Agricultural intensification and groundwater depletion in Haryana: A post-Green Revolution analysis

Introduction

Background

Groundwater is the largest source of irrigation water in the world, and in many aquifers, withdrawals now exceed natural replenishment (Famiglietti, 2014) [1] . India is the largest user of groundwater, and its northwestern plains have become one of the most closely studied cases of aquifer decline. Satellite gravity measurements first showed substantial groundwater loss across Haryana, Punjab and Rajasthan (Rodell et al., 2009; Tiwari et al., 2009) [2,3] , and observations from wells in the Indo-Gangetic Basin have since confirmed that depletion is concentrated in this part of the basin (Macdonald et al., 2016) [4] . Both pumping and changes in monsoon precipitation shape groundwater storage in India. Governance is made harder by the sheer number of small, privately operated wells, a condition Shah (2009) [5] describes as an anarchy of decentralised users, and by the limits of supply-side responses (Shankar et al., 2011) [6] combines these demand and supply influences in one ratio of draft to availability (Kumar et al., 2022; Velis et al., 2017) [19,20] , which is why it is the basis of the CGWB classification of assessment units. Non-parametric tests are the usual instruments for detecting change in hydrological series. The Mann-Kendall test (Mann, 1945; Kendall, 1975) [22,21] and Sen's slope estimator (Theil, 1950; Sen, 1968) [24,23] make no assumption of normality and are not unduly affected by outliers, while the Pettitt test (Pettitt, 1979) [25] locates a single abrupt change in the central tendency of a series. Applied together, they allow a trend to be established, measured and dated. Their main weakness is that serial correlation in a series can overstate significance, which has led to corrected versions of the test (Hamed et al., 1998; Yue et al., 2002) [26,27] .

Research gap and objectives

Much of the evidence on groundwater decline in north-west India comes from satellite products that cover large regions, or from statelevel aggregates. Both can conceal the fact that districts of the same state follow very different trajectories. District-level comparisons across the full range of CGWB stress categories, covering three decades and using formal tests of trend, magnitude and structural break, appear to have received less attention. This is the gap the present paper addresses. The specific objectives are:

1. to describe the long-run growth in tubewells, irrigated area and water-intensive cropping, which together shape groundwater extraction in Haryana;

2. to test the direction, statistical significance and magnitude of groundwater level trends in five districts representing all four CGWB categories; and 3. to identify the year in which each district's groundwater series shifted abruptly, and to draw suggestions for groundwater management from the results.

The remainder of the paper is organised as follows. Section 2 describes the study area, data and methods. Section 3 presents the results, Section 4 discusses them, and Section 5 concludes with suggestions.

Study area, data and methods

Selection of districts

Districts were selected using the CGWB classification of assessment units into four groundwater stress categories. The classification rests on the stage of groundwater extraction, defined as the ratio of annual groundwater draft to net annual groundwater availability, expressed as a percentage (Table 1). Using this classification as the sampling frame ensures that the districts examined span the full range of exploitation conditions in the state instead of a single level of stress. Selection proceeded in stages. In the first stage, the 22 districts of Haryana were assigned to the four categories (16 over-exploited, one critical, two semi-critical and three safe). In the second stage, districts were drawn at random within each category using IBM SPSS Statistics, at approximately 20 per cent for the over-exploited category and 40 per cent for the others, which gave Sirsa and Mahendergarh (over-exploited), Palwal (critical), Hisar (semi-critical) and Jhajjar (safe). In the third stage, blocks were drawn at random within each district at approximately 40 per cent: Rania, Dabwali and Ellenabad in Sirsa; Mahendergarh, Ateli Nangal, Narnaul and Nangal Chaudhary in Mahendergarh; Hassanpur in Palwal; Barwala in Hisar; and Badli, Beri, Matanhail and Machhrauli in Jhajjar. Villages were then chosen purposively, in consultation with the District Agriculture Officer, according to the irrigation technique practised. The block and village stages serve the primary survey of the wider study; the trend analysis reported here uses district-level data.

Data and variables

The analysis uses annual average groundwater level data from the CGWB Ground Water Year Books, which report piezometric levels at the well, block and district levels. District-level series were used for the five selected districts.

