Anthropogenic and geogenic influences on peri-urban aquifers in semi-arid regions: insights from a case study in northeast Jaipur, Rajasthan, India

Jaipur, the capital of the state Rajasthan in northwestern India, is situated between latitude 26.82 to 27.00° and longitude 75.73 to 75.85° (WGS 84) on the eastern border of the Thar Desert (Fig. 1). It is classified as a semiarid area with a long-term average annual rainfall of 565 mm in Jaipur district, based on data from 1977 to 2006 (CGWB 2013). In all, 90% of the annual rainfall takes place during the south-west monsoon from June to September (CGWB 2007). A high variability of the annual monsoonal precipitation amount (300–700 mm) can be observed (Yadav et al. 2016).

The city is situated within the Proterozoic Aravalli Mountain Range which extends over 700 km from North Gujarat in the SW up to Delhi in the NE (Sharma 2009). The steeply rising hills (up to 650 m NN) surrounding Jaipur from NE to W belong to the Alwar series of the Mesoproterozoic Northern Delhi Fold Belt and mainly consist of fractured and partly weathered quartzite. The Geological Survey of India (GSI 2011) reports for the Alwar series a feldspathic quartzite. In the occasionally wide valleys in-between the NNE–SSW striking hills (GSI 2011) the bedrock is covered by Quaternary alluvial sediments, which can reach a thickness of up to 100 m (CGWB 2007). The alluvium does not only consist of fluvial sediments, but also aeolian sediments and denudational material close to the hills, leading to a wide range of sediments from silt and fine sand to coarse sand and gravel near the hills. Three fluvial phases during the Quaternary reworked the sediments and formed the landscape so that today buried river channels as well as undulating fluvial plains can be found (CGWB 2013; GSI 2011). Kankar layers, consisting of secondary calcite rich nodules, and clay lenses occur at various depths within the Quaternary alluvium (Antea International, unpublished report, 1999).

Two peri-urban communities situated in the NE of Jaipur in the subdistrict Amer were chosen as exemplary study areas due to their socio-economic, topographic and geologic heterogeneity. Khara Kuaa (KK) is a well-established Hindu community lying at the foothills of the Aravallis behind Amer Fort, where two historic water reservoirs are located. The main aquifer is the weathered part of the Proterozoic quartzite (fractured hard rock aquifer). The water demand is partly covered by public water supply (water from the surface-water dam Bisalpur mixed with local groundwater is supplied for 1 h/day or via tanker), by private water tanker (local groundwater) and by groundwater drawn from hand pumps. The second study area, Nai Ki Thari (NKT), is a Muslim community located in the flat plain 5 km outside of the current city boundary, which was developed only within the last 20 years. The main aquifer is the Quaternary alluvium. More than 95% of the water supply is based on local groundwater, as there are no public water supply connections.

A detailed mapping of the water infrastructure was conducted to understand the role of groundwater in the target zones and to identify wells suitable for hydrogeological investigations. Mapping activities included interviews about how and where groundwater is used by whom and about possible sources of contamination (e.g. location, construction and maintenance of septic tanks). An emphasis was laid on surface-water bodies and their possible connection to the aquifer. Data regarding the design of the wells were collected during interviews with well owners and users, as no bore logs were available.

Wells for regular hydrochemical sampling were chosen according to the following criteria:

As no official piezometer (only one abandoned bore well used for this study as piezometer) was located within the study area, hand pumps and bore wells were selected as sampling points.

Water samples were taken for analyses of major ions and stable isotopes of water. For hydrochemical analyses, 121 water samples were collected from a total of 34 wells and two surface-water bodies (Fig. 2) from November to December 2017 (Table 1). Thirty of these wells were chosen for regular sampling at a 3–5-month interval, the number varying slightly at each campaign due to some wells falling dry or pumps braking. Additionally, one rain sample was collected for a full hydrochemical analyses in January 2017.

