Tracing groundwater recharge sources in the northwestern Indian alluvial aquifer using water isotopes ( d 18 O, d 2 H and 3 H)

Rapid groundwater depletion from the northwestern Indian aquifer system in the western Indo-Gangetic basin has raised serious concerns over the sustainability of groundwater and the livelihoods that depend on it. Sustainable management of this aquifer system requires that we understand the sources and rates of groundwater recharge, however, both these parameters are poorly constrained in this region. Here we analyse the isotopic ( d 18 O, d 2 H and tritium) compositions of groundwater, precipitation, river and canal water to identify the recharge sources, zones of recharge, and groundwater ﬂow in the Ghaggar River basin, which lies between the Himalayan-fed Yamuna and Sutlej River systems in northwestern India. Our results reveal that local precipitation is the main source of groundwater recharge. However, depleted d 18 O and d 2 H signatures at some sites indicate recharge from canal seepage and irrigation return ﬂow. The spatial variability of d 18 O, d 2 H, d-excess, and tritium reﬂects limited lateral connectivity due to the heterogeneous and anisotropic nature of the aquifer system in the study area. The variation of tritium concentration with depth suggests that groundwater above c. 80 mbgl is generally modern water. In contrast, water from below c. 80 mbgl is a mixture of modern and old waters, and indicates longer residence time in comparison to groundwater above c. 80 mbgl. Isotopic signatures of d 18 O, d 2 H and tritium suggest signiﬁcant vertical recharge down to a depth of 320 mbgl. The spatial and vertical variations of isotopic signature of groundwater reveal two distinct ﬂow patterns in the aquifer system: (i) local ﬂow (above c. 80 mbgl) throughout the study area, and (ii) intermediate and regional ﬂow (below c. 80 mbgl), where water recharges aquifers through large-scale lateral ﬂow as well as vertical inﬁltration. The understanding of spatial and vertical recharge processes of groundwater in the study area provides important base-line knowledge for developing a sustainable groundwater management plan for the northwestern Indian aquifer system. (cid:1) 2018 The Authors. Published by Elsevier B.V. ThisisanopenaccessarticleundertheCCBYlicense(http://


Introduction
Groundwater depletion from major alluvial aquifer systems is a global issue (e.g., Foster and Chilton, 2003;Wada et al., 2012;Gleeson et al., 2015).The global groundwater extraction rate is $1500 km 3 per year (Doll et al., 2012), which is more than the natural groundwater recharge rate.Increasing water demand for economic development, power generation, drinking water, and agriculture has exacerbated groundwater exploitation, leading to the decline of the water table in major alluvial aquifer systems across the world (Konikow and Kendy, 2005;Aeschbach-Hertig and Gleeson, 2012).Large-scale groundwater depletion has resulted in land subsidence, reduction in the base flow of springs and rivers during dry periods, saltwater intrusion, water quality degradation, and damage to aquatic ecosystems in different parts of the world (Fishman et al., 2011).This problem is becoming more challenging given the increasing demands by development and a growing population, along with poorly understood effects of climate-driven changes in the water cycle (e.g., Aeschbach-Hertig and Gleeson, 2012).These broad issues can be addressed only if we know the sources and areas of groundwater recharge along with the interconnectivity and dynamics of aquifer systems at appropriate regional scales.This information, however, is poorly constrained for many of the world's major alluvial aquifer systems.
The northwestern Indian aquifer system (NWIA) underlies the states of Punjab, Haryana, and Rajasthan and represents one of the major alluvial aquifer systems of the Indo-Gangetic basin in northern India and Pakistan (MacDonald et al., 2016).The region is characterized by multiple, semi-confined sand-rich aquifer bodies that are laterally discontinuous but may be highly interconnected (van Dijk et al., 2016a;MacDonald et al., 2016).This is one of the most agriculturally intensive regions in India, where annual food grain production has increased four fold from 50 million tons in 1950 to 203 million tons in 1999-2000 (Kumar et al., 2005).Such intensive agricultural activities are mainly attributed to the so-called 'green revolution' in India, aimed at achieving self-reliance in food production.This intensive food production has led, however, to greatly-accelerated demand for irrigation water.Since India's surface water infrastructure was never adequate to meet this requirement, the focus shifted to groundwater extraction using large numbers of tube wells.The average tube well density in this region is >15 km À2 (Ambast et al., 2006).As a consequence, the groundwater level in the NWIA is declining at a much higher rate than any other comparably-sized aquifer on the Earth (Rodell et al., 2009;Tiwari et al., 2009;Chen et al., 2014;Panda and Wahr, 2015;Long et al., 2016).Satellite-borne gravity measurements suggest that groundwater levels declined at $3.1 ± 0.1 cm per year between 2005 and 2010 (Long et al., 2016) and that water was lost at a rate of $54±9 km 3 per year between 2002 and 2008 (Tiwari et al., 2009); the average rate of groundwater depletion is $ 20.4 ± 7.1 gigatonnes per year for 10 years from 2003 to 2012 (Chen et al., 2014).
Sustainable management of the NW Indian aquifer system needs a comprehensive understanding of the sources and rates of groundwater recharge.In addition, it is also important to know the degree of spatial variability of recharge rates that is imposed by the geological heterogeneity that is inherent within alluvial settings.Water isotopes are commonly used for determining the sources of groundwater and residence times.The isotopes of hydrogen ( 2 H and 3 H) and oxygen ( 18 O) have proved to be particularly useful tools in hydrogeological studies, providing valuable insights into water dynamics in a given basin (Dincer et al. 1970;Fontes, 1980;Clark and Fritz, 1997;Hoque and Burgess, 2012).A few local isotopic studies in Punjab and Haryana states have focused on the provenance of groundwater using isotopic tracers, sources of groundwater salinity (Kulkarni et al., 1996;Lorenzen et al., 2012), and estimation of recharge rate based on the tritium tagging technique (Datta et al., 1996;Rangarajan and Athavale, 2000).These studies provide a broad understanding of the mechanism of local recharge sources and zones, but it is hard to extrapolate these local studies to understand basin-scale recharge mechanisms given the heterogeneity of the alluvial aquifer system (Bowen, 1985;Sinha et al. 2013;van Dijk et al., 2016a).There has been no systematic study of recharge sources, zones and groundwater flow (e.g., isotopic fingerprinting or age) at an appropriate scale in this important region, with the important exception of the studies of Lapworth et al. (2015Lapworth et al. ( , 2017) ) and Rao et al. (2017) who focused on a region between the Beas and Sutlej Rivers in Punjab.Comparable work across the rest of Punjab and Haryana has not yet been done.
To investigate the spatial pattern of groundwater recharge sources in the NWIA, we focus here on the Ghaggar River basin that encompasses parts of the states of Punjab, Haryana and Rajasthan in northwestern India.We investigate the sources of groundwater recharge using the stable isotopes of oxygen and hydrogen in water, coupled with measurements of tritium radioisotopes.Our specific objectives are (1) isotopic characterization of groundwater and surface waters for the Ghaggar River basin; (2) identification of the recharge sources and zones; and (3) understanding recharge processes and groundwater flow dynamics in the aquifer system that underlies the Ghaggar River basin.To achieve our objectives, we conducted systematic sampling of precipitation, surface waters from both rivers and canals, and groundwater across the study area (Fig. 1).

