Introduction

Groundwater has traditionally been the primary source of domestic water in Hanoi city, the capital of Vietnam. With the area’s rapid urbanization and population growth, the city’s water demand is projected to increase from 900,000 m3/day in 2011 to 2,359,000 m3/day in 2030 (VIWASE-HPPC 2012). In order to meet such a growing water demand, a water-supply system using surface water was developed in 2008, and has been gradually expanded in the urban area (full capacity, 300,000 m3/day; 90,000 m3/day as of 2010; VIWASE-HPPC 2012). In comparison, in the suburban and rural areas surrounding Hanoi’s urban center, groundwater is expected to remain the predominant source of domestic water. In the suburban areas, domestic water has been supplied by a few of large-scale water supply systems, a number of community-based piped water systems or household tubewells (Berg et al. 2006; Do et al. 2014; Matsubara et al. 2015). Owing to the excessive abstraction of groundwater, severe land subsidence (>50 mm/year) has occurred, not only in the central urban areas but also in the suburban areas (Dang et al. 2014; Montangero et al. 2007); therefore, appropriate management of groundwater is a pressing issue in the suburban areas of Hanoi.

Hanoi is located in the upper part of the Red River (Song Hong) Delta, 100 km from the sea. The populous urban area lies predominantly on the west bank of the river. The delta typically has two aquifer systems: a topmost Holocene unconfined aquifer (HUA), and an underlying Pleistocene confined aquifer (PCA) (Bui et al. 2012a). In the areas along the Red River, including the central urban area, both the HUA and PCA are considered to be mainly recharged by the neighboring Red River (Berg et al. 2001; Jusseret et al. 2009). In comparison, many of the suburban communes are located far from the river, where surface-water bodies other than the Red River (e.g., ponds, lakes, irrigated farmlands, paddy fields, canals) are suggested as the major sources of groundwater recharge, on the basis of stable isotope analysis (Berg et al. 2008) and chemical wastewater-tracer analysis (Kuroda et al. 2015). Such surface-water bodies are prevalent in the suburban areas of Hanoi city. With increasing urbanization, however, many ponds and lakes have been reclaimed. In addition, land use has also changed, typically from bare land, irrigated farmland or paddy field to built-up areas such as residential and commercial areas. Unknown, however, is the effect that the decreasing ponds and changing land use will have on the groundwater systems and recharge in the future. To determine this, the groundwater system and recharge pathways must be characterized in more detail. In the area south of Hanoi city, beside the Red River, Norrman et al. (2008) investigated the local groundwater recharge from the Red River, by monitoring the groundwater levels for 2 years. In the suburban areas of Hanoi, however, which are more distant from the Red River, there has been no study combining stable isotope analysis with field observations of water levels of surface water and groundwater, to reveal the role of surface waters other than the Red River (e.g., ponds) as a source of groundwater recharge.

The objective of this study is to characterize the groundwater system, and its recharge mechanisms, at two suburban areas of Hanoi in the proximity of the evaporation-affected surface waters (ponds and irrigated farmlands), and to evaluate the effect of decreasing surface-water bodies and land-use change on groundwater recharge in the suburban area of Hanoi. The two study sites had contrasting land use (bare land vs. residential areas) and distance from the Red River (5 km vs. 10 km). In the present study, the water levels of groundwater in the HUA and PCA and nearby ponds were continuously monitored for 3 years. Hydraulic conductivity of aquitard samples was also measured. Moreover, stable isotopes (δ18O, δD) and hydrochemical parameters (e.g., chloride, bromide) were analyzed in the pore water of the soil cores and in the groundwater and surface waters that were routinely taken in the two sites.

Site description

Geology of Hanoi

The geology of Hanoi city is characterized by a fluvial system, with meandering rivers and levee belts, floodplain, and fluvial terraces (Fig. 1; Funabiki et al. 2007; Mathers et al. 1996; Mathers and Zalasiewicz 1999). The Red River Delta typically consists of Quaternary sediments of alluvial and marine origin (50–90 m deep in Hanoi), which unconformably overlies Neogene Tertiary deposits (∼400 m; Mathers et al. 1996; Mathers and Zalasiewicz 1999). Due to the succession of transgressions, regressions and sedimentation process occurring in the Red River Delta, the sediment lithology is highly complex, with varying sediment sequences both vertically and horizontally (Tanabe et al. 2006; Trafford 1996). Nevertheless, the Quaternary formations are generally classified into two aquifers (Trafford 1996; Water-Master-Plan 1993): (1) a topmost HUA, which consists of one or two minor sandy aquifers separated by less permeable clay and silt layers; and (2) a PCA, which consists of coarse sand, gravel and cobbles (Bui et al. 2011, 2012a). These two aquifers are separated by a Holocene-Pleistocene aquitard (HPA; Bui et al. 2011, 2012a). Typically, the PCA is exploited for municipal water supply and the HUA is used by household tubewells (Berg et al. 2008; Do et al. 2014).

Fig. 1
figure 1

Elevation map of the Red River (Song Hong) Delta—adapted from Funabiki et al. 2012; the location of the study area (Fig. 2) was added. The rivers and regional water tables are indicated

Full size image

Hanoi has a humid subtropical climate with annual precipitation of 1,680 mm (World Meteorological Organization 2016). Most (78%) of the rain falls during the rainy season, which is May to September. In this study, the suburban areas of Hanoi are comprised of the region between the 2nd and the 3rd ring roads (Fig. 2). This area, currently agricultural or grassland with some residential areas, has been included in the “central urban area” in the master plan of Hanoi city (PPJ-VIAP-HUPI 2011). Therefore, rapid urbanization is expected in this area, and a number of large-scale projects are in progress, mostly aimed at developing residential areas. The built-up surfaces in Hanoi increased by more than three times between 1999 and 2007 (Dang et al. 2014).

