Complexity in a cellular model of river avulsion

We propose a new model of river avulsion that emphasizes simplicity, self-organization, and unprogrammed behavior rather than detailed simulation. The model runs on a fixed cellular grid and tracks two elevations in each cell, a high elevation representing the channel (levee) top and a low one representing the channel bottom. The channel aggrades in place until a superelevation threshold for avulsion is met. After an avulsion is triggered a new flow path is selected by steepest descent based on the low values of elevation. Flow path depends sensitively on floodplain topography, particularly the presence of former abandoned channels. Several behavioral characteristics emerge consistently from this simple model: (1) a tendency of the active flow to switch among a small number of channel paths, which we term the active channel set, over extended periods, leading to clustering and formation of multistory sand bodies in the resulting deposits; (2) a tendency for avulsed channels to return to their previous paths, so that new channel length tends to be generated in relatively short segments; and (3) avulsion-related sediment storage and release, leading to pulsed sediment output even for constant input. Each of these behaviors is consistent with observations from depositional river systems. A single-valued threshold produces a wide variety of avulsion sizes and styles. Larger "nodal" avulsions are rarer because pre-existing floodplain topography acts to steer flow back to the active channel. Channel stacking pattern is very sensitive to floodplain deposition. This work highlights the need to develop models of floodplain evolution at large time and space scales to complement the improving models of river channel evolution.