Being able to predict the depth and speed of water in a river channel is important for managing in-channel engineering, predicting sediment transport and flood risk, planning river restoration, prescribing minimum flows to preserve river habitats, and predicting carbon dioxide efflux. To make these predictions, we need to understand how the roughness of a river channel's bed and banks slows down the flow within it, i.e. the flow resistance of the channel. We commonly assume that sediment particles are the most important objects obstructing flow in river channels, and we represent their effect using the size of the larger particles relative to the flow depth . This assumption predicts flow resistance reasonably well in larger rivers where particles are small compared to the flow depth. But in headwater streams, which make up 77% of all river networks, flow is usually not much deeper than the largest bed particles. Even the best existing methods for predicting river speed and flow volume from depth, or depth and speed from flow volume, are very unreliable in these conditions, with predictions commonly being wrong by a factor of two. For comparison, predictions of how flow depths will change under climate change scenarios have about half this degree of uncertainty.

In this NERC funded project. led by Dr Rebecca Hodge at Durham University. we are using state of the art technologies for measuring river channel topography at high resolution in the field (terrestrial laser scanning, shallow-water multi-beam sonar) to produce the first comprehensive dataset of rough-bed river topographies, and will use statistical methods to describe the roughness of their beds and banks. We will select representative channels from this dataset, and replicate them in a laboratory flume by 3-D milling them at a reduced scale. In the flume we will sequentially add boulders, sediment and rough banks, and measure how each component affects flow depth and flow resistance. We will also use new numerical modelling methods to simulate flow properties in channels that we have manipulated so that they only contain certain topographic scales, thus allowing us to identify the most important sizes of obstacle. The combined flume and numerical modelling experiments will allow us to determine the physical basis for how different sizes and types of obstacles in a channel combine to set the total flow resistance. From this understanding we will produce new approaches for how best to predict flow speed, depth or volume. Overall, this project will provide a fundamental step change in understanding and prediction of flow in rivers.