At all wells, except the piezometer, the installed pumps (manual or electric) were used for sampling. Because of frequent power cuts, reluctance of well owners, wastage of water at private wells, or environmental conditions (manual pumping at >40 °C), prepumping was difficult at some sampling points. In such cases, sampling was conducted at the usual pumping times of the wells. In this way, it was guaranteed that the well was already running for more than 30 min. The physicochemical parameters —temperature (T), electrical conductivity (EC), pH, oxidation-reduction potential (ORP), dissolved oxygen (O₂)—were measured on-site by customary probes with a digital portable multimeter (Hach HQ40D). After a complete screening in the first sampling campaign, colorimetric tests to determine redox-sensitive species—(ammonium (NH₄⁺), nitrite (NO₂⁻), nitrate (NO₃⁻), sulfide (S₂⁻)—were applied in the field wherever necessary (Table S1 of electronic supplementary material, ESM). Bicarbonate was determined using a titrimetric test (accuracy 0.2 mmol/L). All samples were filtered through a 0.2-μm polyethersulfone membrane filter and samples for cation analysis were acidified with ultra-pure 60% nitric acid to pH <2, before being filled in polyethylene bottles and stored under cold conditions. The storage period lasted from a few days up to 1 month. Cations were analyzed by ICP-OES (Optima 2100 PerkinElmer) and anions by ion chromatography (Dionex ICS 1100) at the hydrogeochemical laboratory at the Institute of Geological Science, Freie Universitaet Berlin, Germany. Samples showing a charge balance error (CBE) below 10% were considered for further analysis (King et al. 2014).

A total of 152 water samples for stable isotope analysis were taken between November 2016 and October 2018. Out of the 152 samples, 112 are groundwater samples, 28 surface-water samples and 12 rain samples. Surface-water samples were taken regularly from two historic reservoirs and one taalaab (lake, where greywater from the surrounding area gets accumulated) and once from two canals (nallahs). The sampling at the lakes was mainly carried out with a 2-L plastic bucket connected to a 1.5-m-long stick. Rain samples were taken on top of a small building to the east of Jaipur City (Lat: 26.903073°, Long: 75.841136°) with a self-made rain isotope collector using the design by Gröning et al. (2012). The samples were collected during seven rainfall events during the monsoon season 2017 and at single rainfall events in January and March 2017. They were filled in tightly sealed glass vials without pretreatment and stored in a dark and cool place until analysis (Picarro L2120-i cavity ringdown analyzer) at the BGR laboratory in Hannover, Germany. All samples were analyzed as replicates, drift-corrected and calculated against international standards (VSMOW/VSLAP). The values are given as δ-value in ‰. Analytical error for a quality check sample measured in each sequence is better than 0.2 and 0.8‰ for δ¹⁸O and δ²H, respectively.

For the regular water-table measurements every 2 months, 30–50 wells were chosen within both communities and their surroundings (~30 km², elevation difference of about 45 m) including hand pumps, piezometers, bore wells, dug wells and step wells. The number of wells varied slightly from campaign to campaign as some wells went dry, were closed, or were completely overgrown. A water level data logger (Solinst) was installed in an unused hand pump tapping the fractured quartzite aquifer. A leveling survey was conducted to determine the absolute elevation of the 31 most important wells. The applied D-GPS Leica Viva GS14 system can give, under favorable conditions, a horizontal accuracy of 8 mm + 1 ppm and a vertical of 15 mm + 1 ppm when applying the RTK mode (real time kinematic) with a single baseline (Leica 2016). To get more information about the thickness of the Quaternary alluvial sediments/aquifer and the distribution of thickness of possible clay layers, 30 geoelectrical resistivity soundings were conducted in Nai Ki Thari. The soundings were recorded with a computerized resistivity meter CRM-500 and the Schlumberger’s configuration was adopted.

