Title: Uncertainty and sensitivity analysis in developing surrogate model of flood control channel design in an urban area
Authors: Masoume Behrouz; Saeed Alimohammadi
Addresses: Department of Civil Engineering, Civil, Water and Environmental Engineering Faculty, Shahid Beheshti University, P.O. Box 16765-1719, Bahar Blvd., Hakimieh, Tehran 1658953571, Iran ' Department of Civil Engineering, Civil, Water and Environmental Engineering Faculty, Shahid Beheshti University, P.O. Box 16765-1719, Bahar Blvd., Hakimieh, Tehran 1658953571, Iran
Abstract: The main purpose of uncertainty analysis is deriving the model output statistical properties and/or its cumulative distribution function (CDF). In this paper, uncertainty and sensitivity analysis of a flood control design problem have been implemented in an urban area using Monte Carlo simulation via Latin hypercube sampling method as well as response surface methodology (RSM) for global sensitivity analysis (GSA). Probability theory along with evidence theory which is a more modern theory especially for epistemic uncertainty analysis has been used here. Several results have been achieved after uncertainty propagation such as the model output CDFs and quantiles. Also, a GSA has been implemented and sensitive parameters determined. Then using RSM, a regression model has been defined as surrogate of the main model. The training and testing of the surrogate model reveals its efficiency that means it's accurate enough for further analysis without any need to the main model.
Keywords: uncertainty analysis; global sensitivity analysis; GSA; aleatory; epistemic; evidence theory; flood control; Latin hypercube sampling; LHS; Monte Carlo simulation; MCS.
International Journal of Hydrology Science and Technology, 2022 Vol.13 No.4, pp.349 - 372
Received: 13 May 2019
Accepted: 02 Jan 2020
Published online: 01 Jun 2022 *