A river is usually not a pixel wide

Posted on

In digital terrain analysis and landscape evolution modelling, we usually represent flow of water and sediment in a very simplified way. Most commonly, we use the D8 algorithm which determines the one steepest neighboring pixel to which all water is passed. This means that flow converges and never diverges, which is at odds with actual river flow. We may use other flow routing techniques such as Dinf or other multiple flow direction algorithms but also these do actually not account for how much water there is. The flow patterns are independent of the flow rate. In a new study spearheaded by Boris Gailleton, we develop a novel technique termed GraphFlood that leverages common flow routing techniques to obtain realistic flow patterns. The study is published as a preprint at EGUsphere.

Comparison of single flow directions (right panel) and GraphFlood (left panel). Source: Gailleton et al. (2024).

The approach is rather simple … at first. We repeatedly determine flow directions and flow accumulations and accordingly elevate pixels by some water amount. At the same time, we calculate for each pixel Manning’s equation which gives us the flow rate and thus the rate at which water moves out of the pixel. After several iterations, flow depth will attain a steady state. Boris put in some tremendous effort to do these calculations as efficiently as possible in C++ and the code works on moderately to really large sized grids. There’s just one minor drawback. The software is currently not yet available for integration in TopoToolbox. Yet, we are working on it. Until then, you can give it a try here.

GraphFlood is just the beginning. There are numerous opportunities to include the hydrodynamic solver into landscape evolution models. Like in EROS (or RiverLab), this will allow running landscape evolution models with rivers whose width is not explicitly given, but evolves in response to geomorphic processes.

Reference

Gailleton, B., Steer, P., Davy, P., Schwanghart, W., Bernard, T.G.A. (2024): GraphFlood 1.0: an efficient algorithm to approximate 2D hydrodynamics for Landscape Evolution Models. EGUsphere [preprint]. [DOI: 10.5194/egusphere-2024-1239]

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.