Assessing the role of groundwater recharge from tanks in crystalline bedrock aquifers in Karnataka, India, using hydrochemical tracers

India, by analysing the isotopic and hydrochemical composition of surface waters and groundwaters, combined with groundwater level observations. The results indicate that tanks have limited impact on regional groundwater recharge and quality in rural areas, where recharge from precipitation and groundwater recycling from irrigation dominate the recharge signal. In the urban setting (Bengaluru), impermeable surfaces increased the relative effect of recharge from point sources such as tanks and rivers, but where present, pipe leakage from public-water-supply accounted for the majority of recharge. Shallow groundwater levels in the inner parts of the city may lead to groundwater discharge to tanks, particularly in the dry season. We conclude that the importance of aquifer recharge from tanks is limited compared to other recharge sources and highly dependent on the specific setting. Additional studies to quantify tank recharge and revisions to the current guidelines for national groundwater recharge estimations, using a less generalised approach, are recommended to avoid over-estimating the role tanks play in groundwater recharge.


Introduction
It is estimated that nearly 700 million Indians living in villages (50% of India's population) depend on groundwater for basic needs (Kulkarni et al., 2015). Groundwater is also the primary water source for over 50% of India's irrigated agricultural land (GoI, 2015). As such, it is a vital resource both for public health and economic growth (Kulkarni et al., 2015;Pahuja et al., 2010;Shah et al., 2003). The Indian Government reports that, on a national scale, groundwater abstraction (of which 90% is used for irrigation) was 249 billion cubic metres (BCM) for the year 2017 -approximately two thirds of the estimated annual net extractable groundwater of 393 BCM (CGWB, 2019a). The scale of groundwater abstraction differs across the country, and since the 1980s, groundwater levels have declined in several regions as a result of increased abstraction due to enhanced and more affordable pump technology. Nearly a third of the groundwater assessment units in peninsular India are therefore in 'semi-critical' (14%), 'critical' (5%), or 'over-exploited' (17%) status (CGWB, 2019a), with strongest declines observed in the sedimentary basins in the northwestern states (Ambast et al., 2006;CGWB, 2019a;MacDonald et al., 2016;Rodell et al., 2009) and in parts of southern India's crystalline basement aquifers (e.g. in the states of Karnataka and Tamil Nadu), which cover about 60% of the country (CGWB, 2019a;Sharma et al., 2018).
The importance of enhanced groundwater management in India has clearly been recognized by the Government of India, acknowledging in the Third National Water Policy document (GoI, 2012b) that water governance needs to be improved to alleviate "a critical situation in many parts of the country". One of the key suggestions is to enhance water availability by using 'tanks' -the local term for typically small, surface water storage structures, such as farm ponds, but also larger lakesfor water harvesting and managed aquifer recharge (MAR) (CGWB, 2007;CGWB, 2013). The use of tanks is a long-standing tradition especially in southern India since ancient and medieval times (Dixit et al., 1993), and they are said to provide flood protection, water storage, ecosystem services, as well as recreational and aesthetic value and potential recharge to local groundwater (D'Souza and Nagendra, 2011;Dixit et al., 1993;EMPRI, 2018;Esha, 2008;Nagendra and Ostrom, 2014;Patil, 2011). Though it is difficult to know the precise number of functional tanks in India (Pant and Verma, 2010), it is estimated that there are between 200,000 (Palanisami et al., 2010;Sharma, 2003) and 500,000 (Reddy, 2015). Recent tank restoration and promotion efforts, with a goal to enhance groundwater availability by means of MAR, have been undertaken by the national and local governments (e.g. Tamil Nadu and Karnataka, often supported by local NGOs), and also received substantial support by international institutions such as the European Union (EU), the National Bank for Agricultural and Rural Development (NABARD), the World Bank, and the International Water Management Institute (IWMI) (Richard-Ferroudji et al., 2018).
However, there are few studies that directly evaluate recharge from tanks or investigate tank-groundwater connectivity in India's basement setting, and therefore, little is known about the effectiveness and potential groundwater quality implications of tank restoration and promotion. In the suggested methodology for groundwater recharge estimations used for national groundwater resource assessment and planning, the Indian Government's Groundwater Resource Estimation Committee (GEC) acknowledges the limited availability of field studies. As a temporary measure, they suggest a recharge estimation method using a fixed recharge rate of 1.4 mm/day/tank-area as approximation (GEC, 1997). This approach was carried forward into the most recent GEC guideline (GEC, 2017) and the latest national groundwater review (CGWB, 2019b). However, while recharge from tanks in the north Indian alluvial plains is relatively high and homogeneous (Bhanja et al., 2019), recharge to the lower-permeability crystalline aquifers dominant in southern Peninsular India is more complicated and likely much more spatially variable. Crystalline basement aquifers are formed by a combination of weathering and tectonic forces and can be best conceptualized as zones of varying degree of weathering, i.e. a weathered upper zone termed 'saprolite', which overlies a less weathered, but fractured rock layer, followed by fresher basement rock (Collins et al., 2020;Singhal and Gupta, 2010). In the southern Indian context, the thickness of the weathered zone is typically a few to several tens of meters, with fracture zones extending to depths of 200 to 300 m (CGWB, 2012). Below the relatively permeable weathered zone, recharge to the crystalline basement depends on the size (aperture) and distribution (connectivity) of fractures. Lateral connectivity varies spatially with depth and depends on the vertical extent of the weathered zone and layout of the fracture network at depth (Banks et al., 2009;Guihéneuf et al., 2014;Ofterdinger et al., 2019). The connectivity may also depend on the type of fissure network, for example on whether the bedrock is fractured or sheared.
Few studies report the local impact of tanks in these crystalline basement settings in India. Investigating recharge from mostly small tanks with < 0.4 km 2 zone of influence ('command area') in Andhra Pradesh, Reddy and Behera (2009) found farmers reported shallower groundwater levels and fewer dry wells in areas with restored tanks than in areas without tanks. A study using macro level data from West Bengal to model the linkage between groundwater levels and tanks found an inverse relationship between tank density and depth to groundwater (Chowdhury and Behera, 2018). However, in both of these studies, it is not clear if the difference in groundwater level can be attributed to recharge by tanks, or is influenced by other factors, such as decreased groundwater abstractions in areas with higher surface water availability (i.e. tanks). Other studies, mainly using water balance approaches, have quantified average daily percolations rates of 5.5-18.4 mm/day Perrin et al., 2012;Singh et al., 2004), and percolation efficiencies, defined as the percolated fraction of tank water depletion over a given time period, of 35-68% Massuel et al., 2014;Metha and Jain, 1997;Perrin et al., 2012;Singh et al., 2004;Sukhija et al., 1997). However, little is known about how much of the infiltrated water actually recharges groundwater, as not all water lost through the base of a tank can be expected to reach the groundwater table, but some may be stored in the vadose zone or taken up by evapotranspiration (De Vries and Simmers, 2002). This is in particular discussed in a study by Boisson et al. (2014), which proposed a combined surface water and groundwater balance method in order to better consider flow within the unsaturated zone and found that only 53-88% of the potential recharge from the tank percolate reach the groundwater table. In summary, while some studies have assessed groundwater recharge from tanks in the Indian crystalline basement using a variety of approaches, most have focussed on quantifying the percolation rates of the tanks. There remains significant uncertainty on how much of this infiltrated water is recharging to deeper groundwater and how much of the water is lost in the unsaturated zone. Therefore, more site-specific studies are necessary to understand the effects of tanks on groundwater recharge in the crystalline basement.
In this study, a suite of direct measurements, i.e. water isotopic ratios, hydrochemistry and water level observations were used to investigate the degree of groundwater recharge and its impact on water quality from tank structures in three study areas in Karnataka State. The areas reflect agricultural land use and different geological settings (fractured versus sheared crystalline basement). Because tanks are traditionally part of both the urban and rural landscape in southern India, we also considered the city of Bengaluruwell known for its numerous interconnected tanksin this study. Specific aims were to: 1) Assess the connectivity of the deeper (fractured) basement aquifer to modern recharge processes from tanks and other sources; 2) assess the significance of tank recharge compared to other recharge sources, and; 3) identify the contamination risk posed by tanks and other recharge sources to groundwater.

