nelle componenti del bilancio idrologico e nella gestione delle risorse idriche, o ancora, con riferimento alla valutazione del rischio idraulico, di comprendere i potenziali impatti delle inondazioni su infrastrutture, ecosistemi e comunità, costituendo i dati di input di modelli idraulici per lo studio della propagazione idraulica di tipo monoo bi-dimensionale. Una corretta analisi delle piene di progetto non può prescindere dagli elementi precedentemente menzionati, beneficiando sia della capacità dei sistemi GIS nel gestire dati spazialmente distribuiti, sia di alcune tecniche di analisi spaziale per la caratterizzazione di un bacino, per integrare queste informazioni all’interno di modelli idrologici e migliorarne l’affidabilità predittiva. Tuttavia, sviluppare un ambiente GIS user-friendly che consenta di stimare le piene di progetto combinando e integrando tutti questi elementi non è del tutto banale. Uno degli obiettivi più impegnativi per tali sistemi è quello di utilizzare modelli idrologici che, pur nella loro semplicità, siano in grado di rappresentare correttamente la formazione della piena di progetto. Nella maggior parte dei casi, i modelli utilizzano un metodo di trasformazione basato sul concetto dell’idrogramma unitario (UH; Sherman, 1932), che si fonda su concetti come il tempo di corrivazione e la curva area-tempo (di corrivazione) del bacino, per i quali l’impiego di un sistema GIS può rappresentare un fattore determinante. Questo compito è tutt’altro che semplice, poiché coinvolge molti fattori, come le velocità con cui l’acqua si muove all’interno del bacino, l’uso del suolo, la morfologia, la scabrezza del suolo, ecc. In questo lavoro presentiamo GETAFLOOD (GEospatial Tool for Automatic derivation of design FLOOD), un tool basato sul linguaggio di programmazione Python, sviluppato in ambiente open-source (QGIS), che integra dati, funzionalità GIS e diverse librerie geospaziali open-source, come GDAL, SAGA e WhiteboxTools (Lindsay, 2016), con tecniche di modellazione idrologica. L’obiettivo è quello di creare un framework integrato che fornisca uno strumento pratico e semplice per la modellazione idrologica delle piene, superando alcune assunzioni semplificative comuni a molti studi idrologici, come quelle utilizzate per stimare il tempo di corrivazione e la curva area-tempo del bacino. La Figura 1 mostra il framework modellistico utilizzato in GETAFLOOD. A partire dalle curve di probabilità pluviometrica (CPP) ottenute secondo una procedura regionalizzata (Forestieri et al., 2018), il tool genera gli ietogrammi sintetici di tipo Chicago per i tempi di ritorno specificati, offrendo la possibilità di considerare un fattore di cambiamento climatico per tenere conto degli effetti attuali e futuri del clima sulle CPP. L’idrogramma di piena è ottenuto attraverso l’applicazione congiunta di un modello di depurazione delle piogge, come quello del CN o di Horton, e del metodo dell’idrogramma unitario distribuito, basato su un nuovo algoritmo sviluppato appositamente. Uno dei principali vantaggi dello strumento è che tutti questi processi, solitamente eseguiti separatamente dall’operatore, sono eseguiti in un unico processo integrato, che parte dai dati di input necessari per restituire in output l’idrogramma di piena, semplificando notevolmente il processo di modellazione e offrendo un supporto concreto per tecnici e ricercatori per l’applicazione in diversi settori, dalla valutazione del rischio idraulico alla progettazione e gestione delle infrastrutture idrauliche. 144 Figura 1. Framework modellistico di GETAFLOOD Bibliografia Forestieri, A., Lo Conti, F., Blenkinsop, S., Cannarozzo, M., Fowler, H.J., Noto, L.V., 2018. Regional frequency analysis of extreme rainfall in Sicily (Italy). Int. J. Climatol. 38, e698–e716. https://doi.org/10.1002/joc.5400. Lindsay, J.B., 2016. Whitebox GAT: A case study in geomorphometric analysis. Comput. Geosci. 95, 75–84. https://doi.org/10.1016/j.cageo.2016.07.003. Sherman, LeRoy K., 1932. Streamflow from rainfall by the unit-graph method. Eng. News Record 108, 501– 505. 