Influence of Uncertainties in the Estimation Procedure of Floodwater Level

Abstract

Despite the recent development of risk-based assessments in flood defense, in practice, flood levels in channels are generally estimated for a design return period by using the discharges of the same return period. The flood levels are also influenced by other factors such as bed roughness, flow area, wetted perimeter, and friction slope which are random in nature. The surveyed cross sections and calibrated values of roughness coefficients are generally used without any allowance for their variability to assess the water levels based on discharge of a selected return period. This simplifies a multivariate random process to a single-variate random process. When the uncertainties of selected influencing parameters were considered in sample studies using Monte Carlo simulation, it was found that the traditional procedures result in an underestimation of water level at high return periods and over estimation of water levels at low return periods. The overall effect is the diffusion of the density function from the center toward the tails of the distribution. Sample studies using the variation in channel roughness and friction slope indicated that the return period of the water level, based on a 100 year return period discharge, varied from 32 to 82 years, depending on the statistical properties of the influencing random parameters. The frequency analysis of water levels was undertaken by analyzing 25–800 million generated data sets.