Study area
The study area for our investigation encompasses a major area of the NWIA in the Indo-Gangetic basin, focusing on the Ghaggar River basin which lies between 29°10 0 N and 30°54 0 N and 74°20 0 E and 77°26 0 E and covers an area of 22,235 km 2 (Fig. 1).The Ghaggar River originates in the Siwalik foothills of the Himalayas at an altitude of 1927 m asl (above sea level) and flows for $330 km across the study area.In the alluvial part of the basin, the elevation varies from 350 to 150 m asl.The Ghaggar is a seasonal river that is set within a large, slightly incised paleo-valley that is more than 3 km wide (Bhadra et al., 2009;Sinha et al., 2013;van Dijk et al., 2016a;Singh et al., 2017).About 90% of the study area is used for intensive agriculture (UNDP, 1985;Ambast et al., 2006), and the main sources of irrigation water supply are precipitation, groundwater, and surface water from canals.Water from the Sutlej River is diverted through a dense canal network for irrigation in the middle and downstream parts of the basin.Water demand in this region has greatly increased since 1970 (Kumar et al., 2014) due to a progressive shift towards rice cultivation in the monsoon season (July-September).As a result of this, a net annual water deficit of 1.63 million-hectare meters for Punjab (estimated for 2008; Kumar et al., 2014) has to be met through groundwater abstraction.As a consequence, there has been a rapid decline in the water table over much of the region over the last four decades (Rodell et al., 2009;Tiwari et al., 2009;Chen et al., 2014;Long et al., 2016;MacDonald et al., 2016).
The Ghaggar River basin extends across four geomorphic units that coincide with distinct subsurface hydrogeological units: the Siwalik hills, the Sutlej and Yamuna sediment fans, the interfan area, and spatially-restricted aeolian deposits (Fig. 1).The Siwalik Hills range in altitude from 400 to 2000 m asl and are made up of sandstone and conglomerate.They are separated from sediments of the Indo-Gangetic foreland basin by the Himalayan Frontal Thrust (HFT).The foreland basin fill is dominated by fluvial fans deposited by the Sutlej and Yamuna rivers, and comprising spatially heterogeneous sand, silt, and clay deposits associated with abandoned, avulsive river channel belts (UNDP, 1985;Saini et al., 2009;Sinha et al., 2013;van Dijk et al., 2016a;Singh et al., 2017).These fans decrease in surface elevation from about 350 to 150 m asl over a distance of about 300 km.Van Dijk et al. (2016a) used aquifer-thickness logs from the Central Groundwater Board (CGWB) to show that aquifer bodies within the fans are typically composed of stacked channel deposits of fine to medium sand, interspersed both vertically and laterally with non-aquifer silt and clay.The aquifer bodies have a median thickness of about 6-7 m (van Dijk et al., 2016a).These sediments form a multi-layer aquifer system (Lapworth et al., 2017).In contrast, the interfan area between the Sutlej and Yamuna fans, east of Patiala (Fig. 1), is characterized by thinner and less abundant aquifer bodies, even at locations that are adjacent to the HFT (van Dijk et al., 2016a).This geomorphic framework contrasts with the traditional division of the Indo-Gangetic basin into Siwalik Hills and distal alluvial deposits, and reflects an important along-strike stratigraphic heterogeneity in the aquifer system.The bulk specific yield varies spatially, with values of 10-26% in the interfan area near the Himalayan foothills and 5-15% in the Sutlej and Yamuna fans (UNDP, 1985;CGWB, 2009CGWB, , 2012)).
The climatic conditions in the Ghaggar basin range from subtropical to semi-arid, with temperatures of 25-48 °C in summer, and 5-19 °C in winter (Kumar et al., 2014).Precipitation shows marked spatial variation throughout the study area.Mean annual rainfall is ranges between 800-1200 mm in the Siwalik Hills, 74% of which is received during July to September.Mean annual rainfall in the foreland basin is 400-800 mm but decreases to less than 400 mm in the southwestern part of the study area.

