Flooding is a natural phenomenon that regularly causes financial and human devastation around the world. In many countries the risk of flooding is managed by society through a combination of governmental agencies and the insurance industry. For both these types of organisation an estimate of the largest, or most widespread, events that can be expected to occur is useful. Such estimates can be used to help in preparing or co-ordinating flood mitigation activities and by the insurance and re-insurance industries to assess financial risk. In this paper we develop a method to simulate a set of synthetic flood events that can be used to estimate the probability of widespread floods. We demonstrate this method using data from a set of UK river flow gauges. The model used in this simulation process is based on the conditional exceedance model of Heffernan and Tawn, extended to incorporate features typically found in the data for extreme river floods. We also present an improved estimation method for the model parameters and demonstrate its advantages through the results of a simulation study. The benefits of the method over previous models used are that it provides a theoretical basis for extrapolation and is flexible enough to account for varying strengths of extremal dependence that are observed in flood data.