Marine and River Dune Dynamics - 1-3 April 2008 - Leeds, United Kingdom 241 1 INTRODUCTION The greatest barrier to using computational models for predicting flood inundation, sediment transport, and large-scale morphologic change in rivers is the presence of bedforms and their complex adjustment to temporally varying flow fields. Even the very best computational flow models in rivers rely heavily on accurate specification of local roughness for making correct predictions, and local roughness is frequently dominated by form drag on bedforms that vary sig- nificantly in space and time. Typically, this rough- ness is parameterized using empirical approaches which have been developed based on constant dis- charges; the evolution of bedforms and the drag they generate is poorly understood in flows with strong temporal variability in flow discharge. Improving the current capability for predicting flow and sediment transport in natural rivers with varying hydrographs requires an improved capability for predicting the spatially distributed adjustment of the bedform fields to variations in discharge. Over the past two decades, improvements in the methods for measuring and visualizing flows over bedforms have driven data collection efforts in the laboratory and in field situations; these measure- ments have greatly improved understanding of those flows. This work has clarified certain aspects of the basic bedform instability mechanism, and has also provided accurate basic data on mean flow and tur- bulence fields for a variety of flow conditions and bedform shapes and sizes for testing computational flow models. Currently, there are several computa- tional codes that have been shown to yield reason- able predictions of the complex flow structure over bedforms using both direct-numerical simulation techniques and turbulence closure methods. There are also coupled models in which tested flow models have been combined with sediment-transport models in order to investigate the initiation and morphologic evolution of bedforms, notably the recently- Bedform Response to Flow Variability J.M. Nelson, B.L. Logan & P.J. Kinzel US Geological Survey, Golden, CO, USA Y. Shimizu University of Hokkaido, Sapporo, Japan S. Giri Delft Hydraulics, Delft, The Netherlands R.L. Shreve University of Washington, Seattle, WA, USA S.R. McLean University of California, Santa Barbara, CA, USA ABSTRACT: Laboratory observations and computational results for the response of bedform fields to rapid variations in discharge are compared and discussed. The simple case considered here begins with a relatively low discharge over a flat bed on which bedforms are initiated, followed by a short period with double the original discharge during which the morphology of the bedforms adjusts, followed in turn by a relatively long period of the original low discharge. For the grain size and hydraulic conditions selected, the Froude number remains subcritical during the experiment, and sediment moves predominantly as bedload. Observations show rapid development of quasi-two-dimensional bedforms during the initial period of low flow with increasing wavelength and height over the initial low-flow period. When the flow increases, the bedforms rapidly in- crease in wavelength and height, as expected from other empirical results. When the flow decreases back to the original discharge, the height of the bedforms decreases in response rapidly, but the wavelength decreases much more slowly. Computational results for the same conditions simulate the formation and initial growth of the bedforms fairly accurately, and also predict an increase in dimensions during the high-flow period. However, the computational model predicts a much slower rate of wavelength increase, and performs less ac- curately during the final low-flow period, where the wavelength remains essentially constant, rather than de- creasing. In addition, the numerical results show less variability in bedform wavelength and height than the measured values. Based on observations, these discrepancies may result from the simplified model for sedi- ment particle step lengths used in the computational approach. Assuming a constant value for the step length neglects the role of flow alterations in the bedload sediment-transport process, which appears to result in pre- dicted bedform wavelength changes smaller than those observed.