Sampling strategy and measurements
We designed our sampling strategy to document spatial and vertical variation in the isotopic signature (d 18 O, d 2 H, and 3 H) of both source water (precipitation, canal water, and rivers) and groundwater across the study area (Fig. 1).Water sampling was carried out throughout the study area during June and July 2013 for d 18 O and d 2 H analysis, and October-November 2012, June 2013, and April and June 2015 for 3 H analysis.We collected groundwater samples from 244 locations from monitoring wells of the CGWB and state groundwater departments, public tube wells, and hand pumps using a grid of about 10 Â 10 km across the study area.The wells were purged for 30-45 min depending on the depth of the well before sampling.The screening depths in the sampled wells vary from $6 to 365 mbgl.It is worth mentioning that screening depths are only recorded for government tube wells and piezometers, while depths for public tube wells and hand pumps are based on local information (tube wells and hand pumps owner).As per the hand pump/tube wells owner, there is generally some displacement (±1 to ±3 m) during the screen installation.So, we have assumed an uncertainty of ±5 m in the screen depths of such wells.
The primary recharge sources of groundwater in the study area are meteoric water, rivers, and irrigation canals.Canal water is typically abstracted from the river network at or near the Himalayan mountain front.To establish the isotopic signatures of these different sources, canal and river water samples were collected from random locations based on accessibility within the study area.Also, three rain gauge stations were set up in Chandigarh, Patiala, and Sirsa (Fig. 1) to develop a local meteoric water line for the Ghaggar River basin.
All water samples were collected in pre-cleaned polypropylene bottles (20 ml for stable isotopic samples and 500 ml for tritium samples).The bottles were rinsed at the site twice with the sample water.To further avoid any diffusive and evaporative losses from the samples bottles, they were tightly sealed and brought immediately to the laboratory for isotopic analysis.Samples for d 18 O and d 2 H were analysed during August-October 2013, and those for 3 H were analysed during March, August, September 2013 (n = 68) and during August-September 2015 (n = 19).Sample latitude, longitude, and altitude were measured using a handheld Global Positioning System receiver during the sampling.Additional parameters such as pH, temperature, and electrical conductivity (EC) were also measured in the field using handheld pH and EC meters.
Stable isotopic (d 2 H, d 18 O) and tritium ( 3 H) analyses were carried out at the Nuclear Hydrology Laboratory at National Institute of Hydrology (NIH) Roorkee, India.Measurements of d 18 O and d 2 H were made using Continuous Flow Isotope Ratio Mass Spectrometry and Dual Inlet Isotope Ratio Mass Spectrometry following standard procedures (Epstein and Mayeda, 1953;Brenninkmeijer and Morrison, 1987).The results are expressed in concentrations per mil (‰) relative to Vienna Standard Mean Ocean Water (VSMOW) on the d scale: where R sample is the 18 O/ 16 O or 2 H/ 1 H ratio of the water samples, and R standard is the corresponding ratio for VSMOW.The overall precision, based on ten repeated measurements of each sample, was less than ±1.0‰ for d 2 H and ±0.1‰ for d 18 O.We also analysed tritium ( 3 H) to place some initial constraints on the age of groundwater samples in the region.Tritium is ideal for the dating of young groundwater (less than 60 years before present) because it is incorporated into water molecules and its activity is not affected by chemical or microbial processes, or by reactions between the groundwater and aquifer material (Stewart and Morgenstern, 2001).Electrolytic reduction was used to concentrate tritium in 500 ml of a water sample using a 20cell standard Tritium Enrichment Unit.The temperature of the sample was maintained at 0°to 5 °C to achieve maximum tritium fractionation and enrichment.After the electrolytic process, the sample volume was reduced to 25 ml.The tritium activity in enriched samples along with the standards was measured using an ultra-low-level liquid scintillation counter (Quantulus Wallac model 1220), and results are reported in tritium units (TU) with 2 sigma errors (Table S2).

