Abstract The occurrence rates of extremes of the natural forces, wave heights and sea levels, are one of the important factors in the design of coastal structures. Especially detecting the difference between the past and future rates is becomingrecently an issue remarked in the discussions of climate change adaptation. However the difference is often so faint that the significance has not been examined appropriately nor even discussed in many previous studies of statistical methods of extreme value analysis. One might wonder if the significance is well shown in selecting the non-stationary model in comparison with the stationary one. However the difference of those models should be considered to be checked in the region where thefrequentlyobserved values are concentrated, while it will not always hold true in the outer region where the observed values are sparse. It is the same point of arguments as existing the limitation of extrapolation even if the parameters of distribution function are estimated. In the previous research, a criterion of possible extrapolation is given by the degree of experience for one sample. This research discussesto compare two populations of sea extremes by extending the Poisson test fortwo-sample.