Abstract:
In a river-centric deltaic country like Bangladesh, an optimal and efficient management of water resources is a challenging task. This is further exacerbated by the lacking coverage and availability of the observation data, such as water level, discharge, and bathymetry. Traditional in-situ observation platforms are costly to install and maintain, and the data is often publicly inaccessible. This does not meet water resources management requirements, where timely and reliable information with high spatial and temporal coverage is preferable. Therefore, this study aims to address these challenges by designing a multi-satellite monitoring system that provides comprehensive, timely, and accessible data on key water resources components i.e. water level, discharge, shoreline, and floodplain bathymetry—thereby enhancing spatial and temporal coverage to meet the critical needs of water resource management in Bangladesh.
The multi-satellite monitoring framework developed in this study integrates altimetry data from Sentinel-3A, Sentinel-3B, Jason-2/3, Synthetic Aperture Radar (SAR) data from Sentinel-1 and spectral data from Sentinel-2. A comprehensive, automated, near-real-time monitoring system is developed that can monitor water levels, discharge, shorelines, and floodplain digital elevation models (DEMs) and implemented over the Jamuna River. By leveraging the multi-satellite technique, the observation period of water level and discharge has been significantly improved from an average of 10 days with single satellites to 3.90 days and a continuous record has been developed for 2008 to 2022. Using SAR and multi-spectral imageries from Sentinel-1 and Sentinel-2 respectively, an automated shoreline extraction technique is applied to monitor the river shoreline and morphology in quasi near real-time. Such continuous monitoring provides an annual floodplain DEM, as well as a yearly assessment of river erosion-accretion.
The dataset generated in this study has been validated based on in-situ observations at BWDB stations i.e. Bahadurabad (SW46.9L) and Kazipur (SW49A) showing good performances in monitoring water level and discharges. Multi-satellite water level monitoring shows an RMSE, MAE, average percent error and correlation coefficient values of 0.53 m, 0.36 m, 2.31% and 0.98 respectively while discharge estimations exhibited RMSE, MAE, average percent error and correlation coefficient of 4,738 m³/s, 2,997 m³/s, 23.69% and 0.95 respectively at Bahadurabad station. Analysis of the DEM reveals that the floodplains along the Jamuna river have been substantially changed over the last ten years, and the global bathymetry product does not reflect the current river topography (particularly the Chars). In addition, the erosion zone identified in the process matches well with secondary newspaper-based field conditions. By leveraging the improved temporal frequency provided by the multi-satellite data, the water level and discharge datasets developed in this study were able to effectively capture flood events. Additionally, continuous monitoring of the shoreline and floodplain using high-resolution satellite imagery provided highly detailed information on flood extents, enabling precise mapping and comprehensive assessment of the affected areas.
This study concludes that multi-satellite remote sensing technique presented here can not only be a reliable operational tool for water resources management agencies in Bangladesh, but also can provide valuable and publicly accessible dataset to the broad water resource research community. Such up-to-date and accessible dataset should be useful for further advancing research on floods, morphological evolution, as well as for more informed decision-making.