Precipitation
Measured d 18 O and d 2 H values of precipitation from three sampling locations (Chandigarh, Patiala, and Sirsa) ranged between À14.8‰ and +5.8‰ for d 18 O, and between À116.3‰ and +51.5‰ for d 2 H (Fig. 2 & S1).The amount-weighted annual precipitation (AWAP) value of d 18 O was À6.5‰ and of d 2 H was À46.8‰.We checked the effect of altitude variation on isotopic values of precipitation from all three stations and this was found to be negligible.Remarkably enriched isotopic values were observed for rainfall events of less than 20 mm per day, particularly from the rain gauge station at Sirsa in the downstream part of the basin where the climate is semi-arid.A cross plot of d 18 O and d 2 H based on the monthly weighted isotopic composition of precipitation from all three sites was used to develop a Local Meteoric Water Line (LMWL) for the study area (Fig. 2).This provides information on the preservation or alteration of the stable isotopic composition of groundwater and various other water sources.The LMWL derived for our study area, along with other LMWLs for the Delhi region and the Global Meteoric Water Line (GMWL), are as follows: ðr 2 ¼ 0:98Þ ½Gourcy etal:; 2005 ð 3Þ The slope of our LMWL (Eq.( 1)) is close to that of the GMWL (Eq.( 3)) defined by Gourcy et al. (2005) and the LMWL from Delhi (Pang et al. 2004) (Eq.( 2)).

River and canal waters
The isotopic values of the Ghaggar River samples (n = 10) varied from À7.5‰ to À6.7‰ for d 18 O and À54.9‰ and À43.8‰ for d 2 H.The d 18 O and d 2 H values of river water samples fall on the LMWL and in close proximity to the AWAP, indicating that modern local precipitation is the primary source of Ghaggar River water (Fig. 3a).In contrast, Sutlej River water ranges from À12.6‰ to À10.5‰ for d 18 O and À87.8‰ to À70.9‰ for d 2 H.The d 18 O of canal water (n = 20) varies from À12.0‰ to À10.6‰ while d 2 H ranges from À81.3‰ to À70.3‰ (Fig. 3a, Table S1).The relative isotopic depletion of canal water compared to the Ghaggar River water and AWAP reflects the Higher Himalayan source of canal water that is derived from the Sutlej River.

Groundwater
The isotopic values of groundwater samples (n = 244) varied between À11.6‰ and À4.7‰ for d 18 O, and À81.4‰ and À34.5‰ for d 2 H (Fig. 3a & b, Table S2).A cross plot of d 18 O and d 2 H of the groundwater samples is used to develop a groundwater regression line for the study area (Fig. 3a, Table S2): The clustering of a large number of groundwater samples around the LMWL indicates that local modern meteoric water is a significant source of groundwater recharge in the study area.However, a cluster of groundwater samples also falls below the GMWL.The slope and intercept of the groundwater regression line (Eq.( 4)) are less than those of the LMWL (Eq.( 1)), indicating the important effect of evaporative enrichment on groundwater (Wassenaar et al., 2011).Importantly, we observe systematic variations in d 18 O values of groundwater with relation to the depth from which the groundwater was sampled (Fig. 4).Groundwater samples taken from depths in general above 80 mbgl show a broad range of d 18 O values (between À11.6‰ and À4.7‰,) and consistently lie very close to the AWAP and canal water samples (Fig. 5c).These samples also show significant spatial variability in d 18 O values (Fig. 5a).In contrast, samples from depths in general below 80 mbgl show a narrower range of values and located closely to the AWAP, varying between À8.2‰ and À4.8‰ for d 18 O and between À58.2‰ and À35.0‰ for d 2 H (Figs. 4, 5b & d).We define the regression lines for groundwater samples taken from both above and below 80 mbgl (Fig. 3a & b) as follows: Both regression lines (Eqs.( 5) & ( 6)) have a lower slope than the GMWL and a negative intercept, suggesting evaporative enrichment of the groundwater.

Deuterium excess
The deuterium excess (d-excess) was defined as d = d 2 H -8 Â d 18 O (Dansgaard, 1964), and quantifies the surplus deuterium about Craig's line (Craig, 1961).The deuterium excess depends on conditions prevalent during primary evaporation, including variation in humidity, ocean surface temperature, and wind speed, and thus gives information on the sources of water vapour (Gat, 1983;Clark and Fritz 1997).While equilibrium processes do not change the d-excess for any of the phases, non-equilibrium evaporation causes a decrease in the d-excess which indicates an increase in the vapour phase.Most groundwater samples have dexcess values close to that of precipitation, and a few samples are similar to the d-excess value of canal water (Fig. 6).The dexcess values of groundwater samples derived from depths in general above 80 mbgl are spatially variable and vary from À4.1‰ to  +14.9‰ (Fig. 8a), which depends upon the mixture of canal recharge or precipitation and groundwater irrigation return flow.In contrast, samples below 80 mbgl have d-excess values in the range of À0.8‰ to +11.3‰, and most samples have values very close to that of precipitation (Fig. 6).The negative and very low d-excess values of a few groundwater samples derived from below 80 m depth mainly correspond to the downstream part of the study area.

