Engineering systems, such as power or water networks, are exposed to hazards whose impact is spatially distributed. Hazard scenarios can be characterized through triggering parameters θ, for example the magnitude and location of a seismic event, or the spatially and temporally distributed precipitation for a flood event. For risk management, e.g. for contingency planning and emergency response preparation, it is relevant to identify hazard scenarios that are in some ways representative of a given hazard magnitude. For example, one would like to identify the earthquake magnitude and location that is representative of the 100-year event. More generally, one may want to identify the parameter values θ that produce a scenario with a return period of interest T. This return period is with respect to the losses in the system, and hence the scenario is a property both of the hazard but also the system response. In this paper we propose a methodology for determining representative hazard scenarios associated to the system response. It is based on applying inverse First Order Reliability Method (inverse FORM) for finding the scenario with the largest losses among those corresponding to an exceedance probability of 1/T (Winterstein et al. 1993, Berk et al. 2017). The methodology is demonstrated by application to an idealized lifeline system.