In proximity of a seismogenic source (near-fault or near-source), meaning for maximum distances ranging from about one km to a few tens of km as a function of earthquake magnitude, the ground motion produced by strong earthquakes presents typical characteristics in terms of amplitude, duration and frequency content, and it may be characterised by the forward-directivity (FD), neutral-directivity (ND) or backward-directivity (BD) phenomena. As a consequence of such directivity phenomena, for the case of forward-directivity the velocity waveform includes one or two pulses, typically of long duration, producing significant motion amplification for the sites in the direction of rupture propagation, while for the case of backward-directivity it includes a long duration and lower amplitude motion for the sites in the opposite direction. Any element at risk located in a region close to an active fault may also be subjected to a permanent static displacement due to fault slip, namely fling step. These characteristics may significantly affect the seismic response of structures in nonlinear field, and in general the extent and distribution of damage on urban infrastructural systems (such as buildings, lifelines, critical facilities). This paper aims to estimate the role played by near-fault motions in the seismic risk assessment of lifelines composed of buried pipelines, such as water supply systems (WSS). The employed methodology involves four levels: seismic hazard, seismic fragility of components, systemic performance and uncertainty. First, several near-fault seismic records, characterized by the presence of an evident velocity pulse, are selected in order to highlight and take into account the FD effects; in particular, the pulses extracted from the records are used to obtain the probability distribution of a novel modification factor for Peak Ground Velocity (PGV), computed by an existing ground motion prediction equation. Also, a fling step model is selected from the ones currently available in the literature. Successively, the seismic demand is applied to the components of infrastructural systems by means of fragility or vulnerability functions, accounting for the uncertainty related to the physical damage state as a function of local seismic intensity. Then, a number of performance metrics are introduced to quantitatively measure the performance of a single system or the whole infrastructure (considering the interdependencies between systems). Two Monte Carlo simulations, in which the near-source effects are neglected and considered, respectively, are then employed to carry out a probabilistic analysis including the treatment of several uncertainties in the problem (in both hazard and system parts). The comparison of results from the two simulations allows to assess the importance of including near-source phenomena in seismic risk assessments. The methodology was implemented as an extension of a civil infrastructure simulation tool, namely Object-Oriented Framework for Infrastructure Modeling and Simulation (OOFIMS), coded in MATLAB® language and recently developed in the SYNER-G (2012) European project. The tool was used to carry out an example application on a realistic WSS. The findings of this study are of use for emergency managers and lifeline asset managers in tectonically-active urban settings, in order to increase the seismic resilience of communities.