Fig. 2
figure 2

a Locations of study sites, b sampling locations of site Linh Dam (LD) and c sampling locations of site Tay Mo (TM)

Full size image

Study sites

The study was conducted in two suburban communes, Linh Dam (LD) and Tay Mo (TM; Fig. 2). The LD site is located in a recreational area (including, for example, an outdoor football field) which has been rapidly developed. Located 5 km west of the Red River, the land was formerly used as irrigated farmland, but currently, the land use consists predominantly of bare land, grassland, sport fields and ponds, surrounded by irrigated farmland, with only a few houses and restaurants. The ponds were excavated for recreational fishing, and are approximately 1–1.5 m deep. In April 2012, the pond closest to the well drilling site was dried, dredged and filled again with rainwater; work which has been conducted every few years on all the ponds. In comparison, the irrigated farmland was covered with water all year round. The TM site is located in a typical residential area, 10 km west of the Red River. A series of ponds (∼1.5 m deep) is surrounded by densely populated houses with paved gardens. It is thought that the ponds had originally formed a natural oxbow lake, which was later divided by reclamation. In both sites, piped water supply is not extended to households, and thus groundwater from household tubewells is the main source of domestic water (Do et al. 2014; Matsubara et al. 2015).

Drilling, soil core sampling and installation of monitoring wells

In 2012, drilling was conducted at the two field sites, for soil core sampling and installation of monitoring wells. The drilling was conducted beside the ponds (distance from the nearest pond: 15 m for LD, and 5 m for TM; Fig. 2). In each site, a pair of monitoring wells, one for the HUA and the other for the PCA, were installed. In the course of drilling for the PCA wells, the soil and sediment were surveyed to a depth of 48 and 56 m in LD and TM, respectively, penetrating the PCA by several meters. Undisturbed cores were taken from silty or clayey layers by inserting a 2-m-long tube core sampler using hydraulic pressure. Sandy or coarse soils from which undisturbed cores could not be obtained in this manner were drilled by rotary boring. For each representative facies, fresh core samples were taken in parallel at the same depths; one for pore water sampling and the other for soil analysis. The well screens were installed at 8–10 m depth and 42–46 m depth in LD, and 19–23 m depth and 50–54 m depth in TM. Note that, in LD, although the drilling was conducted down to the PCA, the well screen for the PCA was installed at the bottom of the HPA, above the PCA (see Fig. 3). Therefore, the monitoring well was treated as an HPA well. The water quality analysis also suggested that the well water represented water from the HPA (see section ‘Hydrochemical compositions’ for details).

Fig. 3
figure 3

The depth profile of chloride, sulfate and chloride-to-bromide mass ratio (Cl/Br) in the pore waters and groundwater in sites a LD and b TM. The geological column and the grain size distribution (Kuroda et al. 2016) are also shown

Full size image

Geological characteristics of cores

Upon drilling, the recovered cores were characterized by the combination of physical-chemical analysis, radiocarbon dating and X-ray radiography. The details of these core geology and sedimentary analysis results are described in Kuroda et al. (2016). Briefly, the cores were classified into three sedimentary units in ascending order (Fig. 3): (1) late Pleistocene fluvial sediments, (2) Holocene, estuarine/deltaic sediments, and (3) modern fluvial sediments.

Unit 1 (LD: 48.0–46.0 m depth; TM 56.0–40.0 m depth) mainly consists of pebbly sand. This unit, which was not bioturbated and was poorly sorted, is interpreted as Pleistocene fluvial sediments (Tanabe et al. 2006); therefore, these layers of sand and gravel are evaluated as belonging to the PCA. The water table was found at 13 m depth in LD, and at 14 m depth in TM, when the monitoring wells were installed.

Unit 2 (LD: 46.0–12.0 m depth; TM 40.0–9.0 m depth) is characterized by upward-fining sand, silt and clay in the lower part (unit 2-1), and similar but upward-coarsening elements in the upper part (unit 2-2). Unit 2-1 (LD: 46.0–24.0 m depth; TM 40.0–23.5 m depth) is interpreted as Holocene estuarine sediments, and unit 2-2 (LD: 24.0–12.0 m depth; TM 23.5–9.0 m depth) as Holocene terrestrial sediments. The silt and clay sediments in the unit 2-1 and the lower half of unit 2-2 correspond to the HPA, whereas the upper sandy sediments of unit 2-2 correspond to the HUA.

Unit 3 (LD: 12.0–0 m depth; TM 9.0–0 m depth) consists of sand, silt and clay with organic matter, and is interpreted as modern fluvial sediments. The lower sandy part of this unit is considered to be part of the HUA. The water table was found at 5 m depth in LD, and at 8.5 m depth in TM, when the monitoring wells were installed. The topmost, upper part of this unit is covered by a silty clayey layer, which is typical of most of the Holocene-covered areas in Hanoi (Jusseret et al. 2009).

The geology explained in the previous is in accordance with the typical geology of the southern and western suburbs of Hanoi, noted in the literature (Bui et al. 2011, 2012a).

Materials and methods

Soil hydraulic conductivity analysis

Soil hydraulic conductivity was measured in a total of five aquitard samples selected from major facies, for both shallow (3–4 m depth) and deep (33–34 m depth) sediments, in LD and TM. The measurement was performed at the University of Tokyo, using the flow-pump method with a constant water flow (Esaki 1996; Kameya and Tokunaga 2003). The experimental conditions and results are summarized in Table S1 of the electronic supplementary material (ESM).

Collection and on-site analysis of pore water, groundwater and surface water

Pore water was taken from each core by applying Rhizon soil moisture samplers (Eijkelkamp, the Netherlands) inside a gas-tight pack (Escal® Neo; Mitsubishi Gas Chemical Company, Japan) with oxygen absorbers (RP-3 K; Mitsubishi Gas Chemical Company, Japan), in order to prevent oxidation of the pore water. The obtained pore water was not additionally filtered, because the water had already been filtered through the 0.2-μm-pore-size membranes attached to the Rhizon samplers at the interface with the sediments.