Tritium concentrations
Tritium concentration was measured for 91 samples of river, canal, and groundwater in the study area (Fig. 7a, Table S3), and 22 samples of precipitation for Roorkee in north India (Table S4).
The average tritium concentrations of precipitation and canal water samples were 8.2 ± 0.3 and 9.0 ± 0.3 TU, respectively.The tritium concentration in groundwater varies between $0.1 TU (minimum) and 12.9 ± 0.5 TU (maximum).These values vary both vertically (i.e. with depth) and spatially (i.e. with distance from the Himalayan mountain front) (Fig. 7b & c).Tritium concentration shows a clear decline with depth (Fig. 7b), although groundwater samples taken generally above 80 mbgl show a much broader range of values, from 0.3 ± 0.2 up to 12.9 ± 0.5 TU, than samples from below that depth (Fig. 7b & c).Spatially, Tritium concentrations vary from 1.3 ± 0.2 to 8.4 ± 0.4 TU particularly in the area within 80 km of the Himalayan front (Fig. 7c).In the middle and downstream regions of the study area between 80 km and 320 km downstream of the Himalayan front (Fig. 7c), TU values for samples from generally above 80 mbgl lie between 0.3 ± 0.2 to 10.2 ± 0.5.In contrast, tube well samples from generally below 80 mbgl show lower TU values, typically less than 4.4 ± 0.2.We observe three downstream patterns of TU value variation for those deeper samples, (Fig. 7a & c): (1) In the area within 80 km of the Himalayan front, the majority of samples have TU values less than 1.2 ± 0.1 TU; (2) in the middle area between 80 to 200 km downstream, TU values vary from $0.1 to 4.4 ± 0.2 TU; and (3) beyond 200 km downstream, TU values are less than 1.9 ± 0.2 TU (except for 3 measurements).Thus, the deeper samples generally show downstream TU variation compared to groundwater samples derived from in general above 80 mbgl.

Estimation of canal water contribution to groundwater recharge
The stable isotopic composition of groundwater samples from generally above 80 mbgl in a number of tube wells (especially in the middle and downstream parts of the basin) indicate important recharge from highly depleted canal water.We have quantified the proportion of this recharge using a two-component separation approach (Clark and Fritz, 1979;Rai et al., 2009;Engelhardt et al., 2014).For this, we assume that the isotopic composition of groundwater samples is derived by mixing of two end-member components, namely AWAP (average d 18 O value of À6.5‰) and canal water (À11.6‰).The volumetric ratio of canal recharge to total recharge, p, is then given by:  The canal water contribution to groundwater recharge is spatially variable across the study area (Fig. 8b).Most tube wells with depths of up to 80 mbgl in the downstream part of the basin have an estimated canal recharge component of more than 50% of total recharge, whereas this value is quite variable (25% to 50%) in the middle part of the basin.

Spatial variation in isotopic characteristics of groundwater
Our analysis of isotopic characteristic of groundwater in the Ghaggar basin shows that d 18 O and d 2 H of groundwaters vary in space as well as depth.To understand the relationship between isotopic and aquifer characteristics, we have plotted aquifer thickness logs from the basin margin in Himalayan front to the distal part of the basin along cross-section A-B along with d-excess and isotopic composition (d 18 O and d 2 H) of groundwaters (Fig. 9a-e).The d 18 O values of groundwater samples range from À5.2‰ to À7.7‰, and from 1.2 to 8.4 TU for tritium concentration (generally above 80 mbgl) in the upstream parts (from the basin margin up to 80 km downstream, Fig. 7c & 9d); these values clearly reflect that precipitation is the main recharge source.However, in the middle and downstream parts of the basin ($80 to 320 km downstream of the basin margin), the d 18 O value and tritium concentration of groundwater samples generally above 80 mbgl are much more variable than those from the generally below 80 mbgl.This could reflect multiple recharge sources such as irrigation return flow, canal water leakage, and precipitation in this part of the basin.
The cross-plot of d-excess vs. d 18 O for the groundwater samples shows an inverse correlation (Fig. 6).The positive d-excess (>10‰) of the groundwater samples (generally above 80 mbgl) is consistent with source water derived from the Higher Himalayas (Pande et al., 2000;Rai et al., 2014) and provides additional evidence of recharge from Sutlej River water that has been redistributed through the canal network.A few samples have lower d-excess values, which suggests evaporative enrichment of groundwater during the recharge process.The d-excess values of groundwater samples that lie between canal water and AWAP values suggest mixing of groundwater from different sources.These d-excess results are broadly consistent with our d 18 O and d 2 H measurements.Low d-excess values in downstream part (between 280 and 320 km from the northern basin margin) suggest a significant evaporative enrichment of recharged groundwater (Fig. 9c).This is likely due to the prevailing semi-arid climate in a southwest part of the study area and slow rate of recharge (56 mm/yr for Punjab and 70 mm/yr for Haryana; Rangarajan and Athavale, 2000).Towards the distal part of the basin, the thickness of non-aquifer material in the subsurface increases (van Dijk et al., 2016a).The spatial variation of d 18 O and d 2 H with distance suggests that in-situ vertical recharge at a variable rate takes place throughout the study area.This is obvious on account of the variable water table (Supplementary Fig. S2) and heterogeneity in the subsurface in NWIA system (Fig. 9a).The isotopic data is in conformity with the observed geological heterogeneity of the NWIA system and the decrease in bulk aquifer body percentage from proximal to distal parts of the basin, which results in limited lateral connectivity of aquifer bodies (van Dijk et al., 2016a,b).

