From pixels to processes: linking automated mapping of rivers to meander- and bar morphodynamics

Although satellite imagery is increasingly used to study river mobility, two challenges remain: (1) fast, accurate tracking of channel banks and centerlines, and (2) linking these observations to predictive models and stratigraphic architecture. We developed ‘rivabar,’ a graph-based Python tool that automatically extracts centerlines and banklines. Using this tool on time-lapse Landsat and Planet Labs imagery, we visualize the evolution of single-thread and multi-thread rivers and investigate how well simple models predict channel migration.

In single-thread rivers, rates of accretion are more variable than rates of erosion. In rivers with low sediment loads, point-bar accretion is more uniform in time and closely matches erosion; curvature-based models predict migration rates well. The morphology and stratigraphy of the inner banks is relatively simple. In contrast, high sediment loads result in the rapid growth of bars that are accreted to the inner bank. The higher the sediment load, the more likely that mid-channel bars and bifurcations develop. Bank accretion is episodic and bar morphology and stratigraphy are complicated. Simple curvature-based models of meandering are less predictive.

Predicting channel movement in multi-thread rivers is even more challenging. However, preliminary results show that the upstream sides of mid-channel bars typically erode and migrate downstream, while the downstream sides grow irregularly. This mimics the asymmetry between erosion and accretion seen in single-thread channels.

These results suggest that sediment load is an important driver of stratigraphic complexity. While curvature-based models work for slowly migrating single-thread rivers, high-sediment-load and multi-thread systems require more sophisticated approaches. The ‘rivabar’ approach offers a scalable way to better understand channel mobility and stratigraphic complexity from widely available remote sensing data.