Abstract: Floods are natural events that cause significant morphological changes in riverbeds and adjacent lands. Therefore, predicting the process of river changes under overbank flows is one of the important issues in river engineering. Conducting laboratory studies, in addition to time, requires high cost and accuracy, so using a numerical and analytical model is more reliable and requires less cost due to providing accurate results. In this study, the behavior and hydraulic geometry of the trapezoidal compound channel have been investigated using MIKE 21 software for the type of bankfull and overbank flow. First, the compound trapezoidal channel section with erodible boundaries was developed under bankfull flow, and then the behavior and geometry of the developed trapezoidal channel were evaluated under overbank flows. The results showed that when the cross-section of the channel is full of water, there was no significant change in the width of the main channel, but the width of the main channel in overbank flows is higher than the full section and increases rapidly. The mean relative differences obtained at the upstream and downstream width of the trapezoidal section and flow depth are 0.96, 0.88, and 0.98, respectively. The width, depth, and sediment concentration values predicted by the above software are better estimated than those of stabable channel design methods such as Brownlie, Van Rijn, and WBP.

Introduction: Floods are one of the natural events that cause significant morphological changes in riverbeds and adjacent lands, which result in financial, human, and other losses. A flood occurs when the water level rises above the main section of the river, and the water enters the flood plains. In such a situation, the section of the river or channel is called a compound section. When a flood occurs, the interference of the flow in the main channel and the flood plain in the compound channel causes momentum transmission and complexity in the flow pattern. Due to the dynamic nature of these events, statistical, experimental, and semi-experimental criteria are used to determine the dimensions of stable regime channels. Nowadays, the use of two-dimensional models to study the behavior of the regime channels has a superior advantage over the empirical and theoretical approaches, which are based on one-dimensional. Despite the redundancy of experimental studies, there have been only a few numerical studies on the flow pattern and sediment transport in alluvial channels. In this study, the behavior and hydraulic geometry of the trapezoidal compound channel have been investigated using MIKE 21 software for the type of bankfull and overbank flow.

Methodology: In this study, the flow in a rectangular channel was simulated using MIKE 21 software. This channel has been studied in a laboratory at the University of Newcastle, England. It consists of two floodplains and a main channel (river model) and has an erodible bed, ‎Fig. 2. The channel is 2.5 meters wide and 22 meters long, of which 18 meters are filled with sand to a depth of 0.6 meters. The median diameter of its particles is 0.1 mm≈D50 (Haidera, 2002).
To study the behavior of the main channel during the flood flow, the flow is adjusted in such a way that the flood zones are on the threshold of water intake. This work led to measuring the intensity of sediment transfer only in the main channel during the flood, and erosion will not happen in the flood zones. ‎Table 1 shows the basic information of modeled channels of bankfull flow predicted by the WBP method (Haidera, 2002).
The hydrodynamic module of MIKE 21 software was used for simulation. This module is used in a wide range of hydraulics and related phenomena. It uses the obtained output results as input for other MIKE 21 modules, such as the sediment transport-spreading module (MIKE 21 User's Manual, 2014). ‎Figs. 3 and ‎4 show a view of the 2D channel model created in Mesh Generator and the mesh applied, respectively.
To evaluate the results obtained from the MIKE 21 model with laboratory data, four statistical methods, including Correlation Coefficient (R²), Root Mean Square Error (RMSE), Relative Difference (DR), and Average Absolute Value Of Relative Errors (MARE), have been used.

Results and Discussion: The comparison of the obtained results shows that among the three sediment transport equations of England and Hansen, Van Rijn, and Meer-Pieter Muller that are available in the MIKE 21 software, the England-Hansen sediment transport equation estimates the channel dimension parameters with the highest correlation coefficient of 0.95.
Among the six examined channels (‎Table 1), three channels (‎Table 2) were used for verification, and the others (‎Table 4) were used for validation. The results reveal that the model can estimate the geometric parameters of the channel well in both stages (‎Tables 3 and ‎5).
When the cross-section of the canal is full of water (first state), there is no significant change in the main channel width, but in overbank currents, the width of the main canal is bigger than it is in the first state and increases rapidly.
Average cross-section changes at the distances of 6, 11, and 13 meters from the beginning of the channel for the developed channel (‎Fig. 7) show that during the flood flow, the channel cross-section estimated by the model is close to the observational cross-section. Changes in sediment and cross-sectional concentrations in the laboratory model after the first 6 hours of the water flow are negligible and have reached stable conditions, while in the mathematical model, after 3 hours, the channel's width rate decreases and eventually reaches stability. This indicates more changes by simulation.

Conclusion: A stable alluvial channel with a bankfull, when subjected to flooding, tilts towards a new stable channel. Also, the width, depth, and sediment concentration values predicted by the above software are better estimated than those of stabable channel design methods such as Brownlie, Van Rijn, and WBP.