Trend and change-point analysis

Three complementary non-parametric techniques were applied to the annual series of each district. Using several tests, rather than one, gives a fuller picture of how groundwater levels changed over three decades and of the different features of that change. All tests were run in R through the Posit Cloud environment.

Mann-Kendall test

The Mann-Kendall test is a rank-based test for the presence of a monotonic trend. It does not require normally distributed data and is relatively insensitive to extreme observations. For a series x 1 , x 2 , …, x n , the test statistic S compares each observation with all later ones:

where sgn(θ) equals +1, 0 or -1 according to whether θ is positive, zero or negative. A positive S indicates a predominance of increasing pairs and a negative S a predominance of decreasing pairs. The variance of S, allowing for ties, is where q is the number of tied groups and t p the number of observations in the p-th group. The standardised statistic is

The null hypothesis (H 0 ) is that the series has no monotonic trend; the alternative (H 1 ) is that a significant monotonic trend exists. Kendall's τ is reported alongside Z as a measure of the strength of the association between level and time. Because levels are measured as depth, a significant positive Z signals deepening of the water table and hence depletion, and a significant negative Z signals shallowing.

Sen's slope estimator

Sen's slope is the median of the slopes between all pairs of observations, which makes it resistant to outliers. The slope between observations i and j is

There are N = n(n-1)/2 such slopes. When arranged in ascending order, the Sen's slope is their median:

A positive slope indicates a rising depth to groundwater, that is, depletion, and a negative slope indicates shallowing. The slope is expressed in metres per year and is reported with its 95 per cent confidence interval.

Pettitt test

The Pettitt test detects a single abrupt change in the central tendency of a series. It builds on the Mann-Whitney statistic between the two parts of the series on either side of each candidate point t:

The test statistic is the largest absolute value of U t,n , and the year at which it occurs is the probable change point:

An approximate two-sided p-value is given by

Throughout, significance is reported at the 1 per cent ( * * * ), 5 per cent ( * * ) and 10 per cent ( * ) levels. The straight lines in Figures 1 and2 are ordinary least squares fits and serve only as a visual guide; inference rests on the tests described above.

Results

Expansion of extraction capacity and water-intensive cropping

Before the groundwater series are examined, it is useful to see what has happened to the factors that drive extraction. Table 2 shows the change between 1966-67, the first year after the state's formation, and 2023-24. Gross cropped area grew by almost half, but gross irrigated area more than tripled, so that cultivation became far more dependent on irrigation. The area under rice rose about nine-fold and that under wheat by more than three-fold, whereas the area under bajra fell by about two-fifths and that under jowar almost vanished. [28] .

Extraction capacity moved in the same direction.

Graphical patterns in district groundwater levels

Over-exploited and critical districts: Sirsa, Mahendergarh and Palwal

Figure 1. plots the annual groundwater depth of the two over-exploited districts and the critical district, with fitted linear trends. Despite differences in aquifer conditions and CGWB category, the three series share a common shape. From the early 1990s to the late 2000s, levels were relatively stable or fluctuated within a band, with short dips and recoveries that appear to reflect rainfall variability, cropping intensity and recharge and not sustained depletion. A structural transition is visible around 2011-2012 in each district, after which depth rises along a deeper and more persistent path, with abrupt increases followed by only partial correction, which is characteristic of aquifers under heavy tubewell extraction. In Sirsa, the water table stood near 11.5 mbgl in the early 1990s and eased to about 10.5 mbgl by 1996. Between 1997 and 2008 it oscillated within a narrow band of roughly 9.8 to 11.8 mbgl. After 2008, depth rose more steeply and year-to-year variability grew, with a sharp rise to nearly 15 mbgl around 2012 and peaks above 16 mbgl around 2018-2019, the deepest in the series. Each peak was followed by a partial recovery to a level still deeper than before, which indicates that the aquifer does not fully recover between dry spells. From 2021 to 2023 levels stayed above 13 mbgl. The fitted trend rises from below 10 mbgl in 1990 to above 14 mbgl in 2023. Mahendergarh began the period at a much greater depth, around 22. The three districts thus differ in magnitude but not in direction. Mahendergarh is the most acute case, Sirsa shows moderate but persistent deepening accompanied by growing variability, and Palwal has the shallowest absolute depths but follows the same upward path. The deepening is therefore not confined to the over-exploited districts.

