Observed data of coastal inundation are very rare, yet are essential for testing the performance of simulation models for this significant natural hazard. In this paper we therefore examine the extent to which observed data can constrain predictions of a typical flood inundation model for a coastal flood event that affected the Somerset Levels in the UK on 13th December 1981. In doing so, a detailed reconstruction is made of the flooding that occurred along a 20 km stretch of the North Somerset coast, potentially providing a benchmark data set for comparative model analysis. Information on the extent of flooding was gathered from photographs taken on the day following the storm (‘hard’ or quantitative data), and contemporary newspaper reports and interviews with a number of people who had first hand experience of the storm and the consequent flooding (‘soft’ or qualitative data). Data on tidal levels and wave overtopping rates were collated from contemporary reports. Information on the topography of the study area and the configuration and heights of the sea defences at the time was determined from a variety of sources. Across the whole domain the model correctly predicted 85% of the flooded area; a level of performance comparable to fluvial applications of inundation models. The performance of the model in areas where flooding was dominated by wave overtopping was shown to be very sensitive to relatively small changes (±50%) in the specified flux rate, yet this parameter can plausibly vary by as much as ±300%. The overtopping flux rate was shown to be a much more sensitive control on the model performance than friction parameters. Finally, the study shows that 2D hydraulic models with relatively simple physics can closely simulate real coastal flood events given the availability of high resolution, high accuracy terrain data such as LiDAR. These conclusions mirror those arrived at over the last decade for fluvial flooding and suggest that the hydrodynamics of coastal and fluvial floodplain inundation are similar, if not identical.