Delineating Variabilities of Groundwater Level Prediction Across the Agriculturally Intensive Transboundary Aquifers of South Asia

The paucity of spatial and depth-dependent groundwater pumpinginformation leads to limitations in data-guided groundwater levelmodeling in stressed South Asian aquifers. Groundwater depletion in South Asia's Himalayan,transboundaryIndus-Ganges-Brahmaputra-Meghna (IGBM) rivers basin is among the highestglobally. Given the high irrigation demand and population, groundwatersustainability requires an improved understanding of groundwater systemsfor the accurate prediction of groundwater levels (GWLs). However,the prediction of groundwater system behaviors is a significant challengesince it is dominated by spatiotemporal and subsurface depth-dependentdrivers. Earlier studies that address the challenges are mainly basedon the short spatial and temporal extent and/or do not separate therenewable (i.e., shallow) vs nonrenewable (i.e., deeper) groundwatersignals. Here, we first identified the variable importance of spatialand depth-dependent drivers on GWL in the IGBM basin. Our resultsindicate a greater influence of anthropogenic factors (i.e., widespreadpumping and increased population) in most parts of the IGBM basin,except in the precipitation-dominated basin of the Brahmaputra. Ournext purpose was to delineate a multifactorial approach for GWL predictionusing the two most used machine learning models (i.e., support vectormachine and feed-forward neural network) in the literature. In general,the machine learning model outputs show a good match in comparisonto the GWL from the observation wells (n = 2303 distributedacross India and Bangladesh) with some limitations in areas with increasedgroundwater irrigation. We separately compared the results from shallow(<35 m) and deep (>35 m) observation wells, emphasizing thesignificanceof deep groundwater pumping. Our approach highlights the importanceof spatiotemporal to multidepth factors in GWL prediction and canbe adopted in other parts of the globe to predict GWLs.

Automated summary

The lack of spatial and depth-dependent groundwater pumping information hinders data-guided groundwater level modeling in stressed South Asian aquifers. Machine learning models show promise in accurately predicting groundwater levels in the affected regions, with emphasis on the significance of temporal and depth factors.