145 Validation of specific and total catchment area estimated via flow direction algorithms through a 2D shallow water equations numerical solver Sara Carta*, Federico Prost, Francesca Aureli, Paolo Mignosa Department of Engineering and Architecture, University of Parma, Parma, Italy *e-mail: [email protected] Abstract Specific Catchment Area (SCA) and Total Catchment Area (TCA) are two widely used topographic attributes in the study of hydrological, geomorphological and biological processes at the watershed scale. Examples of their applications include the use of SCA and/or TCA fields to predict patterns of relative soil saturation, potential erosion, or the spatial distribution of different plant species (Moore et al., 1991; Chirico et al., 2005; Li et al., 2023). Hydrologically, the TCA – defined as the flat area whose surface runoff is potentially conveyed through a given contour segment of length w – can be connected with the discharge at that contour segment in the case of uniform, stationary rainfall and when flow is determined primarily by slope gradient (Gallant & Hutchinson, 2011). The SCA, defined as the contributing area per unit of contour length (SCA = TCA/w), is proportional to the specific discharge under the same assumptions (Chirico et al., 2005; Gruber & Peckham, 2009). The two attributes are typically calculated after determining the flow directions, which can be identified using different algorithms from a digital terrain model (Tarboton, 1997). There are several methods available in the literature, but they can be divided into two main categories: Single Flow Direction (SFD) and Multiple Flow Direction (MFD). In the context of a discretized domain with a regular grid, the two categories differ in terms of whether an individual flow direction is assigned to a given cell from the eight neighboring cells (in the case of SFD methods) or whether the flow is distributed among multiple directions with different weights (in the case of MFD methods). The former are distinguished by their simplicity of implementation, especially when used within more complex modeling chains; however, their inability to describe divergent flow limits their application. On the other hand, MFD methods may exhibit excessively dispersive behaviors (Gruber & Peckham, 2009). In this study, six of the most widely used algorithms were selected: two SFD methods, namely D8 and Rho8, and four MFD methods, namely D-Infinity, MFD-Quinn, MFD-md, and FD8 (Gruber & Peckham, 2009; Qin et al., 2007). The performance of these methods is often evaluated on synthetic surfaces described by known equations, in order to derive the analytical expression of the SCA based on its definition, following the method proposed by Zhou & Liu (2002). This procedure provides a reference solution against which the results obtained with the different algorithms can be quantitatively compared. However, the regular surfaces commonly used for this purpose (sometimes even very complex) are not adequately representative of the real morphology of natural watersheds. Since the fields of SCA and TCA values are not known a priori in a real DTM, nor can they be calculated analytically, an evaluation of the distributed results obtained with the different algorithms on a real watershed remains a pending issue (Li et al., 2021). The research here presented proposes an innovative approach to the analysis of the performances of flow direction algorithms for the determination of TCA/SCA, through the hydraulic-hydrological model PARFLOODRain for complete hydrodynamic propagation at the watershed scale. The model is a numerical solver of the 2D Shallow Water Equations, based on an explicit finite volume discretization, according to a well-balanced formulation. Using the PARFLOOD-Rain model, it is possible to simulate a precipitation field over a region of interest, described through a terrain model with a regularly spaced grid. The solver is