Vertical variability of d 18 O and tritium
Groundwater d 18 O values fall within a narrow range near the AWAP for samples generally below 80 mbgl (Fig. 4).This is likely due to the attenuation of d 18 O and d 2 H values with depth, as water from different sources moves downward and attains the isotopic signature close to the AWAP generally below 80 mbgl (Fig. 4).Similarly, a decrease in tritium concentration is observed with depth (Fig. 7b).The comparatively low tritium concentrations below 80 mbgl indicate relatively large travel time in the groundwater flow system, resulting in loss of tritium by radioactive decay.As mentioned earlier, precipitation is the main recharge source, and therefore, we compare our results with tritium concentration of precipitation for the northwest India.The eight precipitation samples measured here yielded values of $3 to 17 TU.Kumar et al. (2010) reported tritium concentrations of 6.6 to 17.6 TU for northwestern India, while tritium concentrations in precipitation measured during 2004 to 2010 at Roorkee in north India, varies between 2.6 and 15.5 TU (Table S4).We therefore consider tritium concentrations of groundwater of 2 to 12.9 ± 0.5 TU as indicating recharge from modern water.However, tritium concentrations below 1 TU indicates groundwater that is likely older than 50 yrs (Liu et al., 2014).The presence of a mixture of modern and old water below 80 mbgl indicates that significant recharge from modern sources reaches down to tube well depth (i.e., 320 mbgl) except in a few locations.This, in turn, suggests that there is significant vertical leakage through the thick but discontinuous non-aquifer layers within the sedimentary sequence (van Dijk et al., 2016a) and that these layers do not provide a barrier to vertical recharge of groundwater (Bowen 1985, Toth, 2009;Sinha et al., 2013;van Dijk et al., 2016a;Hoque et al., 2017: Fig. 9b).An attempt was also made to corroborate our results with hydraulic calculations using the velocity figures for sandy aquifers from Soni et al. (2014) and Shekhar et al. (2015) and our results are satisfactory even at such large scale of aquifer system.Similar results have been reported by Lapworth et al. (2015) from the Beas-Sutlej region of Punjab using anthropogenic tracers.

Spatial variability in groundwater recharge sources and zones
The spatial variation of d 18 O values, tritium concentrations and d-excess of groundwater samples generally above 80 mbgl, and the canal contribution to groundwater recharge (Fig. 5a, b, Fig. 7a, Fig. 8a & b) reveal spatially variable recharge processes in the study area.This leads us to identify four distinct groundwater recharge zones (Fig. 8a).
Zone I represents the upstream area of the basin, where the dexcess ranges from +6‰ to +11‰ (Fig. 8a).A cross-plot of d 18 O vs. d 2 H shows that all groundwater samples (generally above 80 mbgl) fall on the LMWL and are very close to AWAP (Fig. 8c) suggesting recharge predominantly through local precipitation.Tritium concentrations in groundwater samples generally above 80 mbgl in this zone are up to 12.9 ± 0.5 TU.This indicates that the groundwater is of modern age, and derived from local meteoric precipitation.Previous work interpreted Zone I as characterized by high permeability and high hydraulic conductivity (van Dijk et al., 2016a).Our work, however, has shown that this zone is recharged by local precipitation based on the inferred groundwater recharge sources and flow paths.Therefore, the implication of our work is that recharge rates in this zone would be high although we do not provide any quantification at this stage.
Zone II represents an area where the d-excess ranges from +1‰ to +5‰ (Fig. 8a) and groundwater samples fall on or below LMWL in the cross plot of d 18 O vs. d 2 H and very close to AWAP (Fig. 8d).The primary source of recharge is local precipitation.However, a number of samples fall between the isotopic values of canal water and AWAP, indicating recharge through a mixture of canal water and local precipitation.The enriched isotopic composition and low d-excess of groundwater samples reflect fractionation during recharge through rainfall or irrigation return flow.Tritium concentrations of groundwater samples range from $2 to $10 TU, indicate modern recharge at depths generally above 80 mbgl.The canal contribution to groundwater in a number of locations, estimated by using two-component analysis, ranges from 0% to 50% in zone II (Fig. 8b).This zone is characterized by a high tube well density and rapid decline in groundwater levels due to overexploitation of groundwater resources.Therefore, induced recharge to groundwater through irrigation return flow might be one explanation for evaporative enrichment of isotopic signatures of groundwater.
Zone III represents the area where the d-excess ranges from +8‰ to +15‰ (Fig. 8a), and where d 18 O and d 2 H of the groundwater fall mostly above the LMWL and very close to the isotopic composition of canal water.Relatively more depleted values of d 18 O are observed in this zone compared with zones I and II, which we interpret as evidence of active recharge by canal water.Tritium values are also higher than those in zone II, because of the significant contribution of the canal to groundwater.The canal contribution to groundwater recharge ranges up to 100% (Fig. 8b).
Zone IV represents the distal end of the study area where dexcess value ranges from À2‰ to +4‰, and d 18 O and d 2 H values fall below the LMWL but close to AWAP (Fig. 8f).The d 18 O and d 2 H values of groundwater samples, more enriched than the AWAP, are evidence of recharge by local precipitation, modified by evaporative fractionation.This zone is characterized by semi-arid conditions.Here, water gets fractionated during the recharge process and bears enriched values of d 18 O and d 2 H and low d-excess.The canal contribution to groundwater at a number of locations, is less than 50% (Fig. 8b).This is consistent with slow recharge rates in this zone (Rangarajan and Athavale, 2000).

