Spatio-temporal changes of river-aquifer interactions in North-central Bangladesh

Bangladesh, one of the most densely populated countries in the world, relies primarily on its irrigated agriculture to sustain rice production for a population that rose from 90 million in 1981 to about 165 million in 2022. Consequently, over the past three decades, irrigated land in Bangladesh has expanded from 2.58 million ha in 1990 to approximately 5.63 million ha in 2020, increasing the share of irrigated land from 31% to 66%. This rapid expansion has relied heavily on shallow groundwater abstraction, which, together with increasing domestic and industrial abstraction, has led to declining groundwater levels, altering the river-aquifer exchange in the Bengal Delta. Previous studies have estimated the overall change in the total groundwater recharge associated with increased abstraction through freshwater capture (“Bengal Water Machine”). However, these studies have not quantified the change in the focused recharge from river leakage into the groundwater system. To address this issue, we have developed a numerical groundwater flow model to assess the impact of increasing abstraction over the past four decades on the river-aquifer interactions in the North-central Bangladesh.

The 3D unstructured numerical model domain (area: 27670 km²) is delimited by the Shillong Plateau (i.e., Precambrian basement) to the north, and by the Brahmaputra (locally known as Jamuna), Meghna, and Ganges (locally known as Padma) rivers, to the west, east and south, respectively. The groundwater flow model was implemented in MODFLOW-6 and was set up using Flopy environment. The flow model consists of 11 layers (370 m average thickness) based on a regional geological model developed from borehole lithological data, and reflecting the multilayer aquifer distribution described in the literature. Main river networks within the study area, direct (diffuse) recharge from effective precipitation, and abstraction for domestic, irrigation, and industrial purposes were considered as boundary conditions. River-aquifer exchanges were simulated using the RIV package, driven by long-term monthly stage observations at several fluviometric stations within the domain. Hydraulic properties including hydraulic conductivities, vertical anisotropies, and specific storage, along with the riverbed conductance were calibrated using PEST++, based on long-term groundwater levels and drawdowns between 1980 and 2018 in over 80 observation wells located within the model domain.

The preliminary results show that the steady increase in groundwater abstraction for irrigation, especially using wells located in the shallow aquifer, has reversed the direction of flow in most rivers compared to the pre-irrigation (i.e., natural) condition, changing the hydrological system from gaining to losing regime. The magnitude of these changes is subject to high uncertainty, due to the intrinsic heterogeneity of the aquifer system and to the conductances in the riverbed, the morphology of the channels and other factors, especially during the monsoon (wet season).