Groundwater samples were taken from the four monitoring wells, as well as from 10 household tubewells in the neighborhood; and the groundwater was sampled multiple times from March 2011 to September 2013, for a total of 63 samples. The groundwater was slowly taken into a bucket at a rate of 5 L/min, and when the pH and electrical conductivity values stabilized, the groundwater samples were filtered through a 0.45-μm PTFE filter into acid-washed PP containers on site.

Pond water samples (27 samples in total) were taken at seven ponds near the monitoring wells in LD and TM. Irrigated farmland water samples (13 samples in total) were taken from two irrigated vegetable farmland sites in LD, and from a similar site in another suburban commune, Thuong Cat (TC, northwest of Hanoi city center; Fig. 2a). The Red River water samples (nine samples in total) were taken at four locations (Fig. 2a). These surface-water locations were sampled multiple times from March 2011 to September 2013; and each time, the samples were filtered through a 0.45-μm PTFE filter into acid-washed PP containers, similarly to the groundwater samples.

Among the water samples, those for cation measurement were acidified with HNO3 (acid/sample 1%, v/v) on the day of sampling, while those for anions and stable isotope (δD, δ18O) analysis were not acidified. All the samples were kept in the dark at 4 °C until the analysis. The sampling locations of groundwater and surface water are shown in Fig. 2. The information on the pore water, groundwater and surface water, as well as the analytical results for water quality, are summarized in Tables S2 and S3 of the ESM.

Analysis of water samples

Chemical analysis

Cations (Na+, K+, Ca2+, Mg2+ and Al3+) were measured by inductively coupled plasma atomic emission spectrometry (ICP-AES: Optima 3000DV, PerkinElmer, US). Anions (F−, Cl−, NO2 −, Br−, NO3 −, SO4 2− and PO4 3−) were measured by ion chromatography (761 Compact IC, Metrohm, Switzerland; column: SI-90, Shodex, Japan) within 2 weeks after sampling. Bicarbonate (HCO3 −) concentration was estimated on the basis of pH and M-alkalinity, which were measured on the day of sampling. M-alkalinity was measured by titrating the sample with 1.6 M H2SO4 to a colorimetric endpoint, using a digital titrator (Hach, US) and bromocresol green-methyl red indicator (Wako, Japan). All the instrumental analyses were conducted at the University of Tokyo. The quality of the analytical results was considered good because the ion charge balance was less than 5% for most samples.

Stable isotope analysis

The stable isotope ratios of hydrogen and oxygen (δD, δ18O) in the water samples were measured using cavity ring-down spectroscopy (CRDS: DLT-100, Los Gatos Research, US) by Geo Science Laboratory, Japan. The δ18O and δD compositions are conventionally expressed in terms of the per mil deviation from Vienna Standard Mean Ocean Water (VSMOW), using Eq. (1). The overall analytical errors were ±0.1 and ±0.5‰ for δ18O and δD, respectively.

$$ {\updelta}^{18}{\mathrm{O}}_{\mathrm{sample}}=\left[\left(\raisebox{1ex}{${}^{18}\mathrm{O}_{\mathrm{sample}}$}\!\left/ \!\raisebox{-1ex}{${}^{16}\mathrm{O}_{\mathrm{sample}}$}\right.\div \raisebox{1ex}{${}^{18}\mathrm{O}_{\mathrm{VSMOW}}$}\!\left/ \!\raisebox{-1ex}{${}^{16}\mathrm{O}_{\mathrm{VSMOW}}$}\right.\right)-1\right]\times 1000 $$
(1)

The δ18O and δD data for precipitation are known to plot along the meteoric water line; and the local meteoric water line (LMWL) in Hanoi has a slope of 8 and an intercept of 12.5 (WISER Database 2016). In contrast, the δ18O and δD data for water bodies with evaporation processes plot along the so-called evaporation line; and the local evaporation line (LEL) has a slope of 5.5–5.9 in Hanoi (Berg et al. 2008; Norrman et al. 2008).

Water-level monitoring

The water levels of the monitoring wells (LD-HUA, LD-HPA, TM-HUA and TM-PCA; Fig. 2) and the nearby ponds (LD-PA and TM-PB; Fig. 2) were measured using a Mini-Diver® (Eijkelkamp, the Netherlands), which continuously records pressure. The loggers were installed underwater, in the casings of the monitoring wells and in the vertically installed PVC pipes in the ponds. The logger depth (distance from the water surface to the logger) was calculated by subtracting atmospheric pressure, which was simultaneously measured using a Baro-Diver® (Eijkelkamp, the Netherlands), from the pressure measured by the Mini-Divers. The water level from mean sea level (MSL) was calculated by adding the logger depth and the elevation of the logger. The measurements were conducted from 2012 to 2015 at 10-min intervals. Every few months, the loggers were recovered for data collection.

Results

Hydraulic conductivity of cores

The hydraulic conductivity of the clayey surface soil and the silty clay in the HPA was found to be low, and similar regardless of the site or depth (1.7 × 10−8–7.0 × 10−8 cm/s for clayey surface soil, and 4.3× 10−8–5.1 × 10−8 cm/s for HPA; Table S1 of the ESM). These numbers are roughly in the same order of magnitude as the conductivity of superficial silty clayey deposits (1 × 10−7 cm/s), which was estimated by Jusseret et al. (2009). Then, if the entire segment of the HPA in TM is assumed to possess a uniform hydraulic conductivity of 5 × 10−8 cm/s and an effective porosity of 0.1, the vertical infiltration velocity through the estuarine sediments of the HPA at the drilling location (unit 2-1: 40.0–23.5 m depth) is estimated to be 5.7 cm/year, resulting in 290 years of infiltration time (see the ESM S1 for details). Note, however, that these values represent the least infiltration rate and maximum infiltration time, because the HPA had intermittent sandy and silty layers, which would have much higher hydraulic conductivity. Overall, these results show that the respective amounts of vertical infiltration of water through the superficial silty clayey sediments (e.g., precipitation), and from the HUA to the PCA, are usually very small where the aquitards are thick and intact.

