Network concepts to describe channel importance and change in multichannel systems: Test results for the Jamuna River, Bangladesh

Most of the largest rivers on Earth have multiple active channels connected at bifurcations and confluences. At present a method to describe a channel network pattern and changes in the network beyond the simplistic braiding index is unavailable. Our objectives are to test a network approach to understand the character, stability and evolution of a multi-channel river pattern under natural discharge conditions. We developed a semi-automatic method to derive a chain-like directional network from images that represent the multi-channel river and to connect individual network elements through time. The Jamuna River was taken as an example with a series of Landsat TM and ETM+ images taken at irregular intervals between 1999 and 2004. We quantified the overall importance of individual channels in the entire network using a centrality property. Centrality showed that three reaches can be distinguished along the Jamuna with a different network character: the middle reach has dominantly one important channel, while upstream and downstream there are about two important channels. Temporally, relatively few channels changed dramatically in both low-flow and high-flow periods despite the increase of braiding index during a flood. Based on the centrality we calculated a weighted braiding index that represents the number of important channels in the network, which is about two in the Jamuna River and which is larger immediately after floods. We conclude that the network measure centrality provides a novel characterization of river channel networks, highlighting properties and tendencies that have morphological significance.