Seasonal climate predictions are a cornerstone in climate and atmospheric sciences, particularly in a climate change adaptation context. By bridging the gap between short-term forecasts and long-term projections, they offer an interdisciplinary framework essential for building resilience in sustainability-critical sectors like agriculture, water management, and disaster risk adaptation. While advances in computational tools and representation of physical processes have improved their usability, performance remains spatially and seasonally heterogeneous. This thesis addresses these limitations by investigating the predictability of impactful climate events and developing computational techniques to identify margins of prediction skills' improvement. The main goal is to contribute to the development of more reliable tools for sustainable decision-making and climate adaptation strategies within an increasingly variable and uncertain environmental context. In this framework, the thesis focuses on the seasonal prediction of impactful rainfall events in the Mediterranean and the Middle East, specifically aiming at droughts as a critical target. The first analysis performed assesses the drought skill emphasizing the regional differences in predictability, and highlighting the reasons behind enhanced performance of some regions, while in others the prediction remains more challenging. Also, the results demonstrate that large scale climate drivers, mostly sea surface temperature (SST), have a crucial influence in modulating the prediction skill. In fact, years characterized by strong anomalies in tropical SST correspond to windows of opportunity in which predictions are likely more reliable. From an operational perspective, this highlights the value of interpreting large-scale drivers' variability as a diagnostic indicator of prediction confidence, offering a strategy to help decision-makers in climate-sensitive sectors. Beyond these aspects, another development concerns the contribution of multi-model ensembles (MME) combinations in optimizing the prediction skill. Combining predictions from different systems offers improvements compared to relying on individual models, confirming the robustness of MME as a strategy for seasonal prediction optimization. This opens the door to advanced approaches where traditional methods of combination may be complemented by complex algorithms, such as the differential evolution (DE). By employing adaptive weighting strategies, DE allows for further exploitation of individual prediction systems, effectively expanding the range of performance optimization under specific operational constraints. Taken together, these results emphasize both the scientific and practical relevance of seasonal predictions. Predictions can indeed provide actionable information, especially when interpreted with the understanding of teleconnections and optimized through ensemble approaches. Beyond advancing scientific understanding, this thesis indicates the pathways through which seasonal prediction can enhance their societal relevance, ultimately contributing to better management of drought risk, with the potential to expand and apply the same approach to other climate-related challenges. Future advancements in multi-model ensembles and a deeper integration of teleconnection dynamics hold the key to surpassing current predictability limits, ensuring that seasonal predictions provide the required reliability for end-users in operational contexts.
Le previsioni climatiche stagionali rappresentano un pilastro fondamentale delle scienze climatiche e atmosferiche, in particolare nel contesto dell'adattamento ai cambiamenti climatici. Colmando il divario tra le previsioni a breve termine e le proiezioni a lungo termine, esse offrono un quadro interdisciplinare essenziale per costruire resilienza in settori critici per la sostenibilità, quali l'agricoltura, la gestione delle risorse idriche e l'adattamento al rischio di catastrofi. Sebbene i progressi negli strumenti computazionali e nella rappresentazione dei processi fisici abbiano migliorato la loro fruibilità, le performance rimangono eterogenee a livello spaziale e stagionale. Questa tesi affronta tali limitazioni indagando la prevedibilità di eventi climatici impattanti e sviluppando tecniche computazionali atte a identificare i margini di miglioramento della capacità predittiva (skill). L’obiettivo principale è contribuire allo sviluppo di strumenti più affidabili per processi decisionali sostenibili e strategie di adattamento climatico, in un contesto ambientale caratterizzato da crescente variabilità e incertezza. In questo quadro, la tesi si concentra sulla previsione stagionale di eventi precipitativi di forte impatto nel Mediterraneo e nel Medio Oriente, focalizzandosi in particolare sulla siccità come obiettivo critico. La prima analisi condotta valuta la capacità predittiva per la siccità, ponendo l'accento sulle differenze regionali e analizzando le ragioni alla base di performance elevate in alcune aree, a fronte di regioni in cui la previsione rimane più complessa. I risultati dimostrano inoltre che i driver climatici a grande scala, principalmente la temperatura superficiale marina (SST), esercitano un'influenza cruciale nel modulare lo skill previsionale. Infatti, gli anni caratterizzati da forti anomalie nelle SST tropicali corrispondono a "finestre di opportunità" in cui le previsioni tendono a essere più affidabili. Da una prospettiva operativa, ciò evidenzia il valore dell’interpretazione della variabilità dei driver a grande scala come indicatore diagnostico della confidenza previsionale, offrendo una strategia di supporto ai decisori nei settori sensibili al clima. Oltre a questi