Water quality

Stable isotopes

The pond waters and irrigated farmland waters are positioned along the LEL (Fig. 4a), showing that those surface waters were affected by the evaporation processes. In contrast, the Red River water samples, and the weighted average value of precipitation from 2004–2007 in Hanoi (WISER Database 2016), are all plotted along the LMWL (Fig. 4b), showing that these river waters and precipitation in Hanoi are not primarily affected by evaporation processes. In the case of the pore waters and groundwater (from both monitoring wells and household tubewells), both are plotted along the LEL in LD (Fig. 4c) and TM (Fig. 4d). These isotopic signatures suggest that both the pore waters and the groundwater were largely derived from surface-water bodies which had undergone evaporation processes (i.e., the ponds and irrigated farmland in the study sites).

Fig. 4
figure 4

The relationship between δ18O and δD in a water from ponds and irrigated farmlands, b Red River waters and precipitation, and in pore water and groundwater in c LD and d TM

Full size image

The HUA monitoring well in LD showed large variation among the different sampling campaigns (−4.0 to −4.5‰ for δ18O; Fig. 5 and Table S3 of the ESM), probably owing to changes in the quality of the neighboring pond water which recharged the HUA (see section ‘Temporal change’ for details). In contrast, the other monitoring wells in LD and TM showed very little variation among sampling campaigns (within 0.1‰ for δ18O; see Figs. 5 and 6, and Table S3 of the ESM), suggesting a relatively stable groundwater recharge system during the monitoring period.

Fig. 5
figure 5

Water levels of pond and groundwater in site LD

Full size image
Fig. 6
figure 6

Water levels of pond and groundwater in site TM

Full size image

The stable isotopes in the pore waters showed a distinct depth-profile in both sites (Fig. 7). The pore waters in the HUA and the upper part of the HPA had δ18O in the range of −6 to −4‰. The median values of the respective pond waters (−4.6‰ in LD, −6.0‰ in TM) were within this range, whereas the median value of the Red River water (−8.8‰) was clearly beyond it. The lower half of the HPA sediments, however, had lighter δ18O ranging from −8 to −6‰. The δD had a similar depth profile as δ18O, which showed a contrast between the upper and lower sediments (see Fig. S1 in the ESM). The pore water of the PCA could not be sampled, because the coarse PCA sediments were not recovered as undisturbed cores. The δ18O and δD compositions in the HUA groundwater in both LD and TM were similar to those in the pore waters near the depths of the respective well screens. Therefore, the contrasting stable isotope signatures between the HUA and HPA water samples indicate that the HUA water did not infiltrate vertically into the HPA. Interestingly, however, the PCA groundwater in TM had δ18O and δD values in between those in the HUA and HPA. This would suggest an infiltration of HUA water into the PCA, as explained in section ‘HPA and PCA monitoring wells’. The HPA monitoring well in LD had similar stable isotope compositions as the above-lying HPA pore water; this agrees with the hydrochemical analysis results suggesting that the monitoring well water represents HPA water, as explained in ‘Hydrochemical compositions’.

Fig. 7
figure 7

Depth profile of δ18O in a LD and b TM. The data for the Red River water and pond water in the respective sites are recorded at around 0 m depth. The depth profile of aquifers and aquitards is also shown

Full size image

Hydrochemical compositions

The surface-water samples generally exhibited Ca2+-Na+-HCO3 − type water in the Piper diagram (Fig. 8a). The pond and irrigated farmland data are plotted below the freshwater–seawater conservative mixing line (Appelo and Postma 2005), and close to the wastewater data (Kuroda et al. 2015), probably reflecting the discharge of domestic wastewater into the respective water bodies. In contrast, the Red River data are plotted above the mixing line, owing to the abundance of Ca2+, which would be provided from the upstream catchments.

Fig. 8
figure 8

Piper diagrams showing the main hydrochemical compositions of the a surface water and b pore water and groundwater. Wastewater data were obtained from the Yen So wastewater treatment plant in Hanoi (Kuroda et al. 2015). The lines in the piper diamonds indicate the water composition with conservative mixing, where seawater and rainwater are mixed without any reactions (Appelo and Postma 2005)

Full size image

The groundwater from both the monitoring wells and household tubewells was typically of the Ca2+-Na+-HCO3 type (Fig. 8b). This result agrees with previous studies on both the HUA and PCA in the suburban areas of Hanoi (Berg et al. 2008; Nguyen et al. 2014). All the groundwater data, except one point (for a private tubewell), are plotted below the freshwater–seawater conservative mixing line (Fig. 8a). The plots are somewhat close to those of the ponds and irrigated farmlands, but the groundwater data show a general shift toward enrichment of HCO3 −, Ca2+ and Mg2+. This would be because, as the stable isotope results suggest, an appreciable amount of the groundwater was recharged by the ponds and irrigated farmlands, and the quality of the recharged groundwater evolved through dissolution of CaCO3 and possibly degradation of organic matter (Appelo and Postma 2005). Only the HPA monitoring well in LD exhibited Na+-HCO3 − type water, and had twice the HCO3 − concentration of the other groundwaters. This groundwater had a significant amount of dissolved organic matter (∼40 mgC/L), and dissolved gas was continuously emitted from the borehole (the gas content was not measured), which indicated degradation of organic matter. The HPA of this site and that of a neighboring site (Hoang Liet commune) both had highly organic-rich peaty sediments in the HPA (Berg et al. 2008; Kuroda et al. 2016). The hydrochemical characteristics are similar to the pore waters in the above-lying sediments (Fig. 3). Therefore, even though the well was drilled to penetrate the PCA by several meters, the water from the well is more likely to represent HPA water than PCA water in the site.

