Abstract
The global transition toward sustainable energy systems requires sophisticated approaches to address the complex interdependencies between water and energy resources, particularly in the context of hydropower systems. This dissertation presents a comprehensive data-to-decision framework for optimising the water–energy nexus through three interconnected research components: meteorological data processing with uncertainty quantification, climate model bias correction, and optimisation of hybrid renewable energy systems. The focus is on Alpine hydropower infrastructure in the Trentino South Tyrol region of Italy, which represents a critical case study for water-energy integration in complex mountainous terrain. The first component develops the ALPINE-TST-250 dataset, a novel high-resolution gridded dataset featuring hourly precipitation and temperature at 250-meter spatial resolution spanning 1991–2021. Generated using kriging with external drift interpolation that incorporates elevation as an auxiliary variable, the dataset draws from over 300 meteorological stations across the study region and adjacent areas. Comprehensive quality control and cross-validation procedures ensure robustness, yielding average hourly RMSE (MAE) values of 0.21 mm (0.09 mm) for precipitation and 1.79°C (1.24°C) for temperature. Uniquely, the dataset includes spatially explicit uncertainty estimates, providing critical information for hydrological applications and climate impact assessments in data-sparse mountainous regions. The second component conducts a rigorous multi-scale evaluation of bias correction methods for climate model outputs. This systematic review and comparative analysis assess both univariate and multivariate statistical techniques alongside machine learning models across hourly, daily, and monthly temporal resolutions. The investigation reveals how method effectiveness varies with temporal aggregation and identifies optimal bias-correction approaches for different application contexts, directly supporting improved operational planning for hydropower systems under climate variability. The third component develops an advanced optimization framework for hydropower systems using machine learning and reinforcement learning techniques. The Combined-Agent Reinforcement Learning (CARL) framework integrates short-term electricity price forecasting with hydropower operational optimisation, incorporating multiple system configurations including traditional hydropower plants, pumped storage facilities, and floating photovoltaic installations. Application to real-world case studies demonstrates significant revenue enhancements while maintaining environmental and operational constraints. The integration of pumped storage systems with renewable energy shows particular promise for enhancing economic returns and grid stability, with results indicating revenue improvements through optimized pumped storage operations and strategic renewable integration. Together, these research components demonstrate the value of a comprehensive data-to-decision framework that bridges hydrometeorology, climate science, and advanced optimisation techniques. The methodologies developed are directly applicable to Alpine hydropower infrastructure and scalable to other mountainous regions worldwide, contributing to Europe’s renewable energy transition while maintaining water resource sustainability and energy system reliability.