Non-uniqueness in sediment transport in river network hydrology-sediment modelling

Sediment cascades are a convenient way of conceptualizing the transfer of sediment from hillslope production areas, through the river network, to the river basin outlet. Distributed hydrology-sediment models play an important role in the prediction of these source-to-sink links, because they can explicitly connect water and sediment fluxes along topographically-driven pathways. Here, we provide some examples of such cascade-based hydrology-sediment model applications in alpine environments and some problems related to their use.

In particular, we highlight two critical problems with hydrology-sediment modelling that go beyond trivial model calibration difficulties. These address fundamental issues of (a) non-uniqueness in sediment source mixing, and (b) sediment supply limitations. The first problem of non-uniqueness is known in hydrological modelling as the curse when models perform well at basin outlets for the wrong reasons, misrepresenting hydrological processes within the basin. In geomorphology, this concept has not received the same level of attention. Here we show that even a calibrated physically-distributed hydrology-sediment model can be subject to non-uniqueness, and provide the same suspended sediment yields at the basin outlet with completely different combinations of sediment sources. Including sediment tracers in model validation helps to identify this problem, and it is also helpful to check simulations at sub-basin scales where we are closer to distinct sediment sources. The second problem of sediment supply limitations is a challenge for all models that rely on transport capacity formulas for sediment transport. In our experience, both supply and transport capacity limit sediment transport at the basin scale, and we need to include this in our models. For example, we show that supply limitations can completely change the seasonality of sediment yields and render many climate change impact studies worthless.

Finally, we argue that both problems above, at least for suspended load, can be partially addressed by novel monitoring. For example by low cost smart sensors that allow a distributed sensing of sediment fluxes above and below potential sediment sources at high resolutions, or by high resolution remote sensing to capture space-time variability in river turbidity. This kind of data can dramatically improve our ability to calibrate models, reduce non-uniqueness, and over the long term identify the key signatures of sediment supply in river systems. It is our opinion that improving the predictions of climate and environmental change effects on sediment yields requires both better model validation as well as new data.