Assessment and modelling of hydroclimatic extremes for resilient water system management under a changing climate

Water resources are under growing pressure due to the intensifying impacts of climate change, which are altering precipitation regimes, increasing drought frequency, and amplifying the magnitude of extreme hydrological events. These changes directly affect the stability and reliability of water-dependent systems such as agriculture, hydropower production, and hydraulic infrastructures. Understanding and modelling hydroclimatic extremes is therefore crucial for designing adaptive and resilient water management strategies. The main objective of this dissertation is to develop and apply data-driven and physically consistent methodologies, applicable across multiple spatial and temporal scales. These methodologies are designed to evaluate and build models capable of describing hydroclimatic extremes and projecting their future evolution based on the outputs of climate models. The study tackles this challenge through four complementary lines of investigation integrating statistical innovation, large-scale data analysis, and applied modelling, with the aim of advancing the understanding and assessment of hydroclimatic extremes under a rapidly changing climate. The first part of this dissertation focuses on the characterization of precipitation extremes across Europe through the application of the Obesity Index (OB), a non-parametric and empirical indicator used to quantify the heaviness of the tails of precipitation probability distributions. By applying the OB to the high-resolution gridded E-OBS dataset, the study demonstrates the ability of the index to identify regional variations in tail behavior without the need to select explicit thresholds for extremes detection. The analysis provides evidence that the heaviness of the distribution tails is spatially heterogeneous and may be influenced by data interpolation procedures. These findings contribute to improving the understanding of extreme precipitation events, with implications for risk assessment. The second part of the thesis investigates simultaneous drought occurrences across global croplands, addressing a major gap in understanding compound agricultural risks. Using the Standardized Precipitation Evapotranspiration Index (SPEI) over the 1902-2022 period, the analysis examines 17 major crops representing 65% of global farmland and over 70% of global protein production. Results show a significant positive trend in both the total drought-affected area and drought severity, with at least 20% of the cultivated land, weighted by crop yield, experiencing drought in two-thirds of the years. The findings indicate that the frequency of simultaneous droughts has increased, particularly in tropical and subtropical regions, posing serious threats to global food security. This large-scale assessment highlights the growing systemic vulnerability of agricultural production under concurrent climatic shocks. The third research component explores the response of hydropower production to future climatic conditions in the Upper Colorado River Basin (UCRB), focusing on the Lake Powell, one of the largest reservoirs in the United States. Statistical models are developed to relate spatially and temporally averaged precipitation and temperature across the basin to monthly hydropower production between 2005 and 2022. These models are then forced with bias-corrected CMIP6 outputs under four Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) to assess future changes in hydropower production. The projections reveal that higher emission scenarios tend to produce stronger positive trends in generation during warmer months, while lower emission scenarios exhibit greater variability and weaker impacts. These results underscore the sensitivity of hydropower systems to precipitation patterns and temperature variations, and emphasize the need for adaptive management to ensure reliable energy production under future climatic uncertainties. Finally, the dissertation includes a comprehensive review of rainfall nowcasting models, which are critical for enhancing early warning systems and short-term water management. The review compares numerical weather prediction models, stochastic approaches, and emerging deep learning methods, emphasizing their respective strengths, limitations, and potential for integration. Particular attention is given to predictive uncertainty and model blending as key elements for developing robust nowcasting systems capable of supporting decision-making in real time. The discussion outlines promising research directions to improve forecasting accuracy, data assimilation, and communication of uncertainty to end-users. Overall, these four research components provide a coherent framework for evaluating, modelling, and projecting hydroclimatic extremes across multiple spatial and temporal scales. From the statistical characterization of precipitation distributions to the global monitoring of droughts, from climate projections of precipitation and temperature for estimating hydropower production to the evaluation of forecasting models, this work offers both methodological and applied contributions to the field of hydroclimatology. The overall findings highlight that building resilience in water systems requires an integrated approach that combines quantitative analyses, scenario-based modelling, and proactive adaptation strategies. By deepening the understanding of extreme hydroclimatic processes, this dissertation supports the development of sustainable and adaptive management practices, essential to ensure water security in an era of accelerating climate change.