Abstract:
The Brahmaputra-Jamuna River is renowned for its enormous flow and sediment load making its channels very dynamic and unstable. This pattern, which causes channel shifting and erosion of riverbanks, is noticeable every year. During the monsoon, new erosion sites are discovered in vulnerable places, and even the bank protection structures require routine maintenance. The locals that live along the Brahmaputra-Jamuna river's banksides and chars are impacted by this. Because of riverbank erosion, thousands of people in Bangladesh are currently homeless and living in abject poverty. So, a better system for predicting river bank erosion might be one way to mitigate the impact of such a disaster. In physics-based river bank erosion high resolution bathymetry data is needed. Bangladesh Water Development Board (BWDB) gathers river bathymetry at intervals of 4 to 6 km which seems too course for capturing the dynamics of the braided river. Regular mapping of braided river bathymetry is expensive and time-consuming worldwide. Remote sensing can be used to close this gap. Therefore, this study's primary goal is to create braided river bathymetry based on satellite images so that riverbank and char erosion can be predicted. Here, bathymetry generated from Sentinel images utilized in a 2D morpho-dynamic model (Delft3D model) estimate erosion. The created approach was verified for the years 2017 and 2018, and it was calibrated for 2019. For the Brahmaputra-Jamuna, bathymetry derived from satellite images indicates that, among 2017 ,2018 and 2019, the maximum depth was 17.5, 16, 19.5 m, while the average depth was 10, 9, 10 m. However, the derivation of bathymetry from SAR images is insufficient. The simulated model's R2 value was approximately 0.96 in the case of a discharge at Bahadurabad; in contrast, the values are 0.99, 0.97, and.99 for water level stations in Mathura, Serajgani, and Chilmari. The NSE, PB, and RRMSE for the stations in Mathura, Bahadurabad, Serajgani, and Chilmari show the level of satisfaction. The erosion prediction accuracy was expressed by kappa statics and throughout the three years from 2017 to 2019 those were around 67%, 78%, and 72%. Determining bathymetry using such optical images seems reliable to use as an input parameter of numerical model for river erosion estimation in data scarce region.