Flood modelling improvement using automatic calibration of two dimensional river software SRH-2D

Abstract

River model calibration is essential for reliable model prediction. The manual calibration method is laborious and time-consuming and requires expert knowledge. River engineering software is now equipped with more complex tools that require a high number of parameters as input, rendering the task of model calibration even more difficult. This paper presents the calibration tool O.P.P.S. (Optimisation Program for PEST and SRH-2D) and then uses it in multiple calibration scenarios. O.P.P.S. combines PEST, a calibration software and SRH-2D, a bi-dimensional hydraulic and sediment model for river systems, into an easy-to-use set of forms. O.P.P.S is designed to minimise the user’s interaction with the involved program to carry out rapid and functional calibration processes. PEST uses the Gauss–Marquardt–Levenberg algorithm to adjust the model’s parameters by minimising an objective function containing the differences between field observation and model-generated values. The tool is used to conduct multiple calibration series of the modelled Ha! Ha! river in Quebec, with varying information content in the observation fields. A sensitivity study is also conducted to assess the behaviour of the calibration process in the presence of erroneous or imprecise measurements.