Abstract: Different return periods of annual extreme ice thickness are the key indexes for ocean structure design of ice-infested area. Only the point estimations of sea ice thickness have been given in the current practical engineering design. However, estimation of confidence intervals can consider the uncertainty of design values of sea ice thickness, which play an important role to provide guidance for the design and construction of marine structures. Based on the maximum likelihood method of interval estimation, the solving processes of confidence intervals are presented using Gumbel distribution, Weibull distribution with three parameters, log-normal distribution with three parameters and Pearson type III distribution. Taking the data of annual extreme floating ice thickness in the northern area of Bohai Sea as examples, the above four types of distributions are adopted to fit the observed data at Yingkou and Huludao. The confidence interval of the design ice thickness under different return periods are estimated by maximum likelihood method. The results show that the confidence intervals of the design sea ice thickness calculated by Pearson type III distribution are better than those by the other three distributions.