Multivariate statistical analysis (MSA) was applied and compared to graphical methods (Durov plots, binary plots). For that, the original data base was edited. As a first step, only the groundwater samples from one sampling campaign (October 2017) were used to avoid duplicates which could lead to an overestimation for certain wells. The October 2017 sampling campaign was chosen because all samples of this campaign show a CBE within ±3%. Regarding the input parameters, the variables were reduced to 17 variables (main ions, EC, pH, O₂, Br, Sr, Zn, F, Ba, B) including only variables with results for all samples (Güler et al. 2002). Values below the detection limit were replaced by multiplying the respective lower detection limit with 0.55 to avoid nonnumerical values. This simple substitution method is possible if the data set shows only a small portion (<10%) of censored values (Sanford et al. 1993) and if the choice of the substitution method does not have an effect on the final geochemical interpretation (Miesch and Barnett 1976). In a subsequent step, descriptive statistics were calculated and histograms were generated using the statistical software statistiXL 2.0, which can be used as an add-in with Microsoft Excel. Based on this, the data were checked for normal distribution as multivariate statistical methods assume a normal distributed data set. As a result, all variables, except EC, pH and Ba, were log-transformed to achieve a more normal distribution. Both, the transformed and the untransformed values were then standardized by calculating their standard scores to guarantee that all variables are equally weighted in the following multivariate statistical calculations regardless of their magnitude (Güler et al. 2002).

Hierarchical cluster analysis (HCA) was chosen as it provides the possibility to generate homogenous groups, which helps to define hydrochemical facies (Newman et al. 2016). The distance/similarity measure was calculated with the squared Euclidean distance method. In order to avoid a tendency in the subsequent analysis, the HCA was first performed with the nearest neighbor clustering method with which outliers can be identified (Daughney and Reeves 2005, King et al. 2014). Subsequently, the clustering was performed following Ward’s method. Güler et al. (2002) had shown that the combination of these methods gives best results for hydrochemical data. Next to the distance matrix and the clustering strategy, results of the HCA are displayed as a dendrogram giving a visual summary of the grouping. To avoid an unpractically large number of groups and subgroups in the HCA, four locations were classified as outliers in the HCA (KK_OBS02, KK_BW08, NKT_DW02, Amer Inflow) due to their greatly differing chemical properties.

The δ¹⁸O and δ²H values in 12 rain samples range from 2.4 to −14.2‰, and 34.1 to −102.2‰, respectively (Fig. 4a). They result in a local meteoric water line (LMWL) for Jaipur of y = 7.8x + 9.1. The historic reservoirs show a very wide range of δ¹⁸O and δ²H values (from 7.8 to 0.0‰ and 24.5 to −12.3‰, respectively), while the taalaab shows only slight variations, between 1.2 and 0.2‰ for δ¹⁸O, and between −8.9 and –16.5‰ for δ²H (Table S3 of ESM). The nallah samples are more negative than the other surface-water samples and plot close to the groundwater samples. The regression line through the water samples of the historic reservoir (y = 4.6x + 9.9) has a slope of 4.6 which is typical for evaporation lines for arid to semiarid regions (Sharp 2007). All groundwater samples, with the exception of KK_HP26, have isotopic compositions between −2.5 and − 5.8‰ for δ¹⁸O and between −21.9 and − 38.8‰ for δ²H (Fig. 4b). The most negative δ-values belong to the hand pumps situated upgradient of the study areas and grouped into Amer Inflow (KK_HP08, NKT_HP12). These samples plot very close to the LMWL indicating that they are not much influenced by evaporation. The samples of the groundwater subgroups Amer Hillside, NKT East, Amer Plain, NKT North and NKT West (1-1, 1-2, 2-1, 2-2, 2-3) seem to be located on an evaporation or mixing line, while the samples of the subgroup Amer Valley (1-3) plot on a cluster around −3.4 and − 28.9‰ for δ¹⁸O and δ²H, respectively. The sample from hand pump KK_HP26 has much higher δ¹⁸O and δ²H values (5.7 and 17.3‰) than all other groundwater samples, and plots in the region of the historic reservoir samples (Fig. 4b).

Based on the geoelectrical resistivity soundings, two main lithologies could be distinguished in Nai Ki Thari: alluvium and quartzite, whereby the quartzite is divided into a smaller weathered and a hard layer. The weathered layer is situated at the transition between the alluvium and the hard rock (Figs. S1–S3 of ESM). The alluvium has its maximum thickness in the central part of Nai Ki Thari with quartzite at depth of 50–60 m. However, it is probable, that the hard rock was not detected with the given configuration so the thickness of the alluvium may be even greater. The shallowest occurrence of quartzite is encountered in the northeast and southwest of the investigated area (Figs. S1–S3 of ESM) with quartzite at a depth range of 30–40 m below ground level (mbgl). The deeper parts of the alluvium and the quartzite appear to be water saturated. The general groundwater flow direction is roughly from west to east, towards Tributary E and Dhund River (Fig. 5). Superficial N–S striking outcrops of the quartzite in the east of Khara Kuaa probably lead to local changes of the groundwater flow direction towards the southeast.

