GC21D-0572:
Spatiotemporal Dynamics of River Channel Migration on the Ganges-Brahmaputra Delta: 2000-2013

Tuesday, 16 December 2014
Christopher Small1, Soyee Chiu2, Daniel Sousa2, Dhiman Ranjan Mondal3, Michael S Steckler4, Syed Humayun Akhter5, Bodrudoza Mia5, Steven Lee Goodbred Jr6, Carol Wilson7 and Leonardo Seeber4, (1)Lamont Doherty Earth Obs., Palisades, NY, United States, (2)Lamont -Doherty Earth Observatory, Palisades, NY, United States, (3)CUNY Graduate School and University Center, Earth and Environmental Sciences, New York, NY, United States, (4)Lamont-Doherty Earth Obs, Palisades, NY, United States, (5)Dhaka University, Dhaka, Bangladesh, (6)Vanderbilt-Earth & Envir Scies, Nashville, TN, United States, (7)Vanderbilt University, Dept Earth and Environmental Sciences, Nashville, TN, United States
Abstract:
We use multitemporal multiscale satellite remote sensing to complement field observations and subsurface measurements to better understand the relationship between recent and historic fluvial dynamics on the Ganges-Brahmaputra (GB) delta. To provide regional context for the interannual changes in river channel geometry we conduct spatiotemporal (ST) analyses of MODIS Enhanced Vegetation Index (EVI) imagery for 2000-2013 using a new method of Empirical Orthogonal Function (EOF) analysis. We use EVI because it distinguishes water from wet and dry sediment on the basis of the spectral slope at VNIR wavelengths. Water has a negative slope while dry sediment has a small positive slope and vegetation has a large positive slope. To characterize the ST patterns associated with river channel migration we use iterative EOF analysis (iEOF). In iEOF we first conduct a single year EOF analysis for each year in the time series to identify the primary spatial principal component (PC1) for each year and separate this from the spatial structure of the subannual temporal patterns associated with vegetation phenology. We then construct a decadal time series of PC 1 for each single year and conduct a second EOF analysis of the time series of 13 individual year PCs. The standard EOFs of the full (312 images x 16 day) time series only resolve a decadal trend (EOF 8), but the iEOF clearly distinguishs the progressive decadal trend (EOF 2) from the cyclic component (EOF 3) of decadal changes in sediment reflectance. The temporal feature space constructed from PC 2 and PC 3 (corresponding to temporal EOFs 2 and 3) distinguishes pixels with progressive decadal increases and decreases in reflectance from pixels with cyclic changes. Evolution of the annual structure is animated at www.youtube.com/watch?v=UM1UYvdnYXk Despite significant differences in the 2 rivers’morphologies, and the considerable magnitude of flooding every year, we observe year-to-year continuity in the progressive downstream migration of the larger meanders on both rivers. We also observe a pronounced change in the morphology and evolution of the channels in the Brahmaputra braid plain upstream and downstream of the hinge zone with more rapid changes in channel geometry upstream and a more stable configuration of two anastamosing primary channels downstream.