Study locations
This study focuses on three areas in south and southeastern Karnataka. There are seven perennial river basins in Karnataka, but due to the high inter-annual variability of monsoon rainfalls and river flow, and high vulnerability of local surface water sources to contamination, groundwater is used as primary drinking water source in most of the state's rural area (EMPRI, 2018). Groundwater resources in Karnataka are also heavily used for industrial and domestic water supply in major cities such as Bengaluru (population about 8 million) and Mysore (population about 0.9 million), and for agricultural production across the state. Recent (2017) estimates of groundwater withdrawals are 70% of total extractable groundwater (compared to 63% on the national scale), and about 26% of Karnataka's 'taluks' (administrative divisions) are classified as 'over-exploited' with respect to groundwater use, which is 9% higher than on the national scale (CGWB, 2019b).
The selected sites for this study are the city of Bengaluru, two subcatchments (Aralamallige and Hadonahalli) in the rural area of the Thippagondanahalli (TG Halli) catchment about 30 km north of Bengaluru, and the rural Berambadi catchment in the southern part of Karnataka (Fig. 1). All study areas are located in the Cauvery River Basin, with exception of the north-eastern part of Bengaluru, which falls into the South Pennar basin. Table 1 lists key characteristics, such as size, tank density, and climatology of the study areas. Mean annual precipitation in the areas ranges 800-900 mm/year, but with high annual variations (Fig. 1E), and its pattern is composed of two distinct monsoon periods, namely the dominant southwest summer monsoon (June-September) that typically produces most of the annual Areas of low/high percentage of household water supply from piped mains water are indicated in Bengaluru (B), highlighting the high percentage of piped water supply (imported mainly from the Cauvery River) in the inner city. Where samples were taken nearby larger surface water features in Bengaluru, the names of these features are indicated by letter, and fully stated in the legend. Data sources: % piped water coverage in Bengaluru based on (GoI, 2012a), rainfall data provided by the Kabini CZO and ATREE (see acknowledgement), elevation data from USGS SRTM data (http://earthexplorer.usgs.gov). precipitation (~85%), and the northeast winter monsoon from October to December (Rahul et al., 2016). Mean temperatures are around 23 • C, and potential evapotranspiration usually exceeds the annual rainfall with mean values of approximately 1100-1650 mm/year across the study areas (Sekhar et al., 2016;Srinivasan et al., 2015).
Numerous long-established water tanks are characteristic of all three study areas (see Table 1) with region-specific distribution in size and setup. Tanks can be divided into the following broad categories; river-fed tanks constructed by using dams (locally termed anicuts) to check water along a hydraulic gradient, or rain-fed tanks constructed by digging a square cavity of shallow (few metres) depth into the ground (EMPRI, 2018). Tank embankments can be site-specific, but typically, river-fed tanks have little embankment aside from the check dam, while smaller, rain-fed tanks are usually without raised embankments.
Bengaluru has mostly cascading, large (> 1.2 ha), medium (0.4-1.2 ha), and small size (< 0.4 ha) tanks with seasonal to perennial water storage. The city is naturally divided into three main river valleys, and most of the city's tanks have been established by the use of anicuts to intercept rainwater and dam existing streams along topographic gradients; often with drains connecting a series of cascading tanks to each other (D'Souza and Nagendra, 2011). Nowadays, the main tank inflows are from rainfall, and urban runoff, including waste waters (Rao et al., 2020). The depth of water bodies in Bengaluru ranges from 0.5 m to a maximum of 9 m (Sankay Tank). The TG Halli catchment is dominated by two large reservoirs, namely TG Halli reservoir itself (in the most southern part of the catchment) and Hesaraghatta Tank, which is located about 30 km upstream from TG Halli along the Arkavathy River (a tributary to the Cauvery River). In addition, there are an estimate of about 600 smaller, often cascading in-stream tanks, of which most have been constructed by checking monsoon runoff (Srinivasan et al., 2015). In the Berambadi catchment, most of its 53 tanksother than the Berambadi tankare ephemeral and only fill up for short periods after strong monsoon rains (Collins et al., 2020;Sharma et al., 2018). There is no detailed bathymetry available for small tanks in TG Halli and Berambadi, but the depth of these smaller ponds observed in similar studies ranges around 4 m , Massuel et al., 2014, Sharda et al., 2006.