The data of the water level logger shows the development of the water table at the quartzite aquifer from premonsoon season (end of April 2017) until postmonsoon season (end of September 2017; Fig. 6). The data gap is caused by a rapid water-table decline below the depth of the installed logger, which was then installed at a deeper level on 15 July 2017. Based on the data it can be concluded that the water table of the quartzite aquifer dropped by about 20 m within 2 months in the premonsoon season from May to beginning of July (red dotted line). During monsoon, the aquifer reacts within several days to rainfall events; however, all interpretation is only a rough assumption as both the data from the level logger and from the weather station are incomplete, and no corrections for changes in air pressure were done.

Moderately high EC values, high concentrations of SiO₂ and bicarbonate, and very high nitrate concentrations are the characteristics of the groundwater from the Proterozoic hard rock aquifer, which is (undiluted and not mixed) best represented by subgroup 1-1 (Amer Hillside). Samples from this subgroup were collected in an area where the hard rock crops out at the ground surface or is only covered by a thin layer of sediments (<10 m). The SiO₂ average concentration of 44 mg/L is in the range of literature values for groundwater of quartzite rocks, being only slightly above the concentrations of 10–40 mg/L reported for the Delhi quartzite (Datta and Tyagi 1996). The weathering of silicate minerals, converting dissolved CO₂ to bicarbonate (Weaver et al. 1999), is a likely origin of the high bicarbonate values.

The EC values and nitrate concentrations do not lie within the typical ranges expected in hard rock aquifers. One reason for high EC values is the semiarid climate of Jaipur. High evaporation rates and low precipitation can lead to an accumulation of soluble salts from the weathering process in the small soil cover and in the fractured parts of the quartzite which are then leached to the saturated zone during the rainy season (Garduño et al. 2011). The sources for the ions, apart from the rain itself, are feldspar, glimmer and iron-oxide minerals partly grown within the quartzite and partly filling the fractures. Microprobe analysis (results not shown here) show that for nearby hard rock samples, quartz and albite show a speckled appearance. As Na is well transportable in water, it is assumed that the albite developed during hydrothermal activities at greenschist-facies conditions. High nitrate concentrations are usually linked to anthropogenic activities. However, several facts point towards a geogenic origin in the hard rock aquifer. Firstly, some of the hand pumps with the highest nitrate concentrations are located in areas hardly affected by anthropogenic influences as yet (KK_HP16, KK_HP02). Latrines, sewage canals, and agriculture were not found around the wells or upgradient. Secondly, sample concentrations of up to 555 mg/L NO₃⁻ exceed the concentration ranges of typical contamination of anthropogenic origin. Total nitrogen (TN) concentrations in raw sewage influent at wastewater treatment plants (WWTPs) vary strongly, with reported values between 12 and 88 mg/L (Alemu et al. 2018; Hanson and Lee 1971; Reddy et al. 2017). TN concentrations in septic tank effluents are in a similar range (Lusk et al. 2017). Assuming that all of the nitrogen is completely converted to NO₃⁻ without any denitrification taking place (and without dilution) this would result in NO₃⁻ concentrations between 53 and 390 mg/L. Mean nitrate concentrations of 40 and 52 mg/L are reported for intensive agricultural activities (Jacks and Sharma 1983; Vikas et al. 2015). Thirdly, increased nitrogen concentration in groundwater can be associated with weathering of bedrock nitrogen (Holloway and Dahlgren 2002)—for example, a geogenic nitrate contamination is reported for quartzite aquifers in Srinagar, Uttarakhand, India (Gupta et al. 2015). So, although leaking sewers and septic tanks are also likely to contribute to the nitrate contamination as well as to the high salinity, neither problem can be solved by groundwater protection measures alone.

Groundwater samples from the alluvial aquifer locally show different characteristics, described by the three subgroups of group 2 (Amer Plain, NKT North and NKT West). The common characteristic of all group 2 samples is high EC values, ranging from 2,840 to 6,200 μS/cm.