The pore water from the HUA showed high concentrations of chloride and sulfate (Fig. 3), and these were highest (e.g., 86 mg/L Cl− in LD) near the ground surface. This would suggest the influence of domestic wastewater infiltrating from the ground surface (Berg et al. 2008). In LD, the chloride levels were also high in the private tubewells tapping the HUA (34–47 mg/L; Table S3 of the ESM).

The pore water from the HPA had lower chloride and sulfate levels (e.g., < 15 mg/L Cl−) than that from the HUA. The chloride-to-bromide mass ratio (Cl/Br) was lower in the HPA than in the HUA (Fig. 3 and Table S3 of the ESM). The enrichment of bromide relative to chloride is possibly caused by the degradation of organic matter (McArthur et al. 2012). In TM, the PCA groundwater had slightly higher anion levels than the HPA pore water, and had much higher Cl/Br; this is consistent with the stable isotope results (section ‘Stable isotopes’), showing that much of the PCA water in TM is derived from HUA water.

Temporal change

Both the stable isotopes and hydrochemical compositions had similar temporal changes. In the pond water, both the δ18O and Cl− levels gradually increased during the dry season, with the highest values in March or May, and decreased during the rainy season, with the lowest values in August (Figs. 5 and 6, and Fig S2 of the ESM). In the rainy season, the rainwater flowing into the ponds would dilute the ponds; whereas, in the dry season, less rainwater inflow, with more evaporation than in the rainy season, would lead to enrichment of δ18O and Cl−. In contrast to the pond water, the groundwater exhibited much smaller seasonal changes. Interestingly, the δ18O and Cl− levels in the HUA groundwater in LD showed a notable decline from November 2012 to March 2013, probably owing to the dip in the respective levels in the neighboring pond seven months before. The reduced HUA groundwater Cl− level after March 2013 would reflect the reduced Cl− level of the neighboring pond water after the dredging work in April 2012, wherein the Cl−-rich pond water was completely removed and the pond refilled with rainwater. This result would suggest that an appreciable amount of HUA groundwater in LD was derived from the neighboring pond water, owing to the existence of rapid recharge paths; in this case, the apparent flow velocity from the pond to groundwater is estimated to be 10 cm/day. The apparent hydraulic conductivity of the overall sediments between the pond and the aquifer is estimated to be 1.3 × 10−4 cm/s (here, effective porosity is assumed as 0.15), which is much higher than the hydraulic conductivity of the clayey surface soil (1.7 × 10−8–7.0 × 10−8 cm/s; section ‘Hydraulic conductivity of cores’), probably because there were places where the clayey surface soil under the pond was relatively thin, owing to geological heterogeneity.

Water-level monitoring

Ponds

Throughout the monitoring period, the pond water levels were higher than the groundwater levels, by 2–4 m in LD (Fig. 5) and 5.5–7.5 m in TM (Fig. 6), and varied within approximately 1.5 m in both sites. In LD, the water level decreased by more than 1 m with the dredging work in April 2012, and then recovered to the original level in the following month. In other periods, the pond water level in LD showed little fluctuation, even after heavy rain events. In TM, on the other hand, the pond water level fluctuated more frequently and broadly, with sharp rises after rains. This contrasting behaviour in response to precipitation would be due to differences in topography and land use. As noted earlier, the ponds in TM were likely formed as a result of the partial reclamation of an oxbow lake; therefore, the surrounding land was 1–2 m higher in elevation than the level of the pond, and thus rainfall could immediately drain into the pond, causing large water-level rises after rains. In LD, the pond was formed by excavation of land formerly used for irrigated agriculture. The surrounding land, with patches of remaining irrigated farmland, tended to be rather flat, and thus comparatively lesser amounts of surface runoff after rainfall would enter into the pond, resulting in the relatively stable water levels after rains. Furthermore, the LD site had a considerable amount of bare land and grassland, whereas the TM site had a greater proportion of buildings and paved gardens. Hence, the TM site would have more surface runoff discharging into the pond due to rainfall events.

HUA monitoring wells

In both sites, the groundwater levels of the HUA monitoring wells (Figs. 5 and 6) were higher in the rainy season (May–September) than in the dry season (October–April). In LD, the groundwater level ranged 5.0–6.9 m above MSL during the monitoring period; this fluctuation is greater than the reported 1-m fluctuation at a neighboring site (Hoang Liet commune; Berg et al. 2008). At the peak (August) of the rainy season, the HUA groundwater level reached up to the clayey surface soil, which confined the HUA. In addition, the HUA groundwater level showed inter-day fluctuations, and rose shortly after heavy rains (e.g., > 20 mm/day). These results show that the HUA groundwater level in LD is significantly affected by heavy precipitation; and this would result from either (1) direct infiltration of rainwater through the ground surface, or (2) strong influence from ambient surface-water bodies whose water level was significantly affected by precipitation. Regarding the latter (2), the nearest pond (LD-PA; Fig. 2) appeared to have little influence on the groundwater level. The pond water level varied much less (except for during the dredging period), and the HUA groundwater level did not differ between the spring of 2012 (when the pond level was low during dredging) and the spring of 2013 (when the pond level was high without dredging). The other ponds in the site were similar to LD-PA in terms of use (recreational fishing) and physical properties (area, depth), also likely suggesting little influence on HUA water levels. The only surface water which could influence the HUA water level was LD-PD, a large lake (area: ∼1 km2; depth: 4–5 m), 300 m from the site, which receives urban wastewater and stormwater. Regarding (1), the LD site was relatively flat and had a large proportion of permeable surface (e.g., bare land and grassland). Hence, there may have been an appreciable amount of rainwater infiltrating into the subsurface, which influenced the HUA groundwater-level change in the rainy season. Since the surface soil of the site had very low hydraulic conductivity, as suggested in section ‘Hydraulic conductivity of cores’, the infiltration may have occurred by macropore flow or preferential flow in the unsaturated zone, both of which allow for fairly rapid flow (de Vries and Simmers 2002; Yin et al. 2010). Macropore flow occurs through root channels, while preferential flow is caused by unstable wetting fronts and heterogeneous soil physical properties (de Vries and Simmers 2002). Both types of flow are likely to be present in the study site, as the site had been irrigated farmland for vegetables, the modern soil used for reclamation contained many large rocks and broken bricks and tiles, and Hanoi’s indigenous soil has large heterogeneity both vertically and horizontally (Tanabe et al. 2006; Trafford 1996).

