The long-term variation and seasonal variation of sea level have a notable effect on the calculation of engineering water level. Such an effect ts first analyzed in this paper. The maximal amplitude of inter-annual anomaly of monthly mean sea level along the China coast is larger than 60cm. Both the storm surge disaster and cold wave disaster are seasonal disasters in vario us regions, so the water level corresponding to the 1% of the cumulative frequency in the cumula tire frequency curve of hourly water level data for different seasons in various sea areas is different from design water level. for example, the difference between them reaches maximum in June. July and August for northern sea area, and maximum in September, October and November for Southern China Sea. The hourly water level data of 19 gauge stations along the China coast are analyzed. Firstly, the annual mean sea level for every station is obtained; secondly, linear chan ging rates of annual mean sea level are obtained with the stochastic dynamic method; thirdly, the astronomical tide and storm surge tide are obtained by subtracting the linear fitting part from the original hourly data. finally, two distributions corresponding to the astronomical tide and wind tide are obtained according tn whether the astronomical tide and storm tide are correlative or not. So the two check water levels are obtained with the joint probability method. The maximal difference between the two water levels of 100 years' recurrence is more than 30cm. Both of the two check wat er levels have disadvantages in the use of observation data, so the mean value is suggested to be taken as the final check water level. A comparison between the two check water levels indicates that the effect of sea level variation upon design water level and check water level is larger than 80 cm at some stations.