Semi-critical and safe districts: Hisar and Jhajjar

Figure 2 shows a different picture for Hisar and Jhajjar. In both, the fitted trend slopes downward, so that depth to water decreased over the period, and in both the steepest change took place in the mid-1990s, when the water table shallowed sharply within a short span. Later conditions fluctuated, shaped more by rainfall and recharge than by sustained depletion. In Hisar, depth was at its highest point of about 10.4 mbgl in 1990, fell to about 8.7 mbgl by 1994-95 and reached a series low of around 6.8 mbgl in 1999. Between 2000 and 2010 it rose again, to a local peak near 9 mbgl, and from 2011 to 2023 it moved between about 7.2 and 8.7 mbgl without a clear direction before declining to about 7.6 mbgl in 2021 and 6.8 mbgl in 2023, the lowest since 1999. The fitted trend falls from about 8.95 to 7.4 mbgl. This gentle decline owes much to the sharp early recovery of the 1990s, and the series has been fluctuating, and at times reversing, since about 2010.

Jhajjar started at about 6.1 mbgl and was stable until 1994, then fell to about 4.7 mbgl in 1995 and 4.3 mbgl in 1996. It varied between about 4.4 and 5.3 mbgl from 1997 to 2004, and then shallowed steadily to a low of about 2.9 mbgl in 2011. After 2012 it fluctuated, rising to about 4.5 mbgl in 2015-2016, falling to about 2.8 mbgl in 2019, the lowest of the series, and ending at about 3.6 mbgl in 2023. The fitted trend falls from about 5.65 to 3.1 mbgl and is the steepest downward trend among the five districts. Unlike in Hisar, the improvement is spread across much of the period and is not driven by a single early episode.

Statistical assessment of trends and change points

Tables 3 and4 report the formal tests. Each district's series has 34 observations. Districts are listed in order of CGWB category, from overexploited to safe. Palwal deserves particular note. Although it carries the milder "critical" label, its Z and τ are slightly higher than Sirsa's and its rate of deepening is practically identical (0.144 against 0.143 m per year). By the evidence of trend alone, the critical district is deteriorating as steadily as an over-exploited one.

Hisar and Jhajjar move in the opposite direction. Jhajjar records a strong and highly significant fall in depth (Z = -5.011; τ = -0.604), at 0.081 m per year, and its confidence interval lies entirely below zero. Its τ is, in absolute terms, the largest of the five. Hisar's result is much weaker. The Z value of -1.883 and p-value of 0.060 reach significance only at the 10 per cent level, the slope of -0.034 m per year is less than half that of Jhajjar, and its confidence interval (-0.082 to 0.001) includes values at and just above zero. The improvement in Hisar is therefore suggestive and not firmly established. The Pettitt test finds a statistically significant change point in every district (Table 4). In the three depleting districts the break falls between 2007 and 2012: 2007 in Sirsa, 2011 in Mahendergarh and 2012 in Palwal. These dates agree with the visual evidence of a steeper, more variable regime after 2008 in Sirsa and after 2011 in Mahendergarh and Palwal. In Sirsa, the result is best read as an abrupt break and not a slow transition. In Jhajjar the break is dated to 2006, consistent with the sustained downward drift from the mid-2000s. In Hisar the break falls early, in 1995, and is significant at the 5 per cent level only. It marks the tail of the sharp early recovery after the deep starting position of 1990, so that its meaning differs from the breaks in the depleting districts: it separates a phase of rapid recovery from a phase of moderate fluctuation and not a phase of stability from one of accelerating depletion.

Discussion

A stress gradient that follows the CGWB classification, with one exception

The five districts line up along a gradient that broadly mirrors the CGWB categories. Deepening is strongest and most significant in the over-exploited districts, present at a comparable rate in the critical district, weak and marginal in the semi-critical district, and reversed in the safe district. This agreement with the official classification is reassuring, because the classification is based on a different quantity (extraction relative to availability) than the water level used here. The consistency suggests that the stage of extraction captures real differences in the condition of the aquifers. The exception is the closeness of Palwal and Sirsa. If categories were a precise guide to the speed of decline, the critical district would be expected to deepen more slowly than the over-exploited ones. It does not, and its τ and Z are somewhat higher. This is a warning that the category labels describe the state of an aquifer at a point in time, while the trend describes its direction. A district can sit in the milder category and yet be on the same downward path, and there is a case for treating it as a district in which preventive action is still cheap. The shallower absolute depths of Palwal (the fitted line reaches only about 9.4 mbgl by 2023) mean that it has more room than Sirsa and far more than Mahendergarh, but the rate of use of that room is similar to Sirsa's.