Hydrogeology
Previous hydrogeological studies have been undertaken in all three areas: Sekhar et al. (2018) investigated groundwater level dynamics in Bengaluru for the period 2015-2017. At TG Halli catchment, Ballukraya and Srinivasan (2019) investigated variations in groundwater levels in the two sub-catchments Aralamallige and Hadonahalli (about 7 km and 20 km north of Hesaraghatta Tank, respectively), which corresponds to the areas used in this study. The Berambadi catchment is the site of the Kabini Critical Zone Observatory (Sekhar et al., 2016).
The three study areas are dominated by weathered and fractured bedrock aquifers of the Banded Gneiss Complex, with local variations in rock-composition. Bengaluru is mainly underlain by Precambrian migmatites, granodiorites, tonalites and gneiss, with local granitic intrusions, especially in the central part of the city (near Lalbagh Lake). The weathered horizon is about 15 m in thickness and overlain with red loamy to gravelly superficial deposits of varying thickness (Hedge and Chandra, 2012). A particularly thick saprolite formation of about 60 m can be found in the eastern parts of Bengaluru (Hedge and Subhash Chandra, 2014). Groundwater levels in Bengaluru are observed from extremely shallow depths of only 1 m below ground level (bgl) to about 70 m bgl . The nearby TG Halli catchment likewise consists of weathered (to a depth of about 20 m bgl), fissured and fractured granite and gneiss (Srinivasan et al., 2015). Boreholes are often deep (200-300 m), but may recharge from water-bearing joints at shallow depths as well as from deeper fractures, thus creating a complicated flow system with many of the water bearing fractures at shallow depths. An extensive study in TG Halli from borehole scans and groundwater levels of 83 boreholes revealed that groundwater levels vary considerably between 28 and 222 m bgl, often with large differences between adjacent (proximities of < 100 m) boreholes (Ballukraya and Srinivasan, 2019). The Berambadi catchment, 200 km southwest of Bengaluru, is dominated by massive to extensively sheared gneiss, overlain by about 1 to 5 m of regolith (Collins et al., 2020), which is a thinner weathered zone than in the other catchments. Groundwater levels in Berambadi have been monitored since 2010, and the observed depths to groundwater range from 10 to 70 m (Sekhar et al., 2016).

Water and land use
Water demand in Bengaluru has risen in recent decades and cannot be met using local water resources. To overcome acute water shortages, additional water is drawn from the Cauvery River, and imported to the city via a large pipeline. River Cauvery water is now the city's main drinking water source, particularly in the centre of Bengaluru where the density of the piped mains network is the greatest. Additionally, numerous boreholes have been drilled after the onset of more advanced and affordable pump technology in the 1980s to increase access to groundwater across Bengaluru. In the outskirts of the city, these are often used for drinking water as well as for small-scale agricultural activities. In the more rural study area of TG Halli, groundwater abstraction has similarly increased in recent decades, and is likely one of the causes of strong decline in surface water flows (Srinivasan et al., 2015). In the Berambadi catchment, groundwater is the primary source of water for most water uses. Increasing groundwater abstraction for agricultural production has led to water table declines of up to 50 m (Sharma et al., 2018). However, the recent strong monsoon of 2018 has resulted in significant recoveries in groundwater levels across the catchment (Collins et al., 2020). Agricultural activities in the rural catchments are intensive, with 2-3 cropping seasons a year, namely in summer (January to May, always irrigated), kharif (May to September, frequently irrigated), and rabi (September to January, only irrigated if rainfalls are insufficient) (Sekhar et al., 2016). The irrigation pattern means that the highest groundwater abstraction for the summer irrigation coincides with the period of deepest groundwater levels.

Sampling and analysis
Repeat samples (total n = 112) were collected from surface water (piped mains water, river water, and tank water) and boreholes from shallow (a few metres) to about 300 m depth (Table 2) during two sampling periods, reflecting post-monsoonal (October 2017) and pre- monsoonal (March-May 2018) conditions. Groundwater samples were collected from boreholes (and in two locations from open wells), and surface waterwhere possiblefrom nearby tanks, and in few cases from rivers (Vrishabhavathi River in Bengaluru) or canals. Additionally, 14 one-off samples (groundwater and surface water) were taken at different locations (numbers indicated in parentheses in Table 2). The surface water samples were obtained as grab samples (using a sampling beaker attached to a rod) at about 10-20 cm depth below the water surface and about 2-3 m off the river/tank edge. Samplers were faced upstream where applicable, and there was no disturbance to the bottom sediment immediately prior, or during sampling. Care was taken to avoid tank locations that appeared stagnant (based on visual evaluation), and sampling was undertaken near outflows where possible. No detailed information is however available on the potential degree of stratification within the surface water bodies. The samples for inorganic chemistry were filtered (0.45 μm cellulose membrane) into two 30 mL Nalgene™ bottles without airspace, and un-acidified, but prior to analysis, cation samples were acidified and preserved with Aristar™ grade concentrated nitric and hydrochloric acid (0.5% v/v). Water chemistry samples were stored at 4 • C, except during transportation to the UK (24 h at ambient room temperature). Bulk rainwater samples (n = 51) were collected for isotope analysis (δ 2 H and δ 18 O) in two of the study areas (TG Halli and Berambadi) for the one-year period September 2017 to September 2018 using a specifically designed totalizer that minimizes evaporation effects (based on IAEA/GNIP, 2014). Sampling intervals of the rainwater varied slightly depending on rainfall events, but were approximately every two weeks. Because of the close proximity (about 30 km) and similar altitude and climatology of TG Halli to Bengaluru, rainfall samples from TG Halli are considered as representative for the Bengaluru region as a whole in this study. Unfortunately, one sample for the period 11/11/2017 to 29/12/ 2017 (constituting about 10% of the annual rainfall) from Berambadi was lost in transport before analysis.
In each sampling round, dissolved oxygen (DO), redox potential (Eh), temperature, pH, and specific electrical conductivity (SEC), were measured in the field where possible in a flow cell, with bicarbonate alkalinity (HCO 3 ) concurrently determined by micro-titration. All surface waters and groundwaters were analysed for inorganic (major, minor, trace) elements using ICP-MS (Agilent 8900) and Ion chromatography (Dionex ICS5000), and for stable isotopes (δ 2 H and δ 18 O) using an isotope ratio mass spectrometer (VG-Optima) at the Centre for Environmental Geochemistry (CEG), UK. The δ 2 H and δ 18 O ratios were reported in reference to the Vienna Standard Mean Ocean Water (VSMOW), with measurement precision within ± 0.05‰ for δ 18 O and ± 1‰ for δ 2 H. Deuterium excess (d-excess) was calculated using d-excess = δ 2 H -8 * δ 18 O according to Dansgaard (1964). Results were visualised using R version 3.6.1 (The R Foundation) and the ggplot2 package. Spatial plots were undertaken using ARCGIS®.
In addition to the hydrochemical investigation, pressure transducers were installed at six locations in Bengaluru and in one location in Berambadi to record water level variations at high frequency (15-minute intervals). Additional manual water level measurements (dip meter) were taken manually at approximately monthly intervals for five of the sampling boreholes in Berambadi and five boreholes in the southern subcatchment of TG Halli (see Fig. 1 B for detailed locations of all monitored sites, and Supplementary Material Table S1 for additional information on borehole elevations and depths). More comprehensive water level observations for the same period from the Kabini observatory (Berambadi) are available in Collins et al. (2020).

