Non-stationary lognormal model development and comparison with the non-stationary GEV model

Classical flood frequency analysis (FFA) requires stationarity of the observed data set. This hypothesis is not always verified for observed data. This restriction can be lifted if classical distributions used in FFA integrate non-stationarity by considering time dependent parameters. It is also possible to consider other covariates instead of the time. The conditional distribution can then be obtained with respect to given values of the covariates. In the present study the non-stationary lognormal (LN) model with linear and quadratic dependence is presented, and corresponding maximum likelihood equations are developed. These models are compared to the non-stationary generalized extreme value (GEV) models by Monte Carlo simulation. The non-stationary LN model is also applied to a case study to illustrate its potential. The case study deals with the analysis of the annual maximum precipitation at the Tehachapi Station in California, USA, with the Southern Oscillation Index (SOI) as a covariate.