Mahendergarh as the outlier

Mahendergarh stands apart. Its rate of 0.74 m per year, its starting depth of more than 20 m and its observed levels approaching 47 mbgl place it in a different class from the other districts. The finding is consistent with the expectation that the state's most acute depletion is concentrated in a few districts, which is where interventions are likely to yield the greatest returns. Two cautions apply. First, the rise from about 18 to nearly 27 mbgl within a single year is large for an aquifer response, and an increase of this size can also arise from changes in the set of observation wells included in the district average, from well deepening, or from measurement conditions. Checking the well-level records behind the district mean would show how much of the jump is physical. Second, the same sharpness is what gives the 2011 change point its high significance, so the two findings should not be treated as independent evidence.

Timing of the shifts

Four of the five change points fall within 2006-2012, and the three depleting districts cluster in 2007-2012. The period is one in which tubewells continued to multiply, the shift towards rice and wheat was well established, and irrigated area was already extensive. Rainfall variability is another candidate, because years of weak monsoon recharge combined with undiminished pumping will deepen the water table sharply, as the abrupt rises followed by partial correction in the figures suggest. The present data do not allow these explanations to be separated, and the cause of the break is a hypothesis to be tested and not a finding of the paper. One further point of method is relevant. When a series trends steadily, the Pettitt test tends to place its break near the middle of the record, and the dates for Sirsa (2007) and Jhajjar (2006) lie almost exactly at the midpoint of 1990-2023. The dates should therefore be read as approximate markers of a regime change and not as exact years.

Is the improvement in Hisar and Jhajjar secure?

The improvements in Hisar and Jhajjar should be read with care. Both owe much to a sharp shallowing in the mid-1990s, and in both the later record shows reversals: troughs in Jhajjar in 2011 and 2019, a rise to about 4.5 mbgl in 2015-2016, and renewed shallowing in Hisar at the very end of the series. These swings suggest that the water table in both districts remains sensitive to annual rainfall and recharge. A shallower water table is not necessarily a sign of ample water for all purposes. Very shallow levels can bring waterlogging and salinity problems in some irrigated tracts, and the study does not assess groundwater quality. The "safe" status should therefore be understood as a margin to be protected, not as a licence to expand extraction.

Limitations

Several limitations qualify the results. The series are district averages of annual readings, so spatial variation within districts and seasonal behaviour are not captured, and the number and location of wells in the average may have changed over time. The Mann-Kendall test as applied here assumes serially independent observations, and the analysis does not adjust for autocorrelation, which could overstate significance (Hamed et al., 1998; Yue et al., 2002) [26,27] ; the large Z values in the depleting districts make it unlikely that the conclusions would change, but the weaker Hisar result is more exposed to this concern. The Pettitt test finds only one change point, whereas some series may contain more. Finally, rainfall, tubewell density and cropping pattern are used to interpret the findings but are not included as explanatory variables, so the paper describes the pattern of depletion without estimating its causes.

Conclusions and suggestions

Conclusions

This paper has examined three decades of groundwater level data for five Haryana districts chosen to represent the four CGWB stress categories. Groundwater depth has increased significantly in Sirsa (0.143 m per year), Mahendergarh (0.74 m per year) and Palwal (0.144 m per year), each with p < 0.001, and abrupt shifts in the mean level occurred between 2007 and 2012. In the safe district of Jhajjar the water table has become significantly shallower (-0.081 m per year), and in the semi-critical district of Hisar the shallowing is small and only marginally significant (-0.034 m per year, p = 0.060), largely reflecting a recovery in the 1990s. The results therefore support the claim that groundwater depletion in Haryana is spatially uneven, strongly concentrated in a few districts, and not limited to those formally classified as over-exploited. The depleting districts hold the greatest need for intervention, while the districts that are improving need protection of what has been gained.

Suggestions

The following suggestions are drawn from the findings and from the wider literature on groundwater management in north-west India. They are offered as directions for policy and not as outputs of the statistical analysis, and the details of current schemes should be verified before they are cited in a policy context. Palwal should be treated as a case for early action, since its decline matches that of an over-exploited district even though its depths are still modest.

