Modelling Hydrological extremes is an important issue in hydrology and water resources research. In Statistics, there are two types of extreme value theory. Classic extreme value distributions, including Weibull, Gumbel and Fréchet distributions, are derived for the normalized extreme value of a data set and are used to model block maxima such as annual data. The peak over threshold (POT) models the tail distributions directly and therefore can be used for daily data. Under mild conditions, the exceedance over threshold asymptotically follows the so-called generalized Pareto distribution (GPD) as the threshold increases.  In this talk, we discuss the use of the so-called extended Burr XII distribution Their performances are assessed by comparing with the other popularly used distribution is hydrology using both real data and simulation study. However, the parameter estimation for the extended Burr distribution is not straightforward due to the so-called embedded model problem. In this talk we will also provide the solution for this problem.