In terms of total recharge over the full year, on the other hand, the amount of direct rainwater infiltration in the rainy season must have been minor, compared to the year-round, continuous recharge from evaporation-affected surface waters (ponds and irrigated farmland in this site), as the isotope and hydrochemical results suggest. Although the level agreement between the groundwater and the neighboring pond in LD was poor, the Cl− results suggested contribution from the pond water to the groundwater recharge. Other likely recharge sources would include the large and deep pond, 300 m north of the site (LD-PD, Fig. 2), or the irrigated farmland, which is abundant in the site and has water on its surface all year round. In either case, local groundwater flow and recharge systems, involving groundwater flowing laterally from surface waters to the aquifer, were suggested. In a site close to the Red River (Nam Du, < 2 km from the river), the high water level of the Red River is considered to raise the HUA groundwater level in the rainy season (Norrman et al. 2008); however, such a phenomenon is not expected in LD, which is much farther from the river (5 km). The HUA water level in LD is higher than in Nam Du, and in 2012 it was below the Red River water level for only 15 days (Fig. S3 of the ESM).

In TM, the temporal change in the HUA groundwater level was less, and more gradual, than in LD. The water level ranged from 0.1–0.8 m above MSL, except for a pulse on September 2014, which was seemingly caused by preceding heavy rain. The HUA remained unconfined throughout the monitoring period. The water level was lowest in the middle of the rainy season (e.g., June–July 2013), and highest at the end of the rainy season or later (e.g., September and October); seasonal trends which lagged behind those of the neighboring pond. No clear response to heavy rains was consistently observed. These results suggest that the HUA groundwater may well be recharged by the neighboring pond, though there was no rapid path by which it was recharged, probably because of the hardly permeable land use and the presence of thick clayey soil (∼5 m thick) above the HUA. In addition, the HUA groundwater level is considered to be affected by the reduced potential of the PCA, as explained in the following. A similar trend, of groundwater level lagging behind the precipitation pattern, was also observed in the HUA in the southern and western suburban areas of Hanoi (Dang et al. 2014).

HPA and PCA monitoring wells

In LD, the fluctuating groundwater level, caused by continuous gas emission from the well, hindered reliable water-level measurement of the HPA well. Nevertheless, the aquifer would be confined throughout the monitoring period, and the groundwater level seemingly had a seasonal trend, with a delayed peak, comparable to that of the HUA (Fig. 5). In a neighboring site (Hoang Liet commune), the PCA groundwater level has a seasonal variation of 1.8 m, which is attributed to the changing water level of the Red River (Berg et al. 2008).

In TM, the PCA groundwater level was 5–6 m lower than the HUA groundwater level, but the water-level fluctuation patterns of the PCA and HUA were synchronized (Fig. 6). The level difference and seasonal change in the respective HUA and PCA groundwater levels were also similar in other locations in the southern and western suburbs of Hanoi (Dang et al. 2014). This temporally similar behavior of the HUA and PCA groundwater levels, together with the stable isotope results, suggested that the two aquifers were hydraulically interconnected, and the HUA groundwater infiltrated into the PCA. However, the thickness of the HPA (>10 m), as revealed by the cores, suggested very little connectivity between the two aquifers in the drilling location (see also section ‘Hydraulic conductivity of cores’). Therefore, the infiltration of HUA groundwater into the PCA is considered to probably occur in areas where the HPA is relatively thin or nonexistent. This suggests the existence of multiple local flow systems in heterogeneous hydrologeological settings, involving lateral flow, rather than a uniform, continuous vertical flow system, as was suggested in Hanoi (Postma et al. 2012). Studies show that the decline in both the PCA and HUA groundwater levels in Hanoi was largest in the southern and western areas, where large-scale pumping stations are located; and that the decline in the HUA groundwater level was partly due to a serious drop in groundwater levels in the PCA, caused by excessive abstraction (Berg et al. 2008; Bui et al. 2012b; Winkel et al. 2011). In the present study in TM, which is southwest of Hanoi’s city center, the observed groundwater-level temporal trends and stable isotope results are in agreement with the conclusions of the abovementioned studies. Regarding the Red River, which governs the groundwater flow and recharge in riverine areas (Berg et al. 2001; Jusseret et al. 2009), the PCA water level fluctuation in TM did not correspond to fluctuation in the river (Fig. S3 of the ESM). The isotopic and hydrochemical results also suggest a minor contribution from the Red River water. This would simply be because the TM site is too far from the Red River, and the water levels and recharge in TM are more affected by local sources (e.g., water from the HUA and HPA, which would have been recharged by ponds and lakes).