2. Reduce the area under water-intensive crops in stressed districts. The rise in rice area from 192 thousand to 1,778 thousand hectares is the most conspicuous change in the cropping pattern. Incentives for shifting from paddy, such as Haryana's "Mera Pani Meri Virasat" scheme and support for direct-seeded rice, should be concentrated where depletion is fastest, and paired with assured procurement and remunerative prices for alternatives such as maize, pulses, oilseeds and bajra. The decline in bajra area suggests that a return to such crops in dry districts is agronomically plausible, but it will not happen without a reliable market.

3. Promote efficient irrigation, with a cap on total extraction. Drip and sprinkler systems can lower the water applied per hectare, but savings at the field level do not always become savings at the aquifer level, because farmers may expand irrigated area or switch to thirstier crops (Grafton et al., 2018) [29] . Subsidies for micro-irrigation should therefore be linked to crop choice and to limits on district-level draft, so that efficiency gains are not absorbed by additional consumption.

4. Reform the energy-water link. Where electricity for tubewells is supplied at a flat rate or free of marginal charge, the farmer has little reason to economise on pumping. Metering of agricultural connections, rationing of supply hours, feeder separation and pilot schemes that pay farmers for electricity saved deserve consideration. Subsidised solar pumps should be accompanied by caps on extraction, since a zero-marginal-cost source can otherwise encourage more pumping.

5.

Invest in recharge where aquifer conditions allow. Rainwater harvesting, rejuvenation of village ponds, check structures and managed aquifer recharge using surplus canal or treated water can reduce the deficit between draft and availability. Sites should be selected using aquifer mapping, and efforts should focus on Mahendergarh, Sirsa and Palwal. Conjunctive use of canal water and groundwater, where canal supply exists, can ease pressure on tubewells.

Strengthen regulation, registration and monitoring.

Registration of tubewells, volumetric measurement of draft, and the effective use of the state's groundwater authority and community-based programmes such as the Atal Bhujal Yojana would give regulators the information they now lack. A denser network of piezometers with telemetric recording would allow blocklevel trends to be tracked and the sudden jumps noted in Mahendergarh to be verified.

7. Protect the gains in Hisar and Jhajjar. Improving districts should not be treated as free of constraint. Continued monitoring of both levels and quality, caution in expanding paddy, and preservation of recharge areas are advisable, since the later series show that the improvement can reverse in dry years.

