Towards a Global Assessment of Bar Push Versus Bank Pull using Remote Sensing and Cloud Computing.

Rivers and floodplains are hotspots of biodiversity that support a large and growing number of people with food, water, nutrients, and transportation ¹. In these floodplains, processes of erosion on the outer-bank and sedimentation on the inner-bank drive lateral channel migration that can produce highly sinuous, intricate, meandering river landscapes. A key question in river meandering research concerns the debate about whether lateral river migration is driven initially by outer-bank erosion that induces local inner-bank deposition (bank pull), or whether inner sedimentation instead initially diverts the flow and subsequently forces outer-bank erosion (bar push). So far, studies exploring these mechanisms have been based on data from only a relatively few locations, producing divergent and inconclusive results ^2,3

Here we propose a methodology in which a long time series of remote sensing imagery is combined with cloud computing to identify the prevalence of bar push versus bank pull across large expanses of the global river network. In our methodology, each image is analysed to identify the timing of pixels undergoing erosion and/or sedimentation. This was achieved by combining the algorithm LandTrendr, which identifies change detection in a time series, with the Modified Normalized Difference Water Index (mNDWI) extracted from Landsat imagery during the period 1984-2020. For the Amazon region, we created a time series of mNDWI for the dry season (August to October) to have a higher availability of Landsat images and to detect the point bars that would otherwise be submerged during the high-water season. Then, we extracted and polygonized the river shore for different periods of the time series. For every river mask, we analysed the erosion years that intersect the outer-bank polyline and the sedimentation that intersect the inner-bank polyline. Then we compared the mean of the two distributions (Δmean_1990 = mean erosion – mean sedimentation) and used the Z-test to identify if they are distinctive. Positive values of Δmean discriminate episodes of ‘bar push’, while negative values discriminate ‘bank pull’. By repeating this analysis for different years and across multiple Landsat tiles, we were able to extract a distribution of Δmean and Z-test values for a large number of river bends.

Extending this methodology for different rivers along the Earth will enable us to test the bar push versus bank pull theories systematically on a wide range of real environments. This will help to identify whether rivers have a predominant mechanism or if they shift between both processes thoughout the years, as suggested by Mason and Mohrig ². This unprecedented scale of bar push versus bank pull analysis will improve our understanding of meandering rivers.     

 

1             Junk, W. J., et al. A classification of major natural habitats of Amazonian white-water river floodplains (várzeas). Wetlands Ecology and Management 20, 461–475 (2012).

2             Mason, J. & Mohrig, D. Differential bank migration and the maintenance of channel width in meandering river bends. Geology 47 (2019).

3             Van De Lageweg, et al. Bank pull or bar push: What drives scroll-bar formation in meandering rivers? Geology 42, 319-322 (2014).