The research presented in this paper involves the application of the joint probability method to the estimation of extreme water levels resulting from astronomical tides and surge residuals and the investigation of the effects of tide–surge interactions on extreme water levels. The distribution of tide peaks was analysed from field records (<20 years) and a 46-year dataset of monthly maximum tidal amplitudes. Large surges were extracted from both field records and a numerical model hindcast covering the 48 largest storm events in the Irish Sea over the period 1959–2005. Extreme storm surges and tides were independently modelled using the generalised extreme value statistical model, and derived probability distributions were used to compute extreme water levels. An important, and novel, aspect of this research is an analysis of tide–surge interactions and their effects on total water level; where interactions exist, they lead to lower total water levels than in the case of independency. The degree of decrease varies with interaction strength, magnitude of surge peak at a particular phase of tide and the distribution of peaks over a tidal cycle. Therefore, including interactions in the computation of extreme levels may provide very useful information at the design stage of coastal protection systems.