Constraining bedrock erosion during extreme flood events

The importance of high-magnitude, short-lived flood events in controlling the evolution of bedrock landscapes is not well understood. During such events, erosion processes can shift from one regime to another upon the passing of thresholds, resulting in abrupt landscape changes that can have a long lasting legacy on landscape morphology. Geomorphological mapping and topographic analysis document the evidence for, and impact of, extreme flood events within the Jökulsárgljúfur canyon (North-East Iceland). Surface exposure dating using cosmogenic 3He of fluvially sculpted bedrock surfaces determines the timing of the floods that eroded the canyon and helps constrain the mechanisms of bedrock erosion during these events. Once a threshold flow depth has been exceeded, the dominant erosion mechanism becomes the toppling and transportation of basalt lava columns and erosion occurs through the upstream migration of knickpoints. Surface exposure ages allow identification of three periods of rapid canyon cutting during erosive flood events about 9, 5 and 2 ka ago, when multiple active knickpoints retreated large distances (> 2 km), each leading to catastrophic landscape change within the canyon. A single flood event ~9 ka ago formed, and then abandoned, Ásbyrgi canyon, eroding 0.14 km3 of rock. Flood events ~5 and ~2 ka ago eroded the upper 5 km of the Jökulsárgljúfur canyon through the upstream migration of vertical knickpoints such as Selfoss, Dettifoss and Hafragilsfoss. Despite sustained high discharge of sediment-rich glacial meltwater (ranging from 100 to 500 m3 s-1); there is no evidence for a transition to an abrasion-dominated erosion regime since the last erosive flood: the vertical knickpoints have not diffused over time and there is no evidence of incision into the canyon floor. The erosive signature of the extreme events is maintained in this landscape due to the nature of the bedrock, the discharge of the river, large knickpoints and associated plunge pools. The influence of these controls on the dynamics of knickpoint migration and morphology are explored using an experimental study. The retreat rate of knickpoints is independent of both mean discharge, and temporal variability in the hydrograph. The dominant control on knickpoint retreat is the knickpoint form which is set by the ratio of channel flow depth to knickpoint height. Where the knickpoint height is five times greater than the flow depth, the knickpoints developed undercutting plunge pools, accelerating the removal of material from the knickpoint base and the overall retreat rate. Smaller knickpoints relative to the flow depth were more likely to diffuse from a vertical step into a steepened reach or completely as the knickpoint retreated up the channel. These experiments challenge the established assumption in models of landscape evolution that a simple relationship exists between knickpoint retreat and discharge/drainage area. In order to fully understand how bedrock channels, and thus landscapes, respond and recover to transient forcing, further detailed study of the mechanics of erosion processes at knickpoints is required.