Comparison of Artificial Neural Network Methods with L-moments for Estimating Flood Flow at Ungauged Sites: the Case of East Mediterranean River Basin, Turkey

摘要: A regional flood frequency analysis based on the index method is applied using probability distributions commonly utilized for this purpose. The distribution parameters are calculated by of L-moments with data annual peaks series recorded at gauging sections 13 unregulated natural streams in East Mediterranean River Basin Turkey. artificial neural networks (ANNs) models (1) multi-layer perceptrons (MLP) networks, (2) radial basis function (RBNN), and (3) generalized regression (GRNN) developed as alternatives to method. Multiple-linear multiple-nonlinear (MLR MNLR) also used study. these indicates that hydrologically homogeneous a whole. Among tried which Generalized Logistic, Extreme Vaules, Normal, Pearson Type III, Wakeby, Pareto, Logistic Values pass Z statistic goodness-of-fit test Basin, former performing yet better than latter. Hence, outcome distribution, two equations estimate any return periods un-gauged site study region. ANNs, MLR MNLR trained tested gauged sites. results show predicting performance MLP model superior others. application performed special Matlab code, yields logarithm peak, Ln(QT), versus desired period, T.