Stable isotopes
Results of the stable isotope analysis (including one-off samples) are presented in Fig. 2 for rainfall samples (Panel A) and groundwaters and surface water by sampling round (Panel B-C), along with the amount weighted mean rainfall composition for TG Halli (also used for Bengaluru given the proximity of ~ 30 km) and Berambadi. Tabularized data, such as mean values and standard deviations for δ 18 O, δ 2 H, and d-excess of surface waters and groundwaters by study area and sampling round are provided in Table S2 in the Supplementary Material.
The δ 18 O ratio of the rainfall samples (n = 51) ranged between − 9.3‰ and + 1.6‰ (median − 5.3‰), and that of δ 2 H between − 65.9‰ and + 22.3‰ (median − 28.5‰). The amount weighted average for the TG Halli rainfall samples was − 5.2‰ for δ 18 O and − 30.3‰ for δ 2 H, and the (incomplete, i.e. missing data from one sample) amount weighted average in Berambadi was − 3.7‰ for δ 18 O and − 20.0‰ for δ 2 H. The overall distribution of δ 18 O and δ 2 H for the rainfall samples is similar to the local meteoric water line (δ 2 H = 11.74 δ 2 H + 7.82) established by Kumar et al. (2010) in an earlier study in Karnataka and the global meteoric waterline (GMWL with δ 2 H = 8 δ 2 H + 10) (Fig. 2, note that the local and global meteoric waterlines are nearly identical).
Surface water samples in Bengaluru (from piped mains, Vrishabhavathi River, and tanks) have similar isotopic compositions to each other in the post-monsoon season (reflecting wet conditions), but there is a slightly higher evaporative signal in the tanks compared to the piped mains and river water in the drier pre-monsoon sampling. This is indicated for example visually by the slightly more positive δ 18 O (Fig. 2C) of the Bengaluru tank samples, and the lower pre-monsoon median dexcess of the Bengaluru tanks (-8.3‰) compared to the piped and river water (0.7 and 2.9‰, respectively, see Table S2 and Figure S1 in the Supplementary Material). Surface waters in the rural catchments have slightly isotopically lighter composition than groundwaters in the postmonsoon (Fig. 2B), and a stronger, but not significant (Wilcoxon rank sum test) pre-monsoonal enrichment in heavier isotopes (median dexcess of − 28.7 and -23.9‰ for TG Halli and Berambadi, respectively, compared to only − 8.3‰ in Bengaluru, Figure S1, Table S1). Groundwater isotopic compositions show a marked difference in spread between the urban and the rural catchments, with a larger variation Table 2 Overview of number of sites sampled in both sampling campaigns and additional one-off samples (indicated in parenthesis) taken from surface water and groundwater in each round, and total number of samples collected. In addition to the listed samples, a total of 51 rainwater samples were taken in weekly to monthly intervals (35 in TG Halli and 16 in Berambadi). * Only one of the three piped water samples in each round was collected from precisely the same location. However, as the piped water composition is presumed to be independent from the sampling location, all three samples in each round are considered in the repeat-dataset.
between sample locations in Bengaluru compared to a very homogenous compositions in the rural catchments TG Halli and Berambadi (Fig. 2), plotting close to the amount average weighted rainfall. This is reflected for example by the higher (three-fold) standard deviation for δ 18 O in Bengaluru of ± 0.9‰ compared to ± 0.3‰ for TG Halli and Berambadi (Table S1, Supplementary Material). The observed variations in Bengaluru seemed to be slightly greater in shallower boreholes, but there is no significant correlation between isotopic composition and borehole depth at either study area (Fig. 3, R = -0.27 for δ 18 O versus borehole depth). Closer examination of groundwater samples from Bengaluru seems to suggest that samples fall into two separate groups (with distinct behaviour, namely an 'inner-city group' (sampling sites near Lalbagh, IISc, and Sankay tank, Figure 4 right column), which consists of groundwater samples that are more isotopically enriched (i.e. more evaporated) and have similar isotopic ratios to piped and tank waters, and a second group of groundwater samples from the outskirt areas (Allalasandra tank, Vrishabhavathi, and Kaikondrahalli tank, Figure 4 Fig. 1). Piped mains water and the amount weighted mean rainfall is shown for both inner city and outskirts. The local meteoric water line (LMWL) is added in grey.

Water types and major ion chemistry
A piper plot (Fig. 5) of all analysed samples by study area and source type demonstrates that the rural catchments are almost entirely dominated by alkaline earth elements (Ca, Mg), while some samples in the urban setting (Bengaluru) are dominated by Na. The dominant water type in Bengaluru is Ca-Na-HCO 3 , with exception of piped water (Ca-Mg-HCO 3 type) and eight samples with Cl as dominant anion (most of these near Kaikondrahalli tank). The rural catchments (TG Halli and Berambadi) exhibit a more diverse range of water types, with higher proportions of Mg both in tank water and in groundwaters. Contrary to Bengaluru, Cl dominance in the rural catchments only occurred in two pre-monsoonal tank samples in TG Halli, and in none of the samples in Berambadi (for more detailed information on water types by study area, water source, and sampling round see Supplementary Material Table S3).

