Simultaneous river width and river topology extraction by remote sensing

Abstract:

A growing amount of satellite and aerial imagery offers the increasing opportunity to extract river planform morphology, particularly at large scales. Among many quantities of river planform morphology, the river width and network topology are of high interest because their scaling relationship is the basis for estimating river system behaviors, such as river extent, sediment, and carbon emission. To develop a more direct measurement of the river hydromorphology by remote sensing, this study proposes a new method that simultaneously extracts river width and river topology based on satellite imagery. The Multispectral Water Index (MuWI) is used to produces high-accuracy water maps from imagery, whereby specialized algorithms are devised to reduce impacts from non-river water (e.g., lake, reservoir, wetland) and on-channel objects (e.g., bridge, dam, ship), and to enforce the river connectivity. The connected river map is skeletonized into the graph while maintaining geo-references as properties of edges and nodes in the river graph, followed by the river graph pruning to remove false and redundant river tributary in the topologic structure. The cross-sectional measure is conducted on the river-reach (graph edge) basis, where orthogonal to centerlines is determined by the bounding geometry. The proposed method also implements the whole-process automation from image preparation to river width calculation. In our experiments, the results are consistent with the reference river widths (R²=0.98 in the mean-width validation and R²=0.91 in the transient-width validation). Despite the promises of the simultaneous extraction, random uncertainty in water mapping is identified as the major source of measurement errors. The results of the extraction could be applicable in fluvial analysis and satellite-derived discharge estimates.