Preferable probabilistic model for annual maximum snow depths in China

Annual maximum snow depths (AMSD) from 120 meteorological stations in China, each with at least 40 years of data, are considered to assess the preferred probability model for AM SD. It is found that the lognormal distribution is preferable to the Gumbel distribution for data from most of the considered stations, indicating that if a single distribution is to be adopted for statistical modelling of AMSD across the whole country, the lognormal distribution is better to be used. It is found that, on average, the estimated 50-year return period value of the AMSD by using the lognotmal distribution, depending on the different fitting methods, is about 6% to 13% greater than that estimated by using Gumbel distribution.