What does this research mean for the field?

Dugwell density in the Banki watershed is significantly influenced by population density and agricultural areas, with a notable shift towards aquifer-centric ecosystem service delivery due to land use changes. Novelty: ClaimNovelty.NOVEL_FINDING. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

The study aims to evaluate groundwater development in the Banki watershed using hydrosociological indicators and machine learning techniques.

March 16, 2026Open Access

Assessment of groundwater development in watersheds using novel hydrosociological indicators and machine learning approach

Key Points

The study aims to evaluate groundwater development in the Banki watershed using hydrosociological indicators and machine learning techniques.
Assessed 4,830 dugwell developments across 591 grids in the Banki watershed.
Evaluated empirical hydrosociological indicators: dugwell frequency, density, and abundance.
Utilized ensemble random forest algorithm to model dugwell occurrence and density based on environmental and anthropogenic variables.
Most sub-watersheds had dugwell frequency over 50%, indicating high groundwater dependence.
Dugwell density showed a strong positive relation with population density and agricultural areas (R2 > 0.85).
Random forest predictions demonstrated good accuracy for dugwell occurrence (ROC = 0.890) and density (R2 = 0.893).

Abstract

The impact of land use changes on watershed morphometry has shifted the delivery of ecosystem services from river-centric to aquifer-centric in the Ganga River Basin. We assessed the hydrosociology of 4,830 dugwell developments in 591 grids (1 km × 1 km size) of the 518.35 km 2 dry tropical Banki watershed using the empirical indicators: dugwell frequency (DwF), dugwell density (DwD), and dugwell abundance (DwA). Notably, most sub-watersheds exhibited DwF exceeding 50 %, indicating widespread groundwater dependence. The DwD exhibited positive regression (R 2 > 0.85) with population density, the number of households, agricultural areas, and subwatershed areas. Moreover, DwD increased with a decrease in the number of streams, length of streams, and drainage density. These extreme morphometric changes are irreversible tipping points, and native people are overcoming spatiotemporal shortages in surface water availability by constructing dugwells at different depths in and around existing and extinct streams. Further, the ensemble random forest algorithm significantly predicted the dugwell occurrence probability (TSS = 0.563; ROC = 0.890) and dugwell density (R 2 = 0.893; RMSE = 3.189) based on seventeen environmental and six anthropogenic variables, respectively. The variable importance feature identified distance to dugwells and local dugwell density as the key predictors influencing dugwell development, highlighting a complex interaction between hydrogeological suitability and human demand. Our proposed hydrosociological framework offers a scalable and transferable approach to assessing groundwater development at watershed, sub-basin, and basin scales, underscoring the need for immediate attention to aquifer conservation through stream rejuvenation in small watersheds. • Hydrosociology of 4830 dugwells quantified in a critically endangered watershed • Increased aquifer-centric delivery of ecosystem services • Dugwell hotspots formed near existing and extinct streams • RF modeling effectively predicted dugwell occurrence probability and density

Assessment of groundwater development in watersheds using novel hydrosociological indicators and machine learning approach

Key Points

Abstract

Cite This Study