Surface water major ion chemistry and field parameters
Piped water in Bengaluru has median concentrations of Ca, HCO 3 , and SO 4 similar to the city's tank water, but is generally lower in concentrations of K, Na, and Cl (Fig. 6, with tabular data given in Supplementary Material, Table S4). Mg concentrations in piped water (medians around 19 mg/L) are about twice as high as in the city's tank water. Most noteworthy, NO 3 was detected in the post-monsoonal piped water in concentrations of up to 19.7 mg/L, while little NO 3 was observed in Bengaluru's tank water in either sampling round (all concentrations < 1 mg/L NO 3 ). River water samples from Vrishabhavathi River showed generally higher concentrations of all major ions and SEC of around 1400 µS/cm (Fig. 6). Tank waters are relatively similar in composition for the three studied catchments, though the Bengaluru tanks have slightly higher ion concentrations with regards to Ca, Mg, HCO 3 , and most noteworthy higher Na and Cl (Fig. 6). The higher Cl concentrations in Bengaluru exceed those of the rural catchments roughly by 30 mg/L and with higher concentrations at Kaikondrahalli Tank (>90 mg/L in both sampling rounds).
Pre-monsoonal Cl medians are higher than post-monsoonal ones in all study regions. In Bengaluru, this is mainly caused by strong increases in Cl concentrations in the two tanks in the outskirts (increase by 101% and 73% at Kaikondrahalli and Allalasandra Tank). The inner-city tanks Sankay and Lalbagh, surrounded by a more densely piped mains network (Fig. 1B)  respectively. SEC and pH in tank waters are higher in the pre-monsoon, with generally larger increases in the rural catchments than in Bengaluru (Fig. 6).

Groundwater major ion chemistry and field parameters
Groundwaters in all three study areas have a relatively constant major ion composition over time (changes by < 25%) except for distinct decreases in NO 3 (by 27-41% in all study areas) and Mg (by 27% in TG Halli) in the pre-monsoon season. Ca, Mg, and NO 3 concentrations in the rural catchments are significantly higher than in Bengaluru, while the reverse holds true for pH (Wilcoxon rank sum test < 0.5). Mg, NO 3 , HCO 3 , and SEC all follow a distribution of Berambadi > TG Halli > Bengaluru. Median K concentrations in groundwaters are without significant differences across study areas and in time, and range 5.5 to 6.8 mg/L. Median Cl concentrations followed the order TG Halli > Bengaluru > Berambadi (highest/lowest median of 168.4 and 67.9 mg/L), and were slightly lower in the pre-monsoon sampling (decrease by 1-12%). Cl concentrations > 100 mg/L were measured in Bengaluru groundwater near Kaikondrahalli, Lalbagh and Vrishabhavathi River (highest measured concentration 560.9 mg/L; near Kaikondrahalli), and in TG Halli in almost all samples from the Aralamallige southern subcatchment (maximum concentration 475.1 mg/L) but none of the samples from the second sub-catchment Hadonahalli (maximum concentration 61.3 mg/L). In Berambadi, a slight trend of increasing Cl concentrations from west (mostly < 100 mg/L) to east (most concentrations > 100 mg/L) was observed. High concentrations of NO 3 exceeding the Indian drinking water limit (DWL) of the Bureau of Indian Standards (BIS) of 45 mg/L NO 3 (BIS, 2012) were primarily detected in Berambadi (85% of all samples, maximum 446.3 mg/L NO 3 ) and TG Halli (33%, maximum 162.5 mg/L NO 3 ), and only at one site in Bengaluru (54.8 mg/L near Lalbagh tank; post-monsoon). It is noteworthy that, except for a weak significant correlation for TG Halli, NO 3 did not show significant trends with the drilled depth of the sampled boreholes (Fig. 3). For a full list of percent of samples exceeding DWLs by study area, see Supplementary Material Table S3.

Chemical ratios, trends with distance to surface water and estimated impact of tanks
The covariant plot of Na versus Cl (Fig. 7A) demonstrates that Na:Cl ratios in Bengaluru (with the exception of the Cl-rich samples from near Kaikondrahalli) closely follow the 1:1 ratio for halite sources or anthropogenic NaCl inputs, while there is excess Cl (or a depletion in Na) observed in TG Halli and Berambadi, especially in the post-monsoon season. In the covariant plot of Ca versus Mg (Fig. 7B), most groundwater samples in all study areas fall between the 2:1 line and 1:1 line, though there is generally a stronger ion concentration and higher proportion of Mg in the rural catchments compared to Bengaluru. This could be indicative either of local differences in geological setting, or be an impact of agricultural inputs, e.g. application of CaNO 3 -type fertilizer, indicated by a weak to strong correlation between Ca and NO 3 in rural groundwaters (R = 0.53 and R = 0.74 for TG Halli and Berambadi, respectively, with p < 0.0005). Fig. 7C shows the Gibbs plot for TDS versus Na/(Na + Ca) weight ratio, which can give information about the dominant evolution processes of waters (Gibbs, 1970). The results indicate that neither study area exhibits strong evaporation nor precipitation dominance, and there is a stronger clustering of waters from all source types in Bengaluru, while in the rural catchments, most tank waters are separated out from the groundwater samples. Fig. 8A shows the Cl/Br mass ratio versus Cl (mg/L) for each study area by source type. Ratios between Cl and Br in waters from different sources usually have a typical range, and due to these distinct ratios, the Cl/Br ratio can give information on mixing processes (Alcalá and Custodio, 2008) and be used to identify recharge to groundwater. Whilst Cl is usually detected at concentrations well above the detection limit, it should be noted though that low Br concentrations close to detection limits may cause some uncertainty in Cl/Br ratios. In our study, 4 of the tank water samples had Br concentration < 1.5 times the detection limit (2 in TG Halli and 2 in Berambadi, concentrations of around 0.015 mg/ L). We have no repeat analysis of these samples, but a presumed uncertainty of +/-10% would shift the Br/Cl mass ratio at the same percentage, so interpretation of these data will focus on the combination of Br/Cl mass ratio and Cl concentration, without too much emphasis on the Br/Cl mass ratios alone.
The Cl/Br plots show a distinct pattern for each study area. In Bengaluru, groundwaters have a median Cl/Br mass ratio of 552 (indicated by the dashed line, Fig. 8A). Surface water Cl/Br ratios mostly lay within 442-2222. One location, Allalasandra tank, has a markedly low ratio of 66 (post-monsoon) and 104 (pre-monsoon). This is caused by unusually high Br concentrations of up to 1 mg/L (versus the overall median of 0.1 mg/L). Other tank waters in the inner-city of Bengaluru plot similarly to most of the piped water (Fig. 8A). As indicated in the previous chapter, piped water and inner-city tanks in Bengaluru generally have lower Cl than river water samples and samples from Kaikondrahalli tank, which also creates a distinction in the Cl/Br plot. The groundwater samples mostly cluster according to their specific location, and no clear trend can be observed.
In TG Halli, groundwater has a relatively uniform Cl/Br ratio with a median of 239 (Fig. 8A), but with a large range in Cl concentrations (from 14-475 mg/L). Tank waters seem similar in Cl/Br ratio, but with some spread in Cl concentrations (3-89 mg/L). The shift in groundwater composition is dominated by a shift along the x-axis which may indicate mixing with either lower or higher Cl water (but with similar Cl/Br ratio). This could indicate a dilution of Cl-rich groundwater by the lower-Cl tank waters. However, a correlation analysis (Person correlation) between the proximity of the borehole location to tanks does not show significantly lower Cl concentration near tanks (Fig. 8B).
Berambadi groundwater samples show a very pronounced pattern indicative of a mixing line that could be produced by waters with constant Br concentration, but higher/lower Cl. Potentially, this is an indication of recharge by return flow from groundwater irrigation, as one may suspect similar Br, but increased Cl due to evaporation processes (mixing of this type will result in a right shift due to the increased Cl, and upward shift due to the change in Cl/Br ratio). Some of the higher Cl concentrated-tank waters seen on Fig. 8A could be a potential endmember (in this case of lower chloride concentration than the groundwaters, thus diluting their Cl concentration). However, the lack of correlation between a borehole's Cl concentration and its borehole proximity to tanks does not seem indicative of this (R = -0.15, p = 0.45; Fig. 8B), so that the mixing of the groundwater with higher Cl concentrations seem more likely.
Aside from Cl, water isotopes are usually considered a conservative tracer, as they are commonly detected in groundwater with the same isotopic composition as the recharging water, e.g. rainfall recharge (e.g. Darling and Bath, 1988). Table 3 shows the mean d-excess for amount weighted rainfall, tank water and groundwater (mean over both sampling periods) in the three study catchments. It should be noted that the Berambadi rainfall is missing the analysis of one sample, and that a longer-term (several years) amount weighted rainfall would be considered more reliable. Hence, we are only using TG Halli here as example for a 2-endmember mixing. Because there is no significant correlation between d-excess and distance of the boreholes to the nearest tank (Supplementary Material, Figure S2; this is in good correspondence with the lack of significant trends in Fig. 8B), the mean of all groundwater samples is considered. Assuming the fraction of tank water times tank water d-excess plus the fraction of amount weighted rainfall times rainfall d-excess to be equal to the groundwater composition, this would result in a fraction of 2.6% (TG Halli) for tank recharge, which is very low and indicates that the effective recharge of deeper groundwater within the fractured bedrock seems to be negligible compared to other recharge sources. A more specifically designed monitoring network with site-specific installation of deeper monitoring boreholes of varying distance from tanks could provide further details on whether significant trends of conservative tracers might be detectable on smaller scales.  Table S1). Most monitored sites show a decreasing trend in water level between November and April when precipitation is low.

