An automatic graph-based method for characterizing multichannel networks

Liu, Y. and Carling, P.A. and Wang, Y. and Jiang, E. and Atkinson, P.M. (2022) An automatic graph-based method for characterizing multichannel networks. Computers and Geosciences, 166: 105180. ISSN 0098-3004

Text (RivMACNet_clean_manuscript_R3)
RivMACNet_clean_manuscript_R3.pdf - Accepted Version
Available under License Creative Commons Attribution-NonCommercial-NoDerivs.
Download (8MB)

Abstract

Assessment and quantitative description of river morphology using widely recognized river planview measures (e.g., length, width and sinuosity of channels, bifurcation angles and island shape) for multichannel rivers are regarded as fundamental parts of the toolkit of geomorphologists and river engineers. However, conventional assessment methods including field surveys or exiting algorithms for the extraction of multichannel planviews might be suboptimal. More recently, the potential for the application of complex network analysis to the study of river morphology has led to emphasis on the accurate characterization and definition of multichannel network topology. Therefore, we developed a novel algorithm called RivMACNet (River Morphological Analysis based on Complex Networks) that enables the extraction of multichannel network topology using satellite sensor images as the input. We applied RivMACNet to a meandering reach of the Yangtze River and a strongly anastomosing reach of the Indus River to construct their network topologies, and then calculated a series of common topological measures including weighted degree (WD), clustering coefficient (CC) and weighted characteristic path length (WCPL). The network analysis indicated that both networks exhibit poor transitivity with small clustering coefficients. The topological properties of the Indus at the reach scale are independent of flow conditions, while they vary across space at the subnetwork scale. In addition, comparison between RivMACNet and an alternative common river network analysis engine (RivaMap) demonstrated that RivMACNet is superior in terms of representation accuracy and network connectivity and, thus, is more suitable for multichannel fluvial systems with complex planviews. RivMACNet is, thus, a useful tool to support further investigation of multichannel river networks using graph theory.