Detecting floodplain inundation based on the upstream–downstream relationship

Abstract

Summary The rise in river stage (water depth) can lead to disastrous floodplain inundation. On the basis of hydraulic simulation data, this study proposes novel data-analytical methods to infer the threshold river stage and detect floodplain inundation. A quasi-Muskingum model is derived from the classical Muskingum model to characterise the relationship between upstream and downstream river stages. Based on this model, F -test and modified Akaike information criterion AIC c are introduced to test if there is a change of the upstream–downstream relationship. Furthermore, a bootstrap-based calibration–validation experiment is set up to evaluate the performance of the quasi-Muskingum model. The proposed methods are applied to a case study of the 1991 and 2001 floods in the Flinders and Norman Rivers in Northern Australia. The results show that floodplain inundation does change the upstream–downstream relationship as it drastically alters the stage–discharge relationship. To combine the quasi-Muskingum model with F -test and AIC c facilitates an efficient approach to detect the change and infer the threshold river stage. The analytical testing is in concert with visual examination – the time when the river stage becomes higher than the detected threshold coincides with the beginning of floodplain inundation. Despite the change, the quasi-Muskingum model effectively captures the upstream–downstream relationship and requires a small number of samples in calibration. This study highlights the effectiveness of the data-analytical methods in dealing with the change of the upstream–downstream relationship.