Groundwater level observations
In Bengaluru, groundwater levels are shallow (average daily maximum level ranging 1-9 m bgl) at the three sites closest to the city centre, and deeper (19-34 m bgl) at sites towards the outskirts (Fig. 9B). Daily groundwater level fluctuations are highly variable across sites, but can be categorized into Group A: sites that experience high and rapid fluctuations (IISC, Yelahanka, Nimhans), and Group B: sites that show little daily fluctuations (Cubbon Park, Lalbagh Park, Bangalore University). Because rapid changes in groundwater level are most commonly caused by nearby groundwater abstractions, it can be assumed that Group A is more directly influenced by anthropogenic head changes, while Group B reflects a more undisturbed condition. The extremely rapid drawdown in some of the boreholes (level decrease of up to 3.5 m in 15 min) is indicative of a high fracture connectivity and low storage in this part of the aquifer. The maximum daily water level in Bengaluru decreased by 1-5 m in the dry season, and showed no strong overall trends, but possibly some downward tendency at two sites (IISc and Lalbagh).
Groundwater levels in TG Halli (Fig. 9B) were only monitored from September 2017 to end of March 2018, which coincides with the premonsoon water sampling. Water levels are much deeper than those in Bengaluru with averages ranging 79 to 148 m bgl. These levels may have dropped somewhat lower in the summer months that are not included in this dataset. The strong variation in groundwater levels between sites corresponds well to previous studies (Ballukraya and Srinivasan, 2019). Three of the five monitored sites (indicated by the black lines Fig. 9B) show much more rapid level increases after strong rainfalls in September to November 2017 (Fig. 9A), which is indicative of rapid fracture flow into these boreholes.

Table 3
Mean d-excess (‰ VSMOW) for rainfall (amount weighted rainfall), tank water and groundwater (mean over both sampling periods). It should be noted that the Berambadi rainfall is missing the analysis of one sample, and that a longer-term (several years) amount weighted rainfall would be considered more reliable. Hence, these values are here only being used to exemplify thatconsidering rainfall and tank water to be two end-memberstank water seems to play a minor contribution in annual groundwater recharge. In Berambadi, groundwater levels show a distinct regional gradient with the shallowest average groundwater level (14 m bgl) at the westernmost site, and deepest average level (62 m bgl) at the easternmost location, which corresponds to observations from previous studies (e.g. Buvaneshwari et al., 2017;Collins et al., 2020;Sharma et al., 2018). The monitored sites exhibit a temporal trend with post-monsoon peak levels in November 2017 and a gradual decline until the onset of the new recharge in May 2018. Compared to Bengaluru and TG Halli, the seasonal groundwater level declines are more pronounced with a drop of 13-15 m from the peak level in November 2017 to the lowest recorded level in April 2018.

Evidence of connectivity of the deeper basement aquifer to modern recharge processes
The observed lack of strong correlations between isotopic ratios or NO 3 concentrations with borehole depth in all study areas (Fig. 3) is indicative of a well-mixed system, and the detection of typical anthropogenic pollutants, such as NO 3 , in high concentrations even at greater depth indicates that recharge is modern. These findings are in accordance with a parallel study by Collins et al. (2020) who demonstrated by means of a range of hydrogeochemical techniques, including the residence time tracer CFC, a high degree of lateral and vertical connectivity in the Berambadi catchment. For example, a homogenously high concentration of CFCs in both shallow and deep monitoring boreholes in this study was found indicative of recent recharge and of a well-mixed groundwater system without clear vertical stratification. However, depending on the number and position of intercepted fractures (borehole casings in the study area are usually cased down to the base of the weathered zone, which likely ranges between 15 and 60 m bgl, and open bores from thereon), a water sample may represent an integrated sample with contributions from shallower fractures in the upper part of the fresh rock potentially overlaying chemical signatures from deeper fractures. It can be concluded that the deeper basement aquifer has a component of modern recharge in all investigated catchments with either fractures or the boreholes acting as preferential pathways.

