Application of artificial neural network, multiple-regression and index-flood techniques in regional flood frequency estimation

Abstract

Flood frequency estimation is fundamental in both engineering science and engineering hydrology science. Comparing the efficiencies of artificial neural network (ANN), multiple-regression (MR) and index-flood (IF) techniques based on L-moments in Qazvin Province of Iran was the main objective of this study. Using the main variables affecting flood magnitude, the study area was divided into two regions based on the clustering approach. The homogeneity of these regions was confirmed using the homogeneity test of L-moments approach. Using the L-moment ratios and the Z-statistic criteria, generalised logistic (GLO) and generalised Pareto (GPA) distributions were identified respectively for the first and second homogen regions as the most robust distributions among five candidate distributions. Relative root mean square error (RRMSE) measure was applied for evaluating the performance of three methods in comparison with the curve fitting method. In general, ANN method gives more reliable estimations.