Abstract:
Ganges-Brahmaputra-Meghna (GBM) river basins form one of the largest and most populous deltas in the world. These three rivers cover an area of approximately 1.7 million square kilometers including Bangladesh, India, Nepal, Bhutan and China. During the flood period, particularly in the monsoon season, over 138,700 m3/s of water flows into the Bay of Bengal through a single outlet of the GBM river basins. In recent years unpredictable precipitation and extreme events due to climate change have affected the entire basins areas. This situation poses threats to water as a means of survival for about 670 million people. Monitoring water budgets is a primary requirement of effective and sustainable river management. Therefore, it is crucial to perceive an efficient way to monitor the water budget for the entire GBM basins region for distribution and prediction among the sharing countries.
In this context, the study aims to assess the seasonal runoff from the water budget components for GBM river basins management. The study meticulously investigates hydrological patterns in the Ganges-Brahmaputra-Meghna (GBM) river basins from 2003 to 2021, utilizing advanced data from IMERG, GRACE, MODIS, and GLDAS. The analysis reveals pronounced seasonal rainfall patterns, with the wet season showing significantly higher precipitation, particularly in the Meghna basin, which averages 2000 mm between June and October, nearly double that of the Ganges, Ghaghara, Kosi, Son, and Yamuna sub-basins. Over the past two decades, a steady decline in precipitation has been observed across the eastern sub-basins, including the Brahmaputra, Meghna, and Kosi, alongside notable discrepancies between remote sensing data and GLDAS model outputs. These discrepancies, particularly in the Meghna, Kosi, and Son basins, exceed 9.87%, likely due to the coarser spatial resolution of the GLDAS model. Water storage insights from GRACE data highlight a seasonal increase of approximately 400 mm during the wet periods, contrasted with decreases in the dry season. The Meghna sub-basin exhibits the highest evapotranspiration rates during the dry season, reaching nearly 300 mm, while other basins show less than 2% seasonal fluctuation, maintaining steady averages throughout the study period. Runoff volume analysis demonstrates significant seasonal shifts, with the Brahmaputra's dry season runoff peaking at nearly 175 billion cubic meters (BCM) in 2004, according to RS data, compared to a lower estimate of 75 BCM from GLDAS in 2003. The wet season sees dramatic increases, especially in the Brahmaputra, where RS data recorded peaks above 800 BCM, underscoring the monsoon's impact and the differing measurement capabilities of RS and GLDAS. Additionally, the study evaluates four regression models for runoff forecasting, finding that the Linear Regression model explains approximately 92.76% of variability with a reasonable margin of error. The Decision Tree and Random Forest models show a slight decrease in explanatory capability, while the K-Nearest Neighbors (K-NN) models excel with R² values around 94.79%, indicating superior prediction accuracy and minimal error rates.