Relative contribution from surface-water bodies to groundwater recharge

On the basis of the δ18O and Cl− results, the relative contribution from evaporation-affected surface-water bodies to each groundwater location was estimated, using a simple two end-member mixing model, as shown in Eqs. (2) and (3) in the following:

$$ {\updelta}^{18}{\mathrm{O}}_{\mathrm{groundwater}}=\updelta {}^{18}\mathrm{O}_{\mathrm{source}1}\times a+\updelta {}^{18}\mathrm{O}_{\mathrm{source}2}\times b $$
(2)
$$ {{\mathrm{Cl}}^{-}}_{\mathrm{groundwater}}={{\mathrm{Cl}}^{-}}_{\mathrm{source}1}\times a+{{\mathrm{Cl}}^{-}}_{\mathrm{source}2}\times b $$
(3)

where δ18Ogroundwater and Cl− groundwater are the average values in each groundwater location; δ18Osource 1, Cl− source 1, δ18Osource 2, Cl− source 2 are the average values of two recharge sources; and a and b are the relative recharge contributions from the respective sources at each groundwater location (a + b = 100%, a, b ≥ 0). Mixing of evaporation-affected surface waters and rainwater was considered for the HUA; whereas, for the PCA, both mixing of the HUA and HPA waters, and mixing of the HUA and Red River waters, were considered. Here, the values of neighboring pond water samples were assumed as the values of evaporation-affected surface waters at each location. The HUA water values were represented by the values in neighboring HUA groundwaters. The HPA water values were represented by the values of the HPA pore waters in each site. The δ18O value in rainwater was represented by the weighted average value of precipitation in Hanoi from 2004–2007 (WISER Database 2016). In the case of Cl−, the relevant precipitation data was not available; thus, the average values of the Red River water samples were assumed as the rainwater value.

Both the δ18O and Cl− results (Table S4 of the ESM) suggest that most HUA groundwaters were predominantly (>90%) recharged by evaporation-affected surface waters, while recharge from rainwater was minor (<10%). In the case of the PCA groundwater, the two end-members contributed roughly equally to the recharge when mixing of the HUA and HPA waters was assumed. When mixing of the HUA and Red River waters was assumed, the former made a greater contribution (>64%) than the latter (<36%). In the HPA monitoring well in LD, the groundwater was considered to almost solely consist of HPA water (>95%), based on the δ18O results.

Discussion

Groundwater system in the Hanoi suburbs

The present two-site study reveals the distinctive local groundwater dynamics and recharge systems in the suburban area of Hanoi. The recharge is dominated by evaporation-affected surface waters in the neighborhood, and the flow is generally controlled by locally heterogeneous geology. Based on the results of the study, conceptual groundwater flow models for sites LD and TM are depicted in Fig. 9. In the short term, the water level can be affected by heavy precipitation in the rainy season, depending on local land use. In contrast, the seasonal trend is generally controlled by lateral recharge from surface-water bodies and depression cones caused by excessive abstraction from the PCA. The vertical infiltration of precipitation through the ground surface, which constitutes a significant source of recharge in many cases, here makes only a minor contribution to recharge. Even in a location with abundant permeable surface, vertical infiltration of precipitation may be limited due to the general presence of surface sediments with low hydraulic conductivity. The majority of HUA water is derived from evaporation-affected surface-water bodies such as ponds, lakes, irrigated farmlands and paddy fields, recharged mostly laterally. The groundwater of the PCA is also mainly derived from evaporation-affected surface-water bodies, through multiple local flow systems involving infiltration of HUA groundwater through areas where the HPA is relatively thin or nonexistent. In the areas away (e.g., > 5 km) from the Red River, the river water—which was traditionally considered as the primary source of PCA groundwater along the river (Berg et al. 2001; Jusseret et al. 2009)—is a minor source of recharge for the PCA. In these areas, excessive abstraction from the PCA induces infiltration of HUA water into the PCA, especially in the southern and western suburbs of Hanoi. The previous study, on chemical wastewater tracers in the same sites as in this study (Kuroda et al. 2015), agrees well with the present study; caffeine and carbamazepine were detected in the HUA monitoring wells in LD and TM, and even in the PCA monitoring well in TM, at levels comparable to those in the neighboring ponds and irrigated farmlands. The current study also shares similar results with previous studies, conducted in several other sites in suburban Hanoi, on stable isotopes (Berg et al. 2008), groundwater age (Postma et al. 2012) and hydrogeological conditions (Bui et al. 2012a). Therefore, the groundwater system outlined above is considered a plausible model, in general, for the suburban areas of Hanoi which are more distant from the Red River. The large difference in groundwater levels between the HUA and PCA underlines the severe stress on groundwater resources in suburban Hanoi. In comparison, in nearby rural areas, the groundwater-level difference between the HUA and PCA is small (e.g., 1 m; Dang et al. 2014), probably due to the greater number of surface-water bodies, lesser number of built-up areas, and most importantly, significantly less abstraction from the PCA, in the latter region.

Fig. 9
figure 9

Conceptual groundwater flow model in a LD and b TM

Full size image

Significance of evaporation-affected surface-water bodies for groundwater recharge in the Hanoi suburbs

Similar to the previous studies, evaporation-affected surface-water bodies such as ponds and irrigated farmland, were confirmed as important recharge sources of groundwater, making a greater than 90% contribution to total recharge in the suburban area of Hanoi. The relative recharge amount from such water bodies would vary significantly, depending on the characteristics and spatial distribution of the water bodies. In the HUA monitoring well in LD, the influence of changes in pond water quality on neighboring groundwater quality was observed. On a larger scale, however, a limited number of large and deep ponds may dominate both the groundwater recharge and the groundwater flow systems. Ponds of greater depth would have greater seepage rates than shallow ponds, because the base of deep ponds is nearer to the HUA. The previous study, on four shallow ponds (1–2 m deep) in the Hanoi suburbs, including one of the ponds in the TM site (TMP-A), suggested a downward seepage rate of only 0.12 m/year from the ponds, on the basis of the depth profile of δD and δ18O observed in the pore waters of pond sediments (Kuroda et al. 2013; see also Fig. S4 of the ESM for the data); and this rate is one order of magnitude smaller than the reported rates in experimental fish ponds in Alabama (1.0–4.5 m/year; Boyd 1982) and in India (1.2 m/year; Sharma et al. 2013).

