The recent development of high-strength (HSS) weldable steels has enlarged the range of design alternatives for the optimization of high-head steel-lined pressure tunnels and shafts (SLPT&S) in the hydropower industry. With the liberalization of the European energy market and increasing contribution of new renewable volatile energies in the electricity grid due to high subsidies, storage hydropower and pumped-storage plants are subject to more and more severe operation conditions resulting in more frequent transients. The use of HSS allows the design of thinner and thus more economic steel liners. However, welded HSS do not provide higher fatigue resistance than lower steel grades, and may be particularly subject to the risk of cold cracking in the weld material as dramatically illustrated by the failure of the Cleuson-Dixence pressure shaft in 2000. Fatigue behavior may become the leading limit state criterion. This research project aims at improving the comprehension of the mechanical behavior of SLPT&S and at developing a framework for probabilistic fatigue crack growth and fracture assessment of crack-like flaws in the weld material of longitudinal butt welded joints, considering all possible steel grades for high-head hydropower schemes. The influence of anisotropic rock behavior and geometrical imperfections at the longitudinal joints on the structural stresses have been studied by means of the finite element method accounting for the interaction with the backfill concrete-rock multilayer system. Parametric correction factors have been derived to estimate stress concentrations and structural stresses in steel liners with ease in practice, allowing the use of $S$-$N$ based engineering fatigue assessment approaches. Stress intensity factors (SIF) for axial cracks in the weld material of the longitudinal joints have also been obtained by means of computational linear elastic fracture mechanics (LEFM). The use of the previously developed parametric equations in the classical formulas for SIF in cracked plated structures has been validated, and new parametric equations for the weld shape correction have been proposed. A probabilistic model for fatigue crack growth assessment has been developed in the framework of LEFM in combination with the Paris-Erdogan law. The probability of failure is estimated by means of the Monte Carlo simulation procedure, in which the crack growth rate parameters and the crack shape ratio are defined as stochastic variables. A week-long normalized loading spectrum derived from prototype measurements on an alpine pumped-storage hydropower plant in Switzerland is used. This approach provides relative and quantitative results through parametric studies, giving new insights on the fatigue behavior of steel liners containing cracks in the weld material of the longitudinal joints. Finally, a fatigue assessment case study is presented, detailing the entire calculation procedures developed in this research. It aims at ensuring the transfer of knowledge toward practitioners.