undoubtedly a powerful tool in terms of both result accuracy and computational efficiency (Aureli et al., 2020). In the first phase of the study, a constant and uniform rainfall was simulated through the PARFLOOD-Rain model over four synthetic surfaces, selected among the most commonly used ones (inclined plane, saddle, convex and concave surfaces) until a steady-state condition was reached. The discharge and specific discharge fields, representative of the steady-state condition, were then used to construct fields of the corresponding TCA and SCA values. For the same surfaces, SCA and TCA maps obtained through the six selected flow direction algorithms were also determined, alongside the reference analytical solutions. The results obtained with PARFLOOD-Rain and the different algorithms were compared with the exact solutions, in order to derive the error metrics. The results obtained through PARFLOOD-Rain, for different grid resolutions, highlighted the high accuracy of the solver in the reproduction of the patterns of the studied topographic attributes (Figure 1a-d). Unlike all the examined algorithms, the numerical model converges towards the exact solution as grid resolution increases, showing remarkably low mean error values. 146 An overview of the initial findings allows for justifying the use of the numerical solver results as a benchmark for validating flow direction algorithms on real surfaces, thereby addressing the gap found in the current literature. This ongoing study thus proposes a validation of the six aforementioned algorithms on real watersheds. As an intermediate phase of the analysis, SCA and TCA maps were examined for a surface characterized by a convergent basin in a constricted channel, followed by a divergent conical surface, for the purpose of representing a natural alluvial fan at the outlet of a narrow gorge – which is a common morphology in mountain basins. The obtained results highlight the superiority of the MFD algorithms, while also allowing the investigation of the limitations imposed by the simplicity of their algorithm – such as, for example, the tendency towards excessive flow dispersion (Figure 1e-g). Finally, interesting observations can be made regarding the discretion involved in defining the contour length, a crucial factor in calculating SCA from TCA, but often inconsistently defined across the different studies in the literature (Li et al., 2023). Figure 1. Left: divergent surface with elliptical contours (800 m x 600 m, 50 cm resolution): SCA calculated using the analytical relationship (a), with PARFLOOD-Rain results (b), using the D8 method (c) and the MFD-Quinn method (d). Right: convergent-divergent surface identified in the intermediate phase (7600 m x 4000 m, 2 m resolution): TCA using the D8 method (e), MFD-Quinn method (f) and PARFLOOD-Rain (g). References Aureli, F., Prost, F., Vacondio, R., Dazzi, S. and Ferrari, A., 2020. A GPU-Accelerated Shallow-Water Scheme for Surface Runoff Simulations. Water, 12(3), p.637. Chirico, G.B., Western, A.W., Grayson, R.B. and Blöschl, G., 2005. On the definition of the flow width for calculating specific catchment area patterns from gridded elevation data. Hydrol. Process., 19, pp.25392556. Gallant, J. C. & Hutchinson, M. F., 2011. A differential equation for specific catchment area. Water Resources Research, 47, W05535. Gruber, S. & Peckham, S., 2009. Land-Surface Parameters and Objects in Hydrology (Chapter 7 in Geomorphometry: Concepts, Software, Applications). Developments in Soil Science, 33, pp.171-194. Li, Z., Lai, X., Shi, P., & Yang, T., 2023. An algorithm-independent definition of effective contour length and its impacts on pixel-scale specific catchment area. Water Resources Research, 59, e2022WR034036. Li, Z., Yang, T., Wang, C., Shi, P., Yong, B., & Song, Y., 2021. Assessing the precision of Total Contributing Area (TCA) estimated by flow direction algorithms based on the analytical solution of theoretical TCA on synthetic surfaces. Water Resources Research, 57, e2020WR028546. Moore, I. D., Grayson, R. B. and Ladson, A. R., 1991. Digital terrain modelling: a review of hydrological, geomorphological, and biological applications. Hydrological Processes, 5, pp.3-30. Qin, C., Zhu, A.