aspetti, un ulteriore sviluppo riguarda il contributo delle combinazioni di ensemble multi-modello (MME) nell'ottimizzazione della capacità predittiva. L'integrazione di previsioni provenienti da sistemi diversi offre miglioramenti significativi rispetto all'affidamento su singoli modelli, confermando la robustezza degli MME come strategia di ottimizzazione delle previsioni stagionali. Questo approccio apre la strada a metodologie avanzate in cui i metodi tradizionali di combinazione possono essere integrati da algoritmi complessi, come l'evoluzione differenziale (Differential Evolution, DE). Attraverso l'impiego di strategie di pesatura adattiva, la DE permette di sfruttare ulteriormente i singoli sistemi di previsione, espandendo efficacemente i margini di ottimizzazione sotto specifici vincoli operativi. Complessivamente, questi risultati sottolineano la rilevanza sia scientifica che pratica delle previsioni stagionali. Esse possono infatti fornire informazioni azionabili, specialmente se interpretate alla luce delle teleconnessioni e ottimizzate attraverso approcci ensemble. Oltre a far progredire la comprensione scientifica, questa tesi indica i percorsi attraverso i quali le previsioni stagionali possono accrescere la propria rilevanza sociale, contribuendo a una migliore gestione del rischio siccità e offrendo un approccio applicabile ad altre sfide climatiche. I futuri progressi negli ensemble multi-modello e una più profonda integrazione della dinamica delle teleconnessioni rappresentano i concetti chiave per superare gli attuali limiti di prevedibilità, garantendo che le previsioni stagionali forniscano l'affidabilità richiesta nei contesti operativi degli utenti finali.
Previsioni climatiche stagionali di eventi precipitativi di forte impatto: valutazione e opportunità / Dal Monte, Thomas. - (2026 May 13).
Previsioni climatiche stagionali di eventi precipitativi di forte impatto: valutazione e opportunità
DAL MONTE, THOMAS
2026-05-13
Abstract
Seasonal climate predictions are a cornerstone in climate and atmospheric sciences, particularly in a climate change adaptation context. By bridging the gap between short-term forecasts and long-term projections, they offer an interdisciplinary framework essential for building resilience in sustainability-critical sectors like agriculture, water management, and disaster risk adaptation. While advances in computational tools and representation of physical processes have improved their usability, performance remains spatially and seasonally heterogeneous. This thesis addresses these limitations by investigating the predictability of impactful climate events and developing computational techniques to identify margins of prediction skills' improvement. The main goal is to contribute to the development of more reliable tools for sustainable decision-making and climate adaptation strategies within an increasingly variable and uncertain environmental context. In this framework, the thesis focuses on the seasonal prediction of impactful rainfall events in the Mediterranean and the Middle East, specifically aiming at droughts as a critical target. The first analysis performed assesses the drought skill emphasizing the regional differences in predictability, and highlighting the reasons behind enhanced performance of some regions, while in others the prediction remains more challenging. Also, the results demonstrate that large scale climate drivers, mostly sea surface temperature (SST), have a crucial influence in modulating the prediction skill. In fact, years characterized by strong anomalies in tropical SST correspond to windows of opportunity in which predictions are likely more reliable. From an operational perspective, this highlights the value of interpreting large-scale drivers' variability as a diagnostic indicator of prediction confidence, offering a strategy to help decision-makers in climate-sensitive sectors. Beyond these aspects, another development concerns the contribution of multi-model ensembles (MME) combinations in optimizing the prediction skill. Combining predictions from different systems offers improvements compared to relying on individual models, confirming the robustness of MME as a strategy for seasonal prediction optimization. This opens the door to advanced approaches where traditional methods of combination may be complemented by complex algorithms, such as the differential evolution (DE). By employing adaptive weighting strategies, DE allows for further exploitation of individual prediction systems, effectively expanding the range of performance optimization under specific operational constraints. Taken together, these results emphasize both the scientific and practical relevance of seasonal predictions. Predictions can indeed provide actionable information, especially when interpreted with the understanding of teleconnections and optimized through ensemble approaches. Beyond advancing scientific understanding, this thesis indicates the pathways through which seasonal prediction can enhance their societal relevance, ultimately contributing to better management of drought risk, with the potential to expand and apply the same approach to other climate-related challenges. Future advancements in multi-model ensembles and a deeper integration of teleconnection dynamics hold the key to surpassing current predictability limits, ensuring that seasonal predictions provide the required reliability for end-users in operational contexts.| File | Dimensione | Formato | |
|---|---|---|---|
|
thesis_dalmonte_thomas_v2.pdf
embargo fino al 13/05/2027
Descrizione: Tesi def
Tipologia:
Tesi di dottorato
Dimensione
89.65 MB
Formato
Adobe PDF
|
89.65 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