In general, several processes can lead to an increase in salinity of groundwater: evaporation, especially in semiarid and arid areas, infiltration of poor water quality (grey water, irrigation water), and dissolution of minerals along the groundwater flow path (Etcheverry and Vennemann 2009; Gat 1975). The high salinity in the study area is most probably caused by a combination of these processes. Evaporation is likely to occur due to the semiarid climate of Rajasthan, which is supported by the stable isotope data that show the group 2 samples plotting on a local evaporation line. However, taking a closer look at group 2 reveals that there is no apparent correlation between isotopic enrichment and increased EC values, indicating that evaporation is not the sole source of the high salinity in the alluvial aquifer. Cl/Br ratios (Fig. 7) of the samples point towards anthropogenic influences. Davis et al. (1998) report molar ratios for urban sewage in England (UK) and the United States from 670 to 1,350, which is similar to the ratios detected by Vengosh and Pankratov (1998) for domestic wastewater in Israel (920–1,970). The wide ranges in Cl/Br ratios of sewage can be, among others, due to variations in source water or salt intake (Katz et al. 2011). With a Cl/Br ratio range of 550–1,100 and a Cl concentration range of 920–1,700 mg/L, most samples from NKT West (subgroup 2-3) plot in the range of “anthropogenic and urban effects”, more precisely in the range of “leaching of garbage and solid waste”. Subgroup NKT North (Cl/Br ratio: 700–1,100; Cl: 1,350–2,000 mg/L) also falls under “anthropogenic and urban effects” as well as subgroup Amer Plain (Cl/Br ratio: 600–1,560; Cl: 550–1,700 mg/L) which shows more variation. This corresponds with the observation that in most places, sanitation and wastewater infrastructures are missing and that solid waste is dumped at empty plots within the community. Particularly high EC values and nitrate concentrations and relatively positive isotope values of samples taken close to farming areas (subgroup 2-2 NKT North) clearly show the influence of agricultural practices (flood irrigation), and in this case strong evaporation caused by water logging (Saha et al. 2014).

These findings are in agreement with Coyte et al. (2019) who also describe the combination of natural and anthropogenic sources of salinity in Rajasthani aquifers. Therefore, protective measures alone, as in hard rock aquifers, would not solve the problem of high salinity. However, with the available data, it is not possible to determine the contribution of each source, and the success of protection measures or changes in irrigation practices would strongly depend on that.

The characteristics of subgroup 1-2 (NKT East) strongly indicate that the two aquifers are not completely separate units, but that mixing takes place. All samples of subgroup 1-2 were taken in an area covered by Quaternary alluvial sediments; however, it is assumed that the wells tap the hard rock underlying the alluvial sediments. Geoelectrical resistivity soundings point towards a comparable shallow depth of the hard rock layer in the northeast of Nai Ki Thari (30–40 mbgl). Interviews with the well owners revealed well depths of 110–140 m. Relatively high SiO₂ concentrations compared to other wells in Nai Ki Thari further support the hypothesis. Comparably more negative isotope values indicate that there might be only little evaporation taking place (in contrast to subgroup 1-1) or that there is mixing with unevaporated water. According to the groundwater contour map, the groundwater comes from the west where the Aravalli Hill Ridge is located and which is presumably a recharge area. In the hard rock, the infiltration process is faster than in the alluvium and, therefore, less evaporation out of the soil zone can take place. Mixing could occur due to the usual well construction, consisting of an open bore hole within the hard rock and a casing, often with slotted parts, only for the alluvial part. As the use of clay seals in the annular space is not a common practice, hydraulic connections between the aquifers are possible. Furthermore, mixing can take place where no aquitard, e.g. clay lenses, occurs. The available data and information lead to the conclusion that both situations are given in the study area.