-X., Pei, T., Li, B., Zhou, C. and Yang, L., 2007. An adaptive approach to selecting a flowpartition exponent for a multiple‐flow‐direction algorithm. International Journal of Geographical Information Science, 21(4), pp.443-458. Tarboton, D. G., 1997. A new method for the determination of flow directions and upslope areas in grid digital elevation models. Water Resources Research, 33(2), pp.309–319. Zhou, Q., & Liu, X., 2002. Error assessment of grid-based flow routing algorithms used in hydrological models. International Journal of Geographical Information Science, 16(8), pp.819–842. e) f) g) 147 Monitoraggio pluviometrico, radar e satellitare di alluvioni lampo in bacini mediterranei Paolo Colosio1*, Sante Laviola2, Giulio Monte2, Roberto Ranzi1 1 DICATAM, Università di Brescia, Brescia, Italia 2 CNR-Istituto di Scienze dell’Atmosfera e del Clima (ISAC), Bologna, Italia *e-mail: [email protected] Sommario Gli eventi di precipitazione intensa di breve durata, spesso originati da eventi convettivi (Dallan et al., 2022), sono in grado di generare alluvioni lampo e colate detritiche, capaci di provocare danni ingenti e, talvolta, la perdita di vite umane (Luino et al., 2019). Lo studio di questi eventi può essere effettuato grazie alla modellistica numerica meteorologica che aiuta a descrivere fisicamente l’evento e quella idrologica per meglio comprenderne gli impatti al suolo (Bonomelli et al., 2024) e permetterne la previsione in tempo reale (Ranzi et al., 2009), l’analisi idrologica relativa alla valutazione dell’intensità e della durata di questi eventi è di fondamentale importanza per poter comprendere i meccanismi di innesco di questi eventi superficiali e poter definire delle soglie pluviometriche di preallerta (Ioratti et al., 2025). Nella memoria si presentano i risultati del monitoraggio pluviometrico, radar e satellitare di sei eventi di precipitazione intensa che hanno innescato alluvioni lampo e colate detritiche nelle Alpi ed in Sardegna tra il 2012 ed il 2022, causando quattro vittime. Il 27 luglio 2012 una colata detritica in Val Rabbia ha lambito l’abitato di Rino di Sonico in Lombardia causando gravi danni, l’interruzione del gasdotto e di infrastrutture lineari e forte preoccupazione nella popolazione. L’alluvione di Monte Pinu in Sardegna del 18 novembre 2013, associata al ciclone Cleopatra, ha causato il crollo di infrastrutture stradali e tre vittime. A Dimaro, in Trentino, la colata detritica del Rio Rotiano del 29 ottobre 2018 è avvenuta dopo tre giorni di piogge intense, seguiti da una fase convettiva della tempesta Vaia (Davolio et al., 2020) che ha innescato tre impulsi di colata detritica causando danni ingenti nel conoide e una vittima. Altri tre eventi significativi sul territorio Lombardo includono la colata detritica del torrente Vallaro di Vione del 28 agosto 2020, l’evento del torrente Blè di Ono San Pietro del 16 agosto 2021 (Berti et al., 2023) e quello del 27-28 luglio 2022 del torrente Re di Cobello. I sei eventi sono stati studiati sfruttando tre sistemi di monitoraggio: la rete di pluviometri disponibili, il radar meteorologico e immagini satellitari. La rete di pluviometri utilizzata conta venti stazioni di misura distribuite nell’area dei bacini idrografici di Lombardia e Trentino che forniscono il dato di precipitazione a scala temporale di 5 o 10 minuti. Per il caso studio della Sardegna, invece, sono stati utilizzati cinque pluviometri. Sebbene le misure dei pluviometri rimanga la più affidabile fonte di informazione pluviometrica puntuale, per poter osservare la formazione e l’evoluzione spaziale degli eventi, in grado di influire sulla risposta idrologica del bacino si può sfruttare il contenuto predittivo del radar meteorologico (Ranzi e Bacchi, 1994) e il monitoraggio satellitare. Il