A Morphology Independent Methodology for Quantifying River Planform Change and Characteristics from Remotely Sensed Imagery

Abstract:

The ready availability of remotely sensed imagery offers the potential to examine river dynamics and planform characteristics at global scales. The Landsat archive currently offers the greatest spatial and temporal coverage of the entire globe. However, at 30 meter multispectral resolution detailed and accurate examination of planform changes using Landsat imagery is restricted to intermediate (~ 500 m wide) to very large (~ 1 km wide) rivers or smaller rivers with very high rates of change. Many of these larger river systems exhibit multi-threaded or braided channel patterns that present significant challenges for many of the existing methodologies for quantifying changes developed for single threaded meandering river systems. In order to examine planform changes in river systems across all scales and morphologies we developed a set of algorithms for quantifying river mobility and planform attributes using raster-based river masks extracted from remotely sensed data. Unlike many prior methodologies for measuring river migration and erosion that rely on changes in the position of river channel centerlines, our methods adopt river banks as a frame of reference for quantifying change. The choice of a bank-centric reference frame was motivated by both a primary interest in the spatial and temporal patterns of bank change and the significant challenge of extracting and comparing channel centerlines in multi-threaded systems. Unlike prior vector-based analysis of river channels, our analysis retains a raster-based representation of the river from the original imagery source. At each bank pixel, our algorithms compute linear rates of bank change, local channel width, bank curvature, and bank aspect (used for examination of the influence of thermal processes such as freeze thaw and permafrost influence). The spatially distributed measurements are also aggregated along equally spaced river segments to examine spatial patterns in erosion/accretion rates, and channel widths (both mean and effective in multi-threaded systems). Using our set of algorithms, we have successfully analyzed rivers varying in width from 10s of meters to greater than 1000 meters. Imagery sources have included aerial photography, high resolution satellite imagery, and coarser imagery such as SPOT, ASTER, and Landsat.