References

1. Famiglietti JS. The global groundwater crisis. Nature Climate Change. 2014;4(11):945-948.
2. Rodell M, Velicogna I, Famiglietti JS. Satellite-based estimates of groundwater depletion in India. Nature. 2009;460:999-1002.
3. Tiwari VM, Wahr J, Swenson S. Dwindling groundwater resources in northern India, from satellite gravity observations. Geophysical Research Letters. 2009;36:L18401.
4. Macdonald AM, Bonsor HC, Ahmed KM, et al. Groundwater quality and depletion in the Indo-Gangetic Basin mapped from in situ observations. Nature Geoscience. 2016;9:762-766.
5. Shah T. Taming the Anarchy: Groundwater Governance in South Asia. Washington, DC: Resources for the Future Press; 2009.
6. Shankar PSV, Kulkarni H, Krishnan S. India's groundwater challenge and the way forward. Economic and Political Weekly. 2011;46(2):37-45.
7. Singh RB. Environmental consequences of agricultural development: A case study from the Green Revolution state of Haryana, India. Agriculture, Ecosystems & Environment. 2000;82(13):97-103.
8. Central Ground Water Board. Ground Water Resources of India. CGWB, Faridabad; 1995.
9. Bhardwaj M, Dutta T. Analysis of determinants of groundwater depletion in rice-wheat ecosystems of Punjab-a case of India. Environment, Development and Sustainability. 2024;27(12):29969-29987.
10. Bhattarai N, Pollack A, Lobell DB, Fishman R, Singh B, Dar A, Jain M. The impact of groundwater depletion on agricultural production in India. Environmental Research Letters. 2021;16(8):085003.
11. Khara DS, Ghuman RS. Depleting Water Tables and Groundwater Productivity Issues in India's Green Revolution States. Journal of Asian and African Studies. 2024;59(6):2078-2091.
12. Rahman KU, Hussain A, Ejaz N, Shahid M, Duan Z, Mohammadi B, Hoai PN, Pham QB, Khedher KM, Anh DT. Evaluating the impact of the environment on depleting groundwater resources: A case study from a semi-arid and arid climatic region. Hydrological Sciences Journal. 2022;67(5):791-805.
13. Kasana A, Singh O. Groundwater Irrigation Economy of Haryana:A Glimpse into Spread, Extent and Issues. Journal of Rural Development. 2017;36(4):531.
14. Rosencranz A, Puthucherril TG, Tripathi S, Gupta S. Groundwater management in India's Punjab and Haryana: A case of too little and too late? Journal of Energy & Natural Resources Law. 2022;40(2):225-250.
15. Dangar S, Asoka A, Mishra V. Causes and implications of groundwater depletion in India: A review. Journal of Hydrology. 2021;596:126103.
16. Das A, Thangavel M, Chinara M, Paria B. How Groundwater Extraction is Determined in Crop Production? -An Empirical Study in a Coastal District of Odisha. Water Economics and Policy. 2026;12(01):2540016.
17. Meng F, Khan MI, Naqvi SAA, Sarwar A, Islam F, Ali M, Tariq A, Ullah S, Soufan W, Faraj TK. Identification and mapping of groundwater recharge zones using multi influencing factor and analytical hierarchy process. Scientific Reports. 2024;14(1):19240.
18. Moeck C, Grech-Cumbo N, Podgorski J, Bretzler A, Gurdak JJ, Berg M, Schirmer M. A global-scale dataset of direct natural groundwater recharge rates: A review of variables, processes and relationships. Science of The Total Environment. 2020;717:137042.
19. Kumar R, Bharti B. Estimation of Groundwater Recharge and Draft of Arwal District Watershed based on Changing Trend of Rainfall and Population Forecasting. Journal of Water Engineering and Management. 2022;3(1):27-37.
20. Velis M, Conti KI, Biermann F. Groundwater and human development: Synergies and trade-offs within the context of the sustainable development goals. Sustainability Science. 2017;12(6):1007-1017.
21. Kendall MG. Rank Correlation Methods (4th ed.). Charles Griffin, London; 1975.
22. Mann HB. Nonparametric tests against trend. Econometrica. 1945;13(3):245-259.
23. Sen PK. Estimates of the regression coefficient based on Kendall's tau. Journal of the American Statistical Association. 1968;63(324):1379-1389.
24. Theil H. A rank-invariant method of linear and polynomial regression analysis. Proceedings of the Koninklijke Nederlandse Akademie van Wetenschappen. 1950;53:386-392.
25. Pettitt AN. A non-parametric approach to the change-point problem. Journal of the Royal Statistical Society: Series C (Applied Statistics). 1979;28(2):126-135.
26. Hamed KH, Ramachandra Rao A. A modified Mann-Kendall trend test for autocorrelated data. Journal of Hydrology. 1998;204(1-4):182-196.
27. Yue S, Pilon P, Phinney B, Cavadias G. The influence of autocorrelation on the ability to detect trend in hydrological series. Hydrological Processes. 2002;16(9):1807-1829.
28. Department of Economic and Statistical Analysis, Government of Haryana. Statistical Abstract of Haryana (2023-24 and 2024-25 editions). Chandigarh.
29. Grafton RQ, Williams J, Perry CJ, Molle F, Ringler C, Steduto P, Udall B, Wheeler SA, Wang Y, Garrick D, Allen RG. The paradox of irrigation efficiency. Science. 2018;361(6404):748-750.
30. Asoka A, Gleeson T, Wada Y, Mishra V. Relative contribution of monsoon precipitation and pumping to changes in groundwater storage in India. Nature Geoscience. 2017;10:109-117.
31. Gleeson T, Wada Y, Bierkens MFP, Van Beek LPH. Water balance of global aquifers revealed by groundwater footprint. Nature. 2012;488(7410):197-200.
32. How to cite this article: Kadian HS, Kaur S, Kumar S, Sangwan D. Agricultural intensification and groundwater depletion in Haryana: A post-Green Revolution analysis. International Journal of Agriculture Extension and Social Development. 2026;9(10):110-116. doi: https://www.doi.org/10.33545/2618072