radar meteorologico di Monte Macaion è un sistema in banda C operativo dal 2001, aggiornato nel 2023 con tecnologia a doppia polarizzazione. Offre una risoluzione spaziale di 500 m e una risoluzione temporale di 5 minuti. Le immagini di riflettività massima lungo la verticale hanno permesso il monitoraggio dell’evento del 2012, 2018, 2020 e 2021. Per lo studio dell’evento del 2013 di Monte Pinu sono state utilizzati dati ottenuti dal radar meteorologico di Monte Rasu, con caratteristiche simili a quello di Monte Macaion ma con una risoluzione temporale di 15 minuti. Durante l’evento del 2022 del torrente Blè, invece, il dato radar di Monte Macaion non era attivo. Questo evento è quindi stato studiato utilizzando soltanto i pluviometri ed i dati satellitari. Nello studio degli eventi, le osservazioni radar sono state elaborate considerando il ritardo tra la precipitazione rilevata in atmosfera dal radar e l’effettivo istante di impatto al suolo calcolato sulla base di una stima della velocità di caduta delle gocce di pioggia. In questo modo si è potuto valutare la risposta idrologica sulla base di quando la precipitazione ha effettivamente raggiunto il suolo. I dati satellitari sono stati utilizzati, invece, come input per l’approccio multi-sensore denominato MASHA (Multi-sensor Approach for Satellite Hail Advection). La tecnica MASHA è una nuova metodologia ibrida satellitare concepita per il monitoraggio in tempo reale di tempeste violente e nubi grandinigene. Operativamente, MASHA (Laviola et al. 2022) combina i punti di forza del metodo MWCC-H (Laviola et al., 2020a-b), utilizzato per rilevare la grandine attraverso i satelliti in orbita bassa della costellazione GPM, con l’elevata risoluzione temporale (5 minuti) del satellite in orbita geostazionaria Meteosat in modalità Rapid Scan. Le analisi effettuate indicano l’affidabilità della riflettività radar nel rappresentare la tempistica e la distribuzione spaziale della precipitazione, sebbene la trasformazione Z(R) richieda ancora una calibrazione specifica per evento o per tipologia di evento in base ai pluviometri. Le immagini satellitari elaborate tramite l’algoritmo 148 MASHA si sono dimostrate efficaci per l’evento di scala sinottica del 2013 e del 2018, ma non sempre per alcuni eventi convettivi, come quello del 2022. Bibliografia Berti, M., Schimmel, A., Coviello, V., Venturelli, M., Albertelli, L., Beretta, L., Brardinoni, F., Ceriani, M., Pilotti, M., Ranzi, R., Redaelli, M., Scotti, R., Simoni, A., Turconi, and L., Luino, F., 2023. Characterization of a debris flow event using an affordable monitoring system. In E3S Web of Conferences (Vol. 415, pp. 1-4). Bonomelli, R., Pilotti, M., and Heidarian, P., 2024. DEBRA: A multi-rheological 2D steep shallow water finite volume scheme for debris flow propagation in mountain areas, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-4141. Colosio, P., Marmaglio, C., Bonomelli, R., and Ranzi, R. and the Team of debris flow monitoring and control in the Central Italian Alps, 2025. Multisensor monitoring and early warning of precipitation in mountain catchments prone to debris flow events , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16231. Dallan, E., Borga, M., Zaramella, M., and Marra, F., 2022. Enhanced summer convection explains observed trends in extreme subdaily precipitation in the Eastern Italian Alps. Geophysical Research Letters, 49(5), e2021GL096727. Davolio, S., Della Fera, S., Laviola, S., Miglietta, M. M., and Levizzani, V. (2020). Heavy precipitation over Italy from the Mediterranean storm “Vaia” in October 2018: Assessing the role of an atmospheric river. Monthly Weather Review, 148(9), 3571-3588. Ioriatti, E., Reguzzoni, M., Reguzzoni, E., Schimmel, A., Venturelli, M., Albertelli, L., Beretta, L., Brardinoni, F., Ceriani, M., Redaelli, M., Pilotti, M., Ranzi, R., Scotti, R., Simoni, A., Turconi, L., Luino, F., and Berti, M., 2025. Defining rainfall thresholds for debris flows in catchments with short monitoring periods and rare