Experimental and numerical model studies of dike-break induced flood processes over a typical floodplain domain

Farm dikes are of crucial importance for the flood defense system of a floodplain domain in the Lower Yellow River, with devastating losses caused by failure of dikes in record. Detailed investigations are therefore urgently required of dike-break induced flood inundation processes over the floodplain domain. In this study, a laboratory model was designed to reflect the different topographical characteristics of the channel-floodplain system, and a series of laboratory experiments were conducted to investigate the flood inundation processes over the floodplain domain before and after a sudden break of a segment of farm dike. Key water level hydrographs were measured at several points to illustrate the hydrodynamic characteristics of dike-break induced flooding. Furthermore, a two-dimensional (2D) hydrodynamic model with the solution of finite volume method was also developed to reproduce the flood inundation processes and investigate the influences of different bed roughness coefficients in the floodplain domain. Results from the measurements and the model predictions show that: (i) the dike-break induced flood wave developed from an asymmetric ellipse at the breach site, and then rapidly propagated toward the downstream floodplain domain after the impingement against the side wall due to a large transverse bed slope. The floodplain domain downstream of the breach site was the most vulnerable zone during a sudden dike-break induced flood event; (ii) the model predictions were in close agreement with the measured results, with the Nash–Sutcliffe efficiency values of water level close to 1.0 at most measuring points distributed in the channel and floodplain domains. The highest root mean square error value of water level at the measuring point was smaller than one tenth of the mean water level; and (iii) the flood propagation time and peak velocity in the floodplain domain were obviously influenced by different bed roughness coefficients. When the comprehensive roughness coefficient increased from 0.015 to 0.045 under the incoming discharge of 38 L/s, the flood propagation time extended by at least 8.0 s and the peak velocity decreased by at least 50% at the measuring points in the floodplain domain. Therefore, it is necessary to determine a suitable value for the comprehensive floodplain roughness coefficient when modeling the dike-break floods.