Predicting river channel pattern based on stream power, bed material and bank strength

Abstract

Rivers exhibit a wide variety of channel patterns, and predicting changes in channel pattern is important in order to foresee river responses to climate change and river restoration. Many discriminators have been developed to define approximate boundary conditions for different channel patterns, based on channel-pattern-controlling parameters such as discharge and valley gradient. However, presently available discriminators have two main shortcomings. First, they perform poorly for rivers with cohesive, relatively erosion-resistant banks. For this subset, discriminators tend to indicate an actively meandering channel pattern, whereas the river morphology and dynamics show that many of these rivers should be classified as laterally stable. Second, channel pattern discriminators are often used to predict channel patterns, which is only valid when parameters are used that are independent of actual channel pattern. This condition is often not met, as many discriminators use the channel slope or width–depth ratio of the channel as input. To resolve both shortcomings, we first propose an additional class of rivers with scroll bars and tortuous channel patterns, which have an inhibited mobility due to their self-formed cohesive deposits. Second, we compare frequently used empirical and mechanistic channel pattern discriminators, taking into account the success in predicting channel pattern and the independence of causal factors used. Thirdly, we present a novel channel pattern discriminator and predictor that includes the effect of a cohesive floodplain, using the average silt-plus-clay fraction of the river banks as proxy. We show that this new predictor outperforms previously used empirical and mechanistic approaches, and successfully predicts channel pattern for 87% of the rivers from a dataset of 70. This new predictor is widely applicable, as it is relatively simple and based on easily obtainable, and mostly independent, parameters.