Another example of mixing due to the well construction is the hand pump KK_HP20. It shows high Si concentrations similar to the wells from Amer Hillside and is situated close to Khara Kuaa but falls into Amer Plain group, suggesting that it is tapping both the alluvial aquifer and the hard rock aquifer. One sample from this hand pump has a nitrate concentration of 567 mg/L and an ammonium concentration of 20 mg/L. It is supposed that a case of direct infiltration by sewage took place as the other three samples show NO₃ concentrations around 200 mg/L. The hand pump is not protected, is situated near to houses and on a street frequently used by animal herds. In such cases, protective measures (better well design, protection zones around wells) would be suitable measures to improve the quality of the water supply.

Groundwater recharge takes place in the Aravalli mountain range and from the historic reservoirs located in the valleys between the hill ridges. The sample Amer Inflow, collected in the hills in the northern part of the study area (Fig. 2), is understood to most closely represent “pristine” (uncontaminated) and naturally recharged groundwater. It was collected in an area where hardly any anthropogenic influences could be observed, shows a much lower mineralization (576 μS/cm) than all other samples, plots in the range of recharge waters in the Cl/Br diagram (Fig. 7) and plots closest to the LMWL of all samples, indicating that no or only little evaporation took place (Fig. 4).

In contrast, the hand pump closest to the historic reservoir (KK_HP26) shows isotope values much more positive than all other groundwater samples. Because the isotope values lie within the range of the historic reservoir samples, it is assumed that the hand pump draws a large part of infiltrated water from the historic reservoir. The samples of subgroup 1-3 (Amer Valley) also show a number of characteristics which point towards a strong influence of groundwater recharge from the historic reservoirs: Firstly, they have comparatively low EC values, with an average of 1,325 μS/cm; although the reservoir shows a slightly higher EC value (1,457 μS/cm), as can be expected due to high evaporation losses and below-average monsoonal precipitation in the sampling year, the EC is still much lower than the EC of the ambient groundwater (2,822 μS/cm). Secondly, the depth-to-water measurements (Fig. 8) and the groundwater level contour map (Fig. 5) also point towards a strong influence from the historic water reservoirs. Wells tapping the hard rock aquifer close to the historic water reservoirs (blue lines in Fig. 8) show stable water tables throughout the year. Therefore, it is assumed that the groundwater of Amer Valley shares a high proportion of recharge from the historic reservoir.

Dug wells located far away from the reservoirs show high seasonal variations in groundwater levels (Fig. 8, green lines). In March 2017, wells in and around Khara Kuaa ran dry, with only two recovering again after the monsoon, indicating a low storage capacity of the quartzite aquifer. This is further supported by the level logger data showing a rapid water-table decline of about 20 m in 2 months (Fig. 6). The reaction time during the following monsoon season lies only in the frame of several days, implying a very low residence time in the hard rock with a high risk of fast contaminations break-throughs. Protection zones around and downstream of the historic reservoir would be crucial to prevent the deterioration of this comparatively good groundwater source.

For comparison, wells situated near Nai Ki Thari tapping the alluvial aquifer show much less pronounced seasonal variations (red lines in Fig. 8), indicating a higher storage capacity of the alluvial aquifer and/or a bigger catchment area.

The findings are summarized in a conceptual model of the study area (Fig. 9). The visualization gives a comprehensive overview of the main hydrogeological processes and zones of special concern. In Nai Ki Thari, the widespread infiltration of wastewater into the alluvial aquifer increases the EC values of the groundwater. In its surrounding area (Amer Plain), irrigation practices, e.g. flooding of fields, strengthen the evaporation effect and lead to high abstraction rates. To conclude, the groundwater of the alluvial aquifer is not potable and further degradation is likely to occur.

Large seasonal variations in the water table of the hard rock aquifer make it an unreliable source for the local users. Furthermore, the results point towards a contribution to the reported high nitrate contamination through water–rock interaction. However, with the artificial reservoirs, both problems could be solved, as can be seen in the groundwater of Amer Valley, where the infiltration of surface water from the historic reservoirs leads to stable water tables and a dilution of the groundwater. An artificial recharge structure, e.g. a recharge pond, could be built north and upgradient of Khara Kuaa, where there is a small valley between hills. At this geologically advantageous location, the rain that falls in this catchment area during the monsoon season could be retained. Besides the aforementioned positive effects on the groundwater, this would have the additional advantage that Khara Kuaa would be better protected against flooding, which occurs regularly during the monsoon season.