We plot the dike costs divided by their length as a function of height and test four different regression models. Our analysis shows that a linear function without intercept is sufficient to model the costs, i.e. fixed costs and higher-order contributions are less significant. We employ log-normal distributions and calculate that the range between 3x and x/3 contains 95% of the data, where x represents the regression value. We compare estimates from Canada, the Netherlands, US, UK, and Vietnam.