Regional frequency analysis (RFA) of hydro-meteorological variables is a commonly used tool to provide quantile estimates of extreme events at ungauged sites. Given the high complexity of hydro-meteorological processes, it is worthwhile to account for the possible nonlinear connections between hydro-meteorological variables and catchments characteristics in all RFA steps. Moreover, to provide relatively reliable quantiles estimates, it is often recommended to only consider sites with sufficiently long data series which lead to ignoring a considerable part of the available information. A number of regression-based RFA methods are proposed to remedy the limitations of the classical approaches, mainly dealing with the non-linear aspect and the exploitation of the hydro-meteorological data. Comprehensive comparisons are carried out between the classical and the proposed methodologies using 151 hydrometric stations from the province of Quebec. The performances of the proposed methods are assessed using new and classical evaluation criteria in a cross-validation procedure.
About the speaker:
Dr. Dhouha Ouali is a PCIC Research Associate with the Marine Environmental Observation Prediction and Response Network. Her research at PCIC is focused on extratropical cyclones and their potential impacts on British Columbia’s power transmission grid.