A Global Analysis and Comparison of River Delta Channel Kinematics Using Satellite Imagery

While some river deltas are relatively stable, the channel networks of other river deltas are actively changing in response to internal and external forcings. As channels widen, narrow, laterally migrate, meander, and avulse through time, the distribution of water and sediment is changed within the channel network. While the classic definitions of river deltas as being tidally, fluvially, or wave dominated help us predict the morphologic activity of deltas over millennial time scales, quantifying this current channel activity of a delta over decadal time scales is important to aid in protecting the human and ecological presences currently living on these systems. In this study, we classify a subset of the world's deltas according to how dynamic their channel networks are over decadal time scales relevant in short term modeling and decision-making processes. For each delta, channel presence is extracted from historic Landsat imagery using DeepWaterMap, a fully convolutional neural network deep learning tool capable of seeing through clouds and discerning water from land, snow, ice, shadows, and clouds [Isikdogan et. al, 2017, 2019]. Channel centerlines are extracted from the DeepWaterMap water maps and the migration rates for all of the channels in each delta network are determined using techniques from particle image velocimetry [Thielicke and Stamhuis, 2014]. The mechanism by which each channel is migrating is determined by calculating the variance of the time series of water map images and observing signatures of each of the main types of channel movement [Jarriel et. al, 2019, 2020]. Together, these techniques allow us to quickly and accurately determine both the distribution of migration rates and the main mechanisms by which the channels are moving in each delta system. Understanding how relatively active a river delta has been in recent decades aids in river delta related decision making and modeling efforts.