The joint probability of extreme wave height and surge at 9 representatively selected stations along China's coasts are analysed using the results extracted from long-term model simulations over a period of 35 years. Three sampling methods which are used to provide the datasets for extreme analysis are detailed and discussed. It is found that the structural response (SR) sampling method shows the best performance in describing the correlations between extreme wave heights and surges, particularly in the typhoon-affected areas, in comparison with both wave dominated (WD) and surge dominated (SD) sampling methods. The distribution of joint exceedance probability and joint probability density for these 9 selected stations are calculated and presented. The results from the joint probability analysis show that the extreme wave height and surge in the study region are partly related, and their correlation becomes stronger where the sea states become more energetic. The comparisons of the combined water level of the wave height and surge at given return periods obtained from the joint probability density and the empirical method indicate an overestimation from the commonly used empirical method in determining the joint extreme water level, which could make coastal structure designs less economic.