Overbank flow in the river Allier

Abstract

For several years the Department of Physical Geography ofthe University of Utrecht has conducted surveys on the river Allier in France. These surveys always took place during periods of low discharge because at high or even moderate discharges measurements are impossible. As information on the flow during a flood is important to understand the river morphology, a flow model of a part of the Allier was made to simulate the flow during a flood. During a survey in the summer of 1998 bathymetric data and flow measurements were collected. With this data a flow model was made and calibrated. The discharge during the survey was approximately 20 m3Is. During the calibration it became clear that the downstream boundary condition (a water level) could not be generated well. This problem was overcome by moving the boundary to a flow measurement section where the water level for a discharge of20 m3/s was known. However this left a problem for simulation at higher discharges than 20 m3/s. The influence ofan error in the downstream boundary condition was estimated both numerically as well as with the Bresse approximation. Both methods showed the backwater effect introduced by an error to extent for about 1000 m upstream of the boundary. The magnitude of a water level error however, was shown to decrease rapidly in the upstream direction. To simulate flow during a flood several simulations were made, steady- (with a constant discharge) and unsteady state (with a varying discharge). Ten steady state simulations were made, increasing in discharge from 100 to 1000 m3/s. In the unsteady state run the flood of November 1994 was simulated. The simulations showed the flow mainly to follow the main channel, leading to an inbank flow pattern. The position of the secondary flow cells - where the bend radius of curvature is smallest - also indicated an inbank flow pattern. Velocities up to 4 m/s were found in the main channel leading to very large bed shear stresses. At several places the flow was directed onto the point bars. The bed shear stress magnitude here indicated that large grain sizes could be transported onto the point bars. The differences between the steady state and the unsteady state simulations were small. Although there were some differences the flow pattern and the magnitude of the velocity were the same. This means that for a global impression ofthe flow pattern at a certain discharge, a steady state simulation is sufficient. This saves a lot of computation time as the unsteady state simulation has a much larger computation time. Armour layers are layers 0 f coarse grains on top of the bed. They were found at several places in the survey area. During the survey a number of the armour layers were sampled. With the aid of the Oak Creek model by Parker (1990) the threshold of motion ofthe grain sizes within these armour layers was estimated. By combining the Oak Creek model and the bed shear stresses from the flow model it was shown that the threshold of motion was exceeded for all grain sizes within the sampled armour layers. Also a rough indication of the surface grain size distribution was given based on the Oak Creek model and the bed shear stresses derived from the flow model. However, the applicability of Oak Creek model to the river Allier was not tested. This requires sediment transport measurements. For the various coefficients in the Oak Creek model the literature values were used.