A Nonparametric Kernel Regression Estimator for Flood Frequency Analysis

Abstract

A nonparametric kernel regression for estimating flood frequency quantile of annual maximum flood events at a gaged site is presented in this paper. Parametric estimators and a nonparametric kernel regression estimator (NK) are compared for three situations—Gaussian data, Skewed data (3 parameter Gamma) and Gaussian Mixture data. Since the results of parametric estimators varied according to the situation, it is not easy to say which estimator is the best. However, the performance of the nonparametric kernel quantile estimator (NK) was relatively consistent across the estimation situations considered in terms of bias and root mean square error (RMSE).