Two centuries of canal building and tube-well pumping have reshaped the aquifer between the Ganges and the Yamuna north of Delhi. Ganges-Yamuna groundwater decline since the 1970s has pushed the median water table about 2 m below its level before the canals were dug, according to a modelling study. Over the same period, most of the Hindon river has switched from a stream fed by groundwater to one that leaks water into the aquifer.
The paper, “Changes in groundwater-surface water interactions following two centuries of irrigation practices and groundwater use in the Upper Ganges-Yamuna interfluve, North India”, was published in Hydrology and Earth System Sciences in October 2026. The lead and corresponding author is Frank van Broekhoven of the Copernicus Institute of Sustainable Development at Utrecht University. His co-authors are Stefan Dekker, Jasper Griffioen and Paul Schot of Utrecht University and Anjali Bhagwat of the Groundwater Hydrology Division at the National Institute of Hydrology, Roorkee.
Modelling 216 years of water use
The study area covers about 16,500 km² of western Uttar Pradesh and southern Uttarakhand. It runs up to 220 km from north to south and up to 100 km across. The Hindon, a rain-fed tributary of the Yamuna joined by the Krishni and Kali rivers, flows through its centre. Monsoon rain between June and September brings about 700 to 950 mm, and the rest of the year adds only 100 to 200 mm.
The authors built a groundwater model in MODFLOW 6 with cells 1 km square and ran it in yearly steps from 1800 to 2016. They fed it rainfall and leakage from the Eastern Yamuna Canal from 1830 and the Upper Ganga Canal from the 1850s. They added irrigation return flow, municipal return flow and abstraction for farms, towns and industry. They ran 1,500 versions with different aquifer properties. They kept the 196 that matched levels in 94 observation wells between 1990 and 2018 and passed checks on realism. The retained runs reproduced observed levels with an R² of 0.86.
From canal surplus to pumping deficit
The model describes three phases. Before 1830, rain supplied the recharge and the system stayed in balance, with a median water table about 7.5 to 8 m below ground. Canal leakage then raised median water tables by about 0.5 to 1 m across the area, and by up to about 5 m in places. The extra water pushed more groundwater into the rivers. Levels stayed fairly stable from 1900 to 1970.
In the model’s present-day water balance, rainfall supplies about 50% to 80% of recharge. Irrigation water returning to the aquifer adds about 10% to 25%, canal leakage about 10% and municipal return flow about 2.5%. Each 100 m of unlined canal leaks about 0.32 litres a second, the figure the authors use to set canal seepage.
Abstraction rose sharply from the 1970s, with the steepest increase between 1980 and 2010. Irrigation now accounts for about 85% of withdrawals and municipal supply for about 15%, the authors report. In the district data they used, sugarcane covers 35.8% of the cropped area, wheat 30.4% and rice 9.5%. Around 2000, average abstraction overtook recharge. Within the model’s uncertainty range, that crossover may have come as early as 1970.
By 2016 the median water table lay about 10 m below ground, roughly 2 m deeper than before the canals. The modelled median fell by about 2 m between 1985 and 2013. Observation wells show a median trend of about minus 0.25 m a year, and some fall by more than 1 m a year. The authors find that the rate of decline eased from about 2000.
The sharpest change is in the rivers. In most retained runs, the Hindon network switched around 2000 from receiving groundwater to losing water into the aquifer. Some runs place the switch in the 1970s. Reaches near the Shivalik foothills still gain water from the ground. The Ganges and the Yamuna receive less groundwater than they once did, but both remain net gaining rivers in nearly all runs.
Limits listed by the authors
The authors are explicit about the model’s simplicity. It treats the aquifer as a single uniform layer, uses yearly steps that hide seasonal swings and keeps river levels constant. It leaves out evaporation from shallow groundwater and uptake by roots, which may overstate historical high water tables. Rainfall before 1900 relies on long-term averages. Many validation wells sit in towns, where declines tend to be steep. The authors advise reading the results at regional scale and caution against applying them to single sites.
Their results also differ from an earlier study of the same region. Citing MacAllister and colleagues (2022), they note that work found canal-driven rises of several metres, up to 20 m, and current levels above those of a century ago. The new model finds rises averaging 0.5 to 1 m and present levels below natural conditions.
The consequences for the Hindon are practical. With less groundwater feeding it, the authors write, the river depends more on wastewater, dilutes pollution less and risks contaminating shallow aquifers with polluted river water. Restoring the exchange between river and aquifer that existed before 1970 would require structural cuts in irrigation pumping or a large increase in recharge, they conclude. The sugarcane and wheat belt north of Delhi draws most of its irrigation from wells. For that belt, the study dates the point at which Ganges-Yamuna groundwater decline began to take water from the rivers as well as the ground.
Sources: van Broekhoven, F. J. G., Dekker, S. C., Griffioen, J., Bhagwat, A. and Schot, P. P. (2026). Changes in groundwater-surface water interactions following two centuries of irrigation practices and groundwater use in the Upper Ganges-Yamuna interfluve, North India. Hydrology and Earth System Sciences, 30, 6315–6337. https://doi.org/10.5194/hess-30-6315-2026. Licence: CC BY 4.0. Earlier work cited: MacAllister et al. (2022), as cited in the paper.
