Abstract
Most ways of predicting flow resistance in shallow rivers with a partial or complete cover of coarse sediment use a bed‐sediment grain diameter as a roughness length scale. However, beds with the same grain size distribution differ in roughness and flow resistance depending on how the larger grains are arranged, the nature of any bedforms and the possible complications of bedrock or rough banks. This has led to interest in predicting flow resistance using metrics of the topographic roughness of the bed. Some researchers have used the standard deviation of bed elevation as a roughness length scale. An alternative for channels containing boulders is to regard the bed as an array of large roughness elements. Fluvial research to date using these two approaches is limited and inconclusive. We review potentially relevant findings from the much more extensive literature in boundary‐layer meteorology and various branches of engineering and note links between the distribution‐statistics and element‐array approaches. The skewness of the elevation distribution is widely seen as important but it is unclear how best to use it for flow prediction. Other open questions include the scale dependence of topographic metrics, and what type of flow resistance equation to use them in. Calibration and testing of new prediction methods require flow data from reaches with known roughness statistics. This need should be met partly by measurements at field sites or in flume models of them, but also by flume experiments and numerical simulations using synthetic roughness.