Anthropogenic noise emission in the marine environment has always been an environmental issue of serious concern. In particular, the noise generated during the installation of foundation piles is considered to be one of the most significant sources of underwater noise pollution. This is mainly attributed to the recent developments in the offshore wind industry. To meet the increasing demand for energy from renewable resources, a large number of offshore wind farms are planned to be constructed in the near future. Despite the plethora of the available foundation concepts to accommodate the tower of an offshore wind power generator, the steel monopile is the most widely used and economically profitable choice for wind turbines installed in shallow water depths. The installation of foundation piles requires a tremendous amount of input energy and the development of special equipment. The piles are usually driven into the seabed with a hydraulic impact hammer that is positioned at the head of the pile and delivers a series of blows, forcing the pile to gradually progress into the sediment. This process is associated with strong impulsive noise that is emitted into the underwater environment and can be detected tens of kilometres away from the construction site. The latter is considered to be harmful for the aquatic species, especially for mammals who exhibit sensitive auditory behaviour. To date, the available knowledge regarding the physics of noise generation caused by marine piling is limited. Without a proper understanding of the noise generation mechanisms, any attempt to mitigate the noise will fall short of expectation. This study aims to fill this knowledge gap. The primary goal is to shed new light on the underlying physics of the underwater noise generated by marine piling in order to help the practitioners to develop more efficient noise mitigation equipment. To reach this aim, the following objectives were set: (i) development of a vibroacoustic model that is computationally efficient and can reproduce the physical mechanisms of underwater sound generation and propagation; (ii) analysis of data collected during an experimental campaign in order to identify the main sources that contribute to the underwater noise pollution; and (iii) theoretical investigation of the effectiveness of a chosen noise mitigation technique and, if possible, generalisation of the conclusions to several other noise mitigation concepts. Regarding the first objective mentioned above, considerable effort has been placed in the development of computationally efficient models to describe the radiated sound field caused by marine piling. The models consist generally of three subsystems, i.e. the pile, the water and the soil. The hammer is substituted by a distributed force exerted at the pile head. A high-order thin shell theory is adopted for the description of the pile dynamics while the seawater is modelled as a compressible fluid medium. For the soil, several approaches are considered which mark the actual differences between the various models. In the \textit{simplified model}, the soil is represented by linear springs and dashpots. In the \textit{advanced model}, the soil is modelled as a layered three-dimensional elastic continuum which is considered to be a more realistic representation of the actual environment. The solution approach adopted to describe the coupled vibroacoustic behaviour of the pile-water-soil system is semi-analytical. It is based on the expansion of the response of the total system in terms of two complete sets of eigenmodes: the \textit{in vacuo} structural modes and the modes of the exterior domain. The modal coefficients are subsequently determined by an appropriate combination of the kinematic conditions that are imposed at the pile-water and pile-soil interfaces together with the use of the orthogonality relations of each set of eigenmodes. This method of solution is considered to be advantageous for several reasons: (i) the computational time is considerably reduced when compared, for example, to the finite element or the boundary element methods; (ii) different stages of the installation process can be investigated with minimum computational effort; and (iii) considerable insight is gained into the physics of noise generation and into the contribution of the various modes to the total acoustic field. The predictions of the models show that the acoustic field in the seawater consists of pressure conical waves (\textit{Mach cones}) generated by the \textit{wave packet} propagating along the pile after the hammer impact. As the wave packet enters the soil region, both shear and compressional waves are radiated, with the former being much stronger than the latter. At later moments in time, \textit{Scholte} waves are observed along the seabed-water interface. These propagate with a velocity slightly lower than that of the shear waves in the soil medium, their energy is localized in the vicinity of the seabed-water interface, and they experience much less attenuation when compared to other propagating modes. A parametric study was also conducted in order to determine the critical parameters of the system and the way they influence the radiated sound. Among the various parameters examined, the pile diameter and the soil properties were found to be the most influential ones in the determination of the underwater sound field. The former defines largely the frequency spectrum of the radiated sound whereas the latter the energy distribution among the various subsystems. In particular, the shear rigidity of the soil was found to be crucial for the correct estimation of the noise levels close to the seabed surface due to the energy transferred into the interface waves. The higher the shear rigidity of the seabed, the larger the penetration depth of the Scholte wave into the fluid and, consequently, the higher the noise levels close to the seabed-water interface. Soil stratification is also important but the influence is mainly governed by the properties and depth of the upper soil layer which is in direct contact with the seawater. Since the soil modelling was found to be critical for the prediction of the noise levels, the results obtained with the pile-water-soil model were also compared with another model in which the soil was substituted by an acoustic fluid with additional dissipation to account for the energy transferred into the shear waves (\textit{fluid model}). The substitution of the soil by a modified acoustic fluid is often favoured in underwater acoustics in order to improve the computational speed. In this study we show that the acoustic approximation of the seabed can yield inaccurate results when certain conditions are not met. Additionally, it has the tendency to underestimate the noise levels especially close to the seabed level. Thus, the substitution of the soil by an acoustic medium should always be carried out with great cautiousness in applications related to marine piling, even if one is interested only in the prediction of the sound field in the seawater region. With regard to the second objective, time series data collected during a measurement campaign were analysed in the time, the frequency, and the time-frequency domains. The analysis has shown that the dynamic response of the system is blow-invariant provided that the input energy and penetration depth remain almost constant. The vibration modes with frequencies below the ring frequency of the shell structure \textit{in vacuo} were shown to be best coupled to the surrounding fluid and therefore able to radiate considerable energy into the exterior fluid domain. This observation was, in fact, verified by the model predictions. In addition, the sound levels in the near-field region showed a strong depth-dependence. Finally, the analysis of the geophone signals verified the existence of low-frequency oscillations close to the seabed level which are attributed to the Scholte waves that were predicted by the prediction models. Regarding the third objective, the final chapter of the thesis is devoted to noise mitigation. A state-of-the-art review of the available mitigation concepts is included and a final model is developed which includes an air-bubble curtain. A parametric study is conducted in order to reveal the principal mechanism of noise reduction and the optimum system configuration for a specific case. The influence of a number of parameters, i.e the volume of the air content, the thickness of the bubble curtain and the distance from the pile surface, on the predicted sound levels, were investigated. It is found that for piles of large diameter, the main mechanism responsible for the noise reduction is the impedance contrast between the seawater and the air-bubble medium. The dissipation effects due to resonance of the individual bubbles seem not to be important for the relatively low-frequencies associated with the sound radiation of large monopiles. Additionally, the efficiency of the air-bubble curtain increases, the higher the air-volume content, and the larger the horizontal distance it is placed from the pile.