Comparison of relative contribution of tank recharge to other sources
The variability of isotopic ratios in the groundwater samples from Bengaluru ( Fig. 2 and Supplementary Material Table S2), indicates that groundwater recharge in the urban setting is influenced by multiple sources including mains leakage, urban runoff and drainage, and infiltration from rainfall and surface waters (conceptualized in Fig. 10).
As expected for the complex urban setting, hydraulic heads in the monitored boreholes demonstrated distinct patterns, ranging from sites with little daily fluctuations (Cubbon Park, Lalbagh Park, Bangalore University) to sites with high and rapid fluctuations (IISC, Yelahanka, Nimhans) (Fig. 9). From a monitoring perspective, these strong head changes at Bengaluru highlight that manual monitoring in low-storage fractured aquifers is best undertaken by taking a series of measurements over time to ensure that a representative water level is recorded. Monitored boreholes in areas with a higher density of the piped water supply system that delivers mainly water from the Cauvery River as potable water supply generally exhibited shallow groundwater levels, typically about 1 m bgl (Lalbagh and Cubbon Park). The shallower groundwater level is likely caused by substantial recharge from the piped mains water network. This is in accordance with other recent studies Tomer et al., 2021). The similar isotopic signature (Fig. 4) and Cl/Br ratios in inner-city tanks, piped water and groundwater (Fig. 8) also point to a strong connection between these waters -with tanks possibly being recharged by the shallow groundwater in the post-monsoon season. Pre-monsoon, most groundwater levels were below the assumed average tank depth, and tanks would be expected to lose water through infiltration and potentially recharge groundwater (Fig. 10A). However, assumed flow is not based on precise monitoring of the tank water level, and a more detailed study would be necessary to clearly identify the predominant flow direction. In either case however, connections and exchange of tank, piped and groundwater in the inner-city area is of high likelihood. In the outskirts, tank water is mainly fed by direct rainfall and runoff, as well as treated and partially untreated sewage collected in drainage canals (indicated by a higher pollutant load of these sites, e.g. Cl), and to a lesser extent by the sparser mains system (Fig. 10B). This conceptualization is supported by strong post-monsoonal similarity between tank water and the amount weighted average rainfall composition in the outskirts, and conversely the stronger post-monsoonal similarity of isotopic composition in tank waters with piped water and groundwater in the inner-city areas (Fig. 4).
In the rural catchments (TG Halli and Berambadi), recharge pathways differ from the urban city in regards to a lack of water mains and the additional recharge from irrigation return flow (Fig. 11). The rural groundwaters show higher spatial similarity in isotopic compositions (e. g. lower standard deviations across the sample groups, see Supplementary Material, Table S2) than groundwaters in Bengaluru, and very similar isotopic composition to the amount weighted average rainfall (Fig. 2). This could indicate that the majority of groundwater recharge in the rural catchments is constituted either by rainfall orconsidering that most of this area is highly irrigatedalso by (usually slightly more evaporated) return flow from irrigated agriculture. The latter is indicated by the higher evaporative change in rural groundwaters (significant increase in d-excess in the pre-monsoon) compared to Bengaluru (no significant change, Wilcoxon rank sum test; Supplementary Material Figure S1). The analysis of the Cl/Br ratios showed groundwater samples with relatively constant Cl/Br ratio, but with a high range of Cl concentrations, indicative for mixing of low Cl 'pristine' groundwater with a high Cl endmember, for TG Halli (Fig. 8). In Berambadi, a typical mixing line of groundwater with an endmember with higher Cl and higher Cl/Br could be observed. This mixing with Cl-rich water is a good indication that irrigation return flows, which typically have increased Cl from agrochemical inputs, could be the dominant recharge source. High NO 3 concentrations further corroborate this conclusion. Even though potential recharge from Berambadi tanks is somewhat spatially distributed because nearby farms may supply water from the tank to fields via drainage channels (Collins et al., 2020), this neither creates an isotopic signal in the groundwaters nor is there indication from the Cl/Br analysis that much mixing of tank water and groundwater is taking place (tank waters mostly had higher Cl/Br, but lower Cl, and no mixing line with tanks as endmember could be identified). Thus, this indicates that much of the infiltrated water from tanks is lost within the unsaturated zone, for example by lateral flow processes and evapotranspiration (Scanlon and Cook, 2002), and is therefore not significantly adding to groundwater recharge.
This finding is in contrast to those from investigations using recharge estimations based on water level fluctuation methods , and studies relating tank density to groundwater levels (Chowdhury and Behera, 2018) which observed higher groundwater levels near tanks. However, in both of these studies, it is not clear if the higher water levels can be attributed to groundwater recharge by tanks, or to other factors, such as decreased groundwater abstractions in areas with higher surface water availability (i.e. tanks), or higher water storage from tanks in the vadose zone (i.e. less need for irrigation).
Previous studies have calculated percolation efficiencies of tanks, defined as the percolated fraction of tank water depletion over a given time period, of 35-68% Massuel et al., 2014;Metha and Jain, 1997;Perrin et al., 2012;Singh et al., 2004;Sukhija et al., 1997). However, little is known about how much of the infiltrated water actually recharges groundwater, as not all water lost through the base of a tank can be expected to reach the groundwater table, but may be stored in the vadose zone or taken up by evapotranspiration De Vries and Simmers, 2002). A rare study using water level data in combination with water isotopes and chemical data to estimate recharge sources concluded that rainfall was the main source of recharge with only marginal recharge from a nearby tank (Saha et al., 2013). Thus, a revised approach to directly estimate water fluxes below tanks may be beneficial, for example by more local-scale studies using environmental tracers as suggested by  combined with water level observations. Regardless of the recharge rate beneath individual tanks, in the context of the landscape as a whole, tanks cover small areas, especially in urban setting where they are encroached on for development, and hence their overall contribution to recharge may always be limited, but yet, it will be a useful endeavour to quantify better the unit area impacts of tanks. For example, we did not see significant trends of conservative tracers (stable isotopes and Cl, Pearson correlation) in groundwater with distance from tanks, and we estimated a recharge contribution only 2.6% from tank waters in TG Halli (based on a two-end-member mixing for d-excess). Because of constraints on the regional data availability (for example, only a one-year amount weighted rainfall average was obtained for this study, and data on variations in tank surface areas and tank levels/storage for the three regions are not available) there is some uncertainty in these results, and it may go beyond this study to define a per-unit-area value of tank recharge. However, a more specifically designed monitoring network with site-specific installation of deeper monitoring boreholes of varying distance from tanks could provide further details on whether significant trends of conservative tracers might be detectable on smaller scales. We therefore propose that studies following a hydrochemical approach as presented here combined with more site-specific monitoring networks as for example outlined in Knappett et al. (2012) and (Stahl et al., 2014) could be informative for the future.