Conceptual model of the hydrological processes controlling isotopic signatures of groundwater
It has already been established that the hydrological processes operating in the basin have subtle differences as we move downstream.Traced downstream, the variability in the recharge processes including the source and rate of recharge, appears to correlate with both the complex and spatially-variable subsurface sedimentary architecture (van Dijk et al., 2016a) and with anthropogenic forcing such as the canal network.We use this information to develop a conceptual model for the hydrological processes operating in the Ghaggar Basin (Fig. 10).The boundary conditions for the conceptual model are: a) geomorphic setting, (b) the spatial variability of d 18 O and d 2 H , (c) the spatial variation of tritium concentration in groundwater, (d) longitudinal cross-sections based on aquifer-thickness logs, and (e) groundwater flow directions derived from water table.
The available aquifer-thickness data, indicate that the aquifer system is composed of stacked, laterally-discontinuous highpermeability aquifer bodies, separated both laterally and vertically by low-permeability non-aquifer material (Fig. 9a).This interleaved pattern of occurrence, coupled with the basin surface topography, imposes and maintains a hierarchy of groundwater flow systems, from shallow to deep (Hoque et al., 2017).The flow systems can be referred as local, if recharge and discharge areas are contiguous; intermediate, if these regions are separated by one or more local systems; and regional: if they extend over the full extent of the basin (Toth, 2009).Our depth-wise isotopic data suggest that local flow patterns in the Ghaggar Basin extend down to a depth of c. 80 m and are dominated by local vertical recharge from meteoric and canal sources, while intermediate and regional flow patterns extend to deeper depths and show evidence for sustained lateral flow (Fig. 10).A similar flow pattern has been proposed for the Bengal basin by Ravenscroft et al. (2005), andToth (2009) showed similar hierarchically nested groundwater flow systems within the exploited depth of the aquifer.The observed vertical variations in isotopic signature of the groundwater are thus likely due to tortuous groundwater flow paths; surface topography and subsurface lithological variations impose hydraulic heterogeneity and anisotropy, which controls the flow patterns and depth of groundwater penetration (Zijl, 1999).
This conceptual model helps to explain the major lateral and vertical variations in groundwater d 18 O, d 2 H and tritium concentrations, which we interpret as due to the heterogeneous distribution of aquifer and non-aquifer sediments in the subsurface.For example, the tritium results show comparatively higher residence time for groundwater from generally below 80 mbgl compared to the groundwater above 80 mbgl.This indicates the existence of longer flow paths for groundwater below 80 mbgl, which is likely controlled by the aquifer heterogeneity and anisotropy.
Furthermore, as illustrated in the schematic in Fig. 10, water recharged in zones I and II must travel deeper to account for the regional groundwater flow system.The groundwater velocity through sand-rich aquifer bodies like those in the Ghaggar Basin are typically $1 to 2 m/day (Soni et al., 2014;Shekhar et al., 2015).Furthermore, the model (Fig. 10) suggests the hydrological process of mixing of this old groundwater in zone III with vertically recharged groundwater from sources like rainfall and return flow of canal and groundwater irrigation.

Conclusions
This study is the first systematic attempt to identify the recharge sources and zones and to characterise groundwater dynamics in the northwestern Indian aquifer in Punjab and Haryana using water isotopes (d 18 O, d 2 H and 3 H).Our results reveal that recharge sources for groundwater in the region are local meteoric water, canal seepage, and both canal and groundwater-derived irrigation return flow.These recharge sources vary spatially in their importance.The influence of canal water to recharge is most apparent in the Zone II to IV in the study area.The spatial variability in d 18 O, d 2 H value and tritium concentration reflects limited lateral connectivity of the aquifer due to the heterogeneity of aquifer material.However, the variation of d 18 O with depth shows the effect of averaging of the isotopic composition of different groundwater sources at various depths.Variability in tritium concentrations with depth reveals that groundwater generally above 80 mbgl is of modern age.However, groundwater generally below 80 mbgl is a mixture of modern and old water.The spatial distribution of aquifer and non-aquifer material in the subsurface likely forces some recharge water to follow longer, deeper, and more tortuous flow paths resulting in low tritium values.Two kinds of flow Further, recharged water travel deeper and join the regional groundwater flow system in zones I and II.The depleted isotopic value in zone III indicates active recharge source from higher altitude and mixing between groundwater and surface water.The enriched isotopic value in zone IV (located in semi-arid condition where rainfall amount is less than 400 mm/year) indicates evaporative enrichment due to the less aquifer material in the distal margin of Sutlej and Yamuna sediments fans in northwest Indian aquifer system.
patterns are observed in the northwest Indian aquifer system: (i) a local flow pattern down to a general depth of c. 80 mbgl, and (ii) an intermediate and regional groundwater flow pattern in aquifer system generally below 80 mbgl, recharged with water from lateral groundwater flow and vertical infiltration.Results from this study can help in the identification of active recharge zones and for designing sustainable groundwater management strategies.

Fig. 1 .
Fig. 1.Geomorphic map of the study area, modified after van Dijk et al. (2016a).The continuous blue lines represents surface rivers.The continuous orange lines represent the canal network.Three rain gauges were installed at Sirsa, Patiala and Chandigarh (green squares) to collect precipitation samples.Canal water sampling locations are shown by red triangles, river water samples by yellow squares, and groundwater samples by grey circles.Basin margin sampling point is shown as a black square.(For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this article.)