debris-flow events., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12179. Laviola S., V. Levizzani, R. R. Ferraro, and J. Beauchamp, 2020a. Hailstorm detection by satellite microwave radiometers. Remote Sens., 12(4), 621. Laviola S., G. Monte, V. Levizzani, R. R. Ferraro, and J. Beauchamp, 2020b. A New Method for Hail Detection from the GPM Constellation: A Prospect for a Global Hailstorm Climatology. Remote Sens., 12(21), 3553. Laviola, S., Vermi, F., Guarascio, M., Monte, G., Folino, G., and Levizzani, V., 2022. The Multi-sensor Approach for Satellite Hail Advection (MASHA): a new technique for nowcasting applications, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-571. Luino, F., De Graff, J., Roccati, A., Biddoccu, M., Cirio, C. G., Faccini, F., and Turconi, L., 2019. Eighty years of data collected for the determination of rainfall threshold triggering shallow landslides and mud-debris flows in the Alps. Water, 12(1), 133. Prati, A., Menapace, A., and Larcher, M., 2024. Preliminary Statistical Analysis of a Large Hydraulic and Hydrological Dataset for Mudflows and Debris Flows Events in The South Tyrol Region (Italy). 15th International Conference on Hydroinformatics, 394. Ranzi, R., and Bacchi, B., 1992. Analysis and forecasting of rainfall fields observed using radar. In Proc. of Int. Workshop Advances Distributed Hydrology (pp. 327-346). Ranzi, R., B. Bacchi, A. Ceppi, M. Cislaghi, U. Ehret, S. Jaun, A. Marx, C. Hegg and M. Zappa, 2009. Real-time demonstration of hydrological ensemble forecasts in MAP D-PHASE. La Houille Blanche, 5, 95-103. 149 Deep learning-assisted flood depth measurement using flooded car images and video feeds from urban flood events Mayank Mishra1, Raffaele Albano2*, Aurelia Sole1 1 School of Engineering, University of Basilicata, 85100, Potenza, Italy 2 Department of Health Science, University of Basilicata, 85100, Potenza, Italy *e-mail: [email protected] Abstract The objective of the study is to deploy computer vision (CV) techniques to estimate the flood depth levels obtained from video feed (obtained from generic surveillance cameras, citizen captured videos of flood events and/or webcams) to better estimate the flood depth levels from standard objects which are already present in water and whose submergence risk needs to be determined. For example, moving objects such as vehicles/some solid objects/ pedestrians that can be easily detected by “object detection” techniques that can be used as an indirect estimation of flood depths. Previously, most of the applications have been applied on gathered data from the internet and other image data in controlled settings, with limited validation data from citizen videos and real-world scenarios. The deployment of these machine learning models in-situ is limited, and most applications are based on research providing little insights on their use on video feeds, and applicationbased scenarios in real field conditions. Figure 1. Flood depth detection results on YOLOv8 trained algorithm. a) criteria used for annotation (modified from Liu et al. 2024) b-c) sample results on the Matera dataset with boxes showing confidence values for prediction, d) limitations in detection in crowded environments and e) detection results on video feeds with false positives for the sheds of the restaurants (note: b-c, videos from Matera flood events, and d-e video from Genova, Italy floods) For this, we collected image data of cars submerged in flood from literature, internet sources and previous pictures of the case study city of Matera (Southern Italy) that had some previous flood events. The depth levels are divided into levels 0, 1, 2, 3 and 4 as illustrated in Fig. 1a. As per the criteria of Liu et al. (2024), 5 levels 0 (0), 1 (0-35), 2 (35-75), 3 (75-105), 4 (>105 cm) are used and we can take the mean value of car depth for future analysis of the car instability levels. The total number of images we used are 1975, with 6219 annotations with [Document text truncated for crawler view.]