Contamination risks posed by tanks and other sources to groundwater
Distinct hydrochemical patterns were observed in Bengaluru (e.g. Mg and Cl 'hotspots' near Kaikondrahalli tank and Vrishabhavathi River, and higher concentrations of NO 3 near Lalbagh tank), while contamination in the rural catchments seemed more widespread.
In Bengaluru, increased concentrations of Cl were present in tank water and groundwaters, and the dominating 1:1 molar ratio of Na/Cl (Fig. 7A) and very high Cl/Br ratio (Fig. 8A) indicates that Cl contamination is related to anthropogenic sources, such as effluent from households and industry (Alcalá and Custodio, 2008). Increased levels of Cl were found in particular in the western outskirts of Bengaluru, around Kaikondrahalli tank, and our conceptualized model indicates that the lower groundwater levels and lower density of mains network (i.e. lower dilution of infiltrated tank water by mains leakage) will increase the risks to surface water quality. This holds true for tanks, but in particular also for larger perennial drainage systems, such as the Vrishabhavathi River, which was found to have high concentrations of K, SO 4 , Cl and HCO 3 -also indicative of anthropogenic pollution. In the city centre, water levels were extremely shallow in the wetter part of the year, and recharge/mixing of tank water with mains leakage will substantially reduce the impact of tanks on groundwater recharge. However, it should be noted that post-monsoon samples from the piped water showed high NO 3 concentrations (up to 19.7 mg/L NO 3 with a median of 15.2 mg/L). Considering that most of the piped water is imported from the Cauvery River, which flows through agricultural areas, it needs to be considered if the imported piped water might be introducing agricultural pollutants into the groundwater system. Similar to Bengaluru, high Cl concentrations were detected in TG Halli groundwaters (seasonal median up to 168.4 mg/L) and Berambadi (seasonal median up to 76.8). For comparison, Buvaneshwari et al. (2020) only detected Cl concentration of mostly < 10 mg/L in a neighbouring, more pristine area in similar geological setting and at a close distance to the Berambadi catchment. Therefore, the higher concentrations in Berambadi compared to its neighbouring area are unlikely to be of natural origin. The most likely cause of this are applied fertilizers (e.g. Mg(NO 3 ) 2 , Ca(NO 3 ) 2 , KCl, or MgCl 2 ), from which NO 3 and Cl are leached into the groundwater during the monsoon season . This could also explain the higher concentrations of Mg in the rural catchments, especially post-monsoon in Berambadi. NO 3 concentrations exceeding the BIS DWL (drinking water limit) of 45 mg/L were detected most frequently (85% of all groundwater samples) in Berambadi, with a peak concentration of 446 mg/L (Table S5, Supplementary Material). This is slightly higher than observed in the same catchment by Buvaneshwari et al. (2017), who detected peak concentrations of 360 mg/L, and average concentrations of 77 mg/L in September 2012, and 75 mg/L in June 2013. Average concentrations in this study were somewhat higher with 133 mg/L (post-monsoon, October 2017) and 113 mg/L (pre-monsoon, April 2018), which shows Different colours are used for source differentiation (e.g. brownish colour for tank water, lighter blue for rainwater, darker blue for irrigation returns). Arrows indicate recharge pathways, with thickness of arrows corresponding to dominant recharge sources. (Drawings are not to scale.) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) an upward trend in nitrate pollution in recent years and re-iterates that better nutrient (and irrigation) management is crucial to prevent further deterioration of the region's groundwater quality (Sharma et al., 2018).

Conclusions
A suite of chemical analyses (stable isotopes, water chemistry) of samples from piped water (sourced mainly from the Cauvery River), surface waters (tank water and river water) and groundwater in three catchments set in the crystalline basement of the Cauvery Basin, Karnataka, India, was undertaken to investigate recharge patterns and pathways. The three catchments comprised the urban area of Bengaluru City (the state capital), and rural areas with higher levels of agricultural activity north of Bengaluru (TG Halli) and the Berambadi catchment in southern Karnataka. The focus of this study was to use hydrochemical tracers to see if significant groundwater recharge from 'tank water' (water from artificial lakes) could be identified within the groundwater system.
The distinct isotopic ratios and hydrochemistry between tank water and groundwaters at most sitesand especially in the rural catchments observed in this study suggest that these are suitable tracers of tank water recharge. The groundwater chemistry results indicate that tank recharge is of limited importance to regional groundwater recharge and quality in rural areas, where recharge from precipitation and groundwater recycling from irrigation dominate the recharge signal. In the urban setting (Bengaluru), impermeable surfaces increased the relative effect of recharge from point sources such as tanks and rivers, but where present, pipe leakage from public-water-supply accounted for the majority of recharge. Location-specific differences could be observed, e.g. between the city centre, where shallow groundwater levels are predominant and groundwater (as well as mains water leakage) is likely to interact and possibly recharge tanks, and the outskirts of Bengaluru. This means that to sensibly estimate recharge by tank water, site-specific information, especially on water level fluctuations and alternative recharge sources needs to also be considered. Additional studies to quantify tank recharge and revisions to the current guidelines for national groundwater recharge estimations, using a less generalised approach, are recommended to avoid over-estimating the role tanks play in groundwater recharge.
Groundwater contamination in Bengaluru from different sources was observed with distinct regional patterns, e.g. Cl and Mg 'hotspots' near Kaikondrahalli tank and Vrishabhavathi River, and higher concentrations of NO 3 near Lalbagh tank. Due to the shallow water levels in Bengaluru's inner-city, and slight indication of recharge from tank water in the outskirts, tanks in Bengaluru could be a possible source of contamination due to vertical infiltration to groundwater in the dry season. In the rural catchments, especially Berambadi, contamination was more ubiquitous, particularly regarding agricultural pollutants. The exceedance of DWLs by NO 3 in the rural catchments is likely the impact of diffuse recharge under fertilized areas, with tanks posing a negligible risk to groundwater quality compared to other agricultural activities on the land surface.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.