In 2002, water surfaces comprised 3.5%, and agricultural land excluding water surfaces 46% (Anh et al. 2004), of the total area of Hanoi city (921 km2; the city has since expanded to 3,345 km2). Assuming that all the surface waters in Hanoi have a 0.12-m/year seepage rate, and that 3.5% of the current Hanoi city area consists of surface waters, then the total recharge amount from the surface waters would be 38,000 m3/day. In comparison, in 2011, the estimated water consumption in Hanoi’s suburban and rural areas, which are supplied solely from groundwater, was several times larger, at 150,000–200,000 m3/day (Do et al. 2014). Therefore, while the actual seepage rate from surface waters could be larger than the assumed value of 0.12 m/year, there would be a large amount of recharge from other evaporation-affected surface-water bodies such as irrigated farmlands, paddy fields and canals. In areas close to the Red River, infiltration from the river would naturally dominate the recharge. Further studies are required for detailed evaluation of the relative contribution, from ponds and lakes of various sizes, irrigated farmlands and the Red River, to groundwater recharge in various suburban areas of Hanoi.

Effects of decreasing surface-water bodies and land-use change on groundwater recharge: implications for groundwater management

Due to the disappearance of surface-water bodies such as ponds and irrigated farmlands, the suburban area of Hanoi may suffer a long-term, significant decrease in groundwater recharge. As the urban water demand increases, the stress on groundwater will increase even more, leading to groundwater depletion and even more severe land subsidence, unless alternative water resources (e.g., surface water and rainwater) are developed in time. Hence, the surface-water bodies, which make a significant contribution to local groundwater recharge, must be preserved; and countermeasures are needed to enhance groundwater recharge, and to conserve water for domestic consumption. Large and deep ponds and lakes must be preserved, and effective seepage must be maintained through regular dredging. Use of pervious pavement materials can enhance rainwater infiltration to the subsurface. Road runoff can also be infiltrated after proper cleaning. Construction of infiltration ponds (Barbosa and Hvitved-Jacobsen 2001) and recharge wells (Rahman et al. 1969) are also considered effective in enhancing groundwater recharge for the HUA and PCA, respectively. Note that, in introducing such practices to enhance groundwater recharge, care must be taken regarding their potential effects on groundwater pollution; for example, hydrogeological alteration such as excessive abstraction can induce serious groundwater pollution by ammonia and arsenic in Hanoi (Berg et al. 2008; Winkel et al. 2011). The occurrence of arsenic and ammonia in groundwater and sediments in the sites of the present study can be found in the parallel study (Kuroda et al. 2016).

Land use and topography can affect the short-term groundwater level and characteristics of surface runoff in the area, but cannot significantly affect the groundwater recharge. Given that agricultural area, bare land and grassland, which are characterized by permeable surfaces, are still predominant land uses in the suburban areas of Hanoi, the increase in impermeable surfaces due to the construction of built-up areas may have little effect on groundwater recharge in the foreseeable future. However, in areas where the surface clayey deposits are thin or nonexistent (e.g., northwest of Hanoi city center, as reported in Jusseret et al. 2009), rainwater infiltration through the ground surface is considered to be significant; and thus, greater care should be taken to preserve the groundwater recharge here, for example, through the construction of rainwater infiltration facilities.

The ponds and lakes in the Hanoi suburbs are often used for irrigation, fishery and recreational purposes, but they also function in numerous other ways for the benefit of local residents, such as detention of stormwater runoff, and retention and treatment of domestic wastewater. Disposal of domestic wastewater has caused serious deterioration of receiving surface-water bodies and even groundwater, with ammonia, organic matter, heavy metals, polycyclic aromatic hydrocarbons (PAHs), bacteria and viruses being major contaminants (Huong et al. 2008; Hung et al. 2015; Kuroda et al. 2013, 2015; Fuhrimann et al. 2016; Norrman et al. 2015). Further, the bottom sediments of ponds and lakes often contain large amounts of ammonia and organic matter, which may lead to deterioration in the quality of pond seepage (Hung et al. 2015; Kuroda et al. 2013); thus, removal of such bottom sediments may also be useful for improving the quality of recharged groundwater. Note, however, that the removal of bottom sediments may also have adverse effects. Since such sediments may sequester heavy metals in sulfides, and increased residence time may lead to degradation of organic pollutants, removal of the sediments may reduce these positive effects. Therefore, the impact of removing bottom sediments should be carefully studied before altering current practices. Appropriate management of land use, and preservation and maintenance of surface-water bodies, is of great importance for the sustainable use of groundwater in Hanoi, and finally for the sustainable development of the city.

Conclusions

An integrated approach involving water-level monitoring, geological survey, and water-quality analysis of groundwater, sediment pore water and surface water, enabled a detailed evaluation of the groundwater system and recharge mechanisms in the suburban area of Hanoi city. The groundwater for both the HUA and PCA are considered to be primarily recharged, through multiple local flow systems principally involving lateral flows, by evaporation-affected surface-water bodies, such as ponds, lakes, irrigated farmlands and paddy fields. The infiltration of precipitation through the ground surface, on the other hand, seems to make only a minor contribution to groundwater recharge in the area. Thus, the decrease in surface-water bodies will likely account for more of the decrease in groundwater recharge than changes in land use. Further studies, then, should focus on detailed evaluation of these surface-water bodies, in terms of their characteristics (e.g., ponds of varying depths, irrigated farmland, paddy fields) and their respective contributions to groundwater recharge.