Fig. 2 .
Fig. 2. Cross plot of d 18 O and d 2 H for precipitation samples (grey open squares); solid black line shows the LMWL and grey solid line shows the GMWL, big triangle shows the value of Amount Weighted Annual Precipitation (AWAP).

Fig. 3 .
Fig. 3. (a) Cross plot of d 18 O and d 2 H for groundwater samples (generally above 80 mbgl shown by purple circles, and generally below 80 mbgl shown by orange circles) along with canal samples (red triangles), Ghaggar River samples (yellow squares), and Sutlej River samples (blue squares).Black continuous line represents the local meteoric water line (LMWL), grey continuous line shows the global meteoric water line (GMWL), and black dashed line is the groundwater (GW) regression line.(b) Cross plot of d 18 O and d 2 H for groundwater samples generally above and below 80 mbgl, indicating different recharge conditions in the Ghaggar basin in northwestern Indian aquifer system.(For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this article.)

Fig. 4 .
Fig. 4. d 18 O variation with sampling depth for all groundwater samples within the study area.The distinction between samples taken generally above 80 mbgl are wider in range, which shows the multiple sources of recharge up to the depth of 80 m compared to samples below that depth.

Fig. 5 .
Fig. 5. (a) and (b) show spatial variation of d 18 O values for groundwater samples taken generally above and below 80 mbgl, respectively.The continuous blue lines represent river network.Basin margin point is shown by black square.The Ghaggar basin boundary shown by red continuous line, and the administrative boundary between Punjab and Haryana is shown by the grey continuous line.Panels (c) and (d) show cross plot of d 18 O and d 2 H for groundwater samples from two different depth zones -above 80 mbgl (purple circles) and below 80 mbgl (orange circles).Red triangles represent canal water samples, yellow squares are Ghaggar River water samples, and blue squares are Sutlej River water samples.Black continuous line is the LMWL and the grey continuous line is GMWL.(For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this article.)

Fig. 6 .
Fig. 6.Cross plot of d 18 O values and d-excess.Groundwater samples are represented by purple circles (above 80 mbgl) and orange circles (below 80 mbgl).Red triangles represent canal water samples, yellow squares are Ghaggar River water samples, and blue squares are Sutlej River water samples.(For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this article.)

Fig. 7 .
Fig. 7. (a) Spatial distribution of tritium concentrations in groundwater samples (generally above and below 80 mbgl) in the Ghaggar River basin.The continuous blue lines represent river network.Basin margin point is shown by black square.The Ghaggar basin boundary shown by red continuous line, and the administrative boundary between Punjab and Haryana is shown by the grey continuous line.(b) Variation of tritium concentrations with depth of groundwater samples represented by purple circles (above 80 mbgl) and orange circles (below 80 mbgl), and (c) Tritium concentration as a function of distance from the basin margin from the Himalayan front.(For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this article.)

Fig. 8 .
Fig. 8. (a) Spatial variation of d-excess of groundwater samples generally above 80 mbgl.There are four distinct zones, I to IV, shown by black dashed line, indicating different recharge sources for the study area.(b) Spatial variation of canal contribution to the groundwater recharge in the study area derived from the two-component mixing model.Panels (c)-(f) show cross plots of d 2 H and d 18 O of groundwater samples generally above 80 mbgl (purple circles), where d-excess varies between 6‰ and 11‰ in zone I, between 1‰ and 5‰ in zone II, between 8‰ and 15‰ in zone III, and between À2‰ and 4‰ in zone IV. (For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this article.)

Fig. 9 .
Fig. 9. (a) Cross section A-B (see Fig. 1 for location) showing Central Groundwater Board aquifer-thickness logs from the study area, modified from van Dijk et al. (2016a).Depth ranges with non-aquifer material are colored green and ranges with aquifer material are colored yellow.Note that aquifer bodies cannot be correlated laterally with the available data; median log spacing is 7 km.(b) the geographic location of point A and B, which is used to prepare cross section A-B, basin margin point in black color filledsquare and Ghaggar River basin.(c) Downstream variation of d-excess with distance from the basin margin from the Himalayan front.(d) Downstream variation of d 18 O values from the basin margin from the Himalayan front.(e) Downstream variation of d 2 H values from the basin margin from the Himalayan front.(For interpretation of the references to colour in this figure caption, the reader is referred to the web version of this article.)

Fig. 10 .
Fig. 10.Schematic model of the hydrological processes in the groundwater system of the Ghaggar River basin in northwestern Indian Aquifer system.The Sutlej and Yamuna sediments fans indicate thinning of the aquifer material from proximal to distal margin.The schematic model is categorised into four zones (Fig.8a), zone I and II are very close to the proximal part of the basin where rainfall amount varies from 800 to 1200 mm/year, indicating precipitation is the main recharge source in both zones I and II.Further, recharged water travel deeper and join the regional groundwater flow system in zones I and II.The depleted isotopic value in zone III indicates active recharge source from higher altitude and mixing between groundwater and surface water.The enriched isotopic value in zone IV (located in semi-arid condition where rainfall amount is less than 400 mm/year) indicates evaporative enrichment due to the less aquifer material in the distal margin of Sutlej and Yamuna sediments fans in northwest Indian aquifer system.