Often viewed solely as an environmental concern, radioactive tracers such as anthropogenic tritium can also provide powerful insights into groundwater flow. High-frequency monitoring of tritium enables detailed reconstruction of water transit times in managed aquifer recharge facilities.

Due to its short half-life (87.4 days) and omnipresence, cosmogenic radio-sulfur (S-35) is an attractive tracer for investigating subyearly groundwater residence times. S-35 is transported to the lower atmosphere by large-scale air mass circulation and transferred to groundwater by precipitation. For groundwater dating, the variability of S-35 concentration in precipitation requires a S-35 input function. However, the required S-35 data are often not available. To fill this gap, we present an approach to reconstructing S-35 concentrations in precipitation based on proxy parameters of better availability. The tested parameters include natural Be-7 and H-3, parameters that allow quantifying cosmogenic S-35 production, and parameters that are correlated to the intensity of S-35 washout from the atmosphere. In comparison with an unrivaled 4 year time series of S-35 in precipitation, we discuss the correlations of all parameters with S-35, assess their individual applicability as S-35 proxies, and evaluate the predictive power of joint data sets in varying combination. As a result, we present a modeling approach that allows reconstruction of a S-35 input function with a monthly temporal resolution based on proxy parameters. This novel modeling approach provides a valuable tool for groundwater dating using S-35 as a tracer in studies that lack directly measured S-35 input data

Naturally occurring radio-sulphur (35S) is suitable as an aqueous environmental tracer for the dating of groundwater and surface waters with residence times of less than one year. As a β-decaying radionuclide, 35S is detected using liquid scintillation counting (LSC). When extracting 35S as sulphate from large-volume water samples, there is the possibility of unintentional co-extraction of other naturally occurring radionuclides, which interfere with the measurement of 35S by LSC. The most important of these radionuclides are (i) 226Ra, (ii) the short-lived progeny of 222Rn, (iii) 210Pb and its progeny and (iv) 3H. In addition, 14C, which might be present in scintillation cocktails or LSC plastic vials, and 40K, which is likely to be present in LSC glass vials, can have a significant impact on the LSC detection result. There are a few publications that address sample preparation for 35S detection with LSC. However, the published datasets do not contain sufficiently detailed information to pursue the issue of potential interferences of the said naturally occurring radionuclides with the 35S signal. In our study, we measured standardized samples containing the said radionuclides by LSC, evaluated location, shape and overlap of the associated energy peaks, and assessed the possible influences of the individual nuclides on the 35S detection results. The findings of our study show that when measuring 35S obtained from a natural water sample, counts detected in the 35S energy window cannot be unconditionally interpreted as actual 35S counts. Interfering nuclides (especially 3H, 14C, 40K, and 210Pb) can lead to counts in the 35S energy window and thus to an overestimation of the 35S activity concentration of the water sample (and consequently to an underestimation of the water age). We therefore recommend generally a complete evaluation of the LSC spectrum in order (i) to be able to infer the potential presence of other radionuclides in the measured sample, and (ii) to evaluate the DPM counted in the 35S energy window accordingly, as these may contain counts of the aforementioned interfering radionuclides.

Research on groundwater residence times is essential for evaluating groundwater abstraction rates and aquifer vulnerabilities, and hence, for sustainable water resources management. Naturally occurring radionuclides are suitable tools for related investigations. While the applicability of several long-lived radionuclides for the investigation of long-term processes has been demonstrated frequently, tracer-based approaches for studying residence times of less than one year have not been fully exploited. That is due to the rather small number of applicable radionuclides that show adequately short half-lives. A promising approach for investigating sub-yearly residence times applies radioactive Sulphur (35S). Radio-Sulphur is naturally produced by high-energy cosmic radiation in the upper atmosphere from where it is transferred with precipitation to the groundwater. As soon as the meteoric water enters the subsurface its 35S activity concentration decreases with an 87.4-day half-life. This makes 35S suitable for investigating sub-yearly groundwater residence times. However, the low 35S activities in natural waters require sulphate pre-concentration for 35S detection by means of liquid scintillation counting. This is done by sulphate extraction from large water samples with anion-exchange resins or/and precipitation as BaSO4. The resulting samples are usually associated with background interferences and quenching. The presented experiments aim at (i) optimizing the sample preparation procedure by simplifying the pre-concentration of sulphate to make it applicable for field sampling and at (ii) reducing quench and background during measurement. We will discuss the different sample preparation methods and lessons learned for the detection and quantification of 35S pre-concentrated from natural water samples that contain a wide range of SO42− concentrations.

Due to its short half-life (87 days), naturally occurring radio-sulfur (35S) is applicable as aqueous environmental tracer for investigating groundwater residence times shorter than one year. Being a pure β-decaying radionuclide, 35S is detected straightforwardly by means of liquid scintillation counting (LSC). The rather low 35S activities in natural waters require (i) a careful sample preparation aiming at extracting 35SO42- from large-volume water samples (ca. 20 L) resulting in samples ready for LSC measurement and (ii) an optimal device-specific setup of the LSC to maximize the 35S signal-to-noise-ratio. A few publications that discuss approaches for sample preparation and device-specific LSC setup optimization are available. This paper presents a summarizing step-by-step instruction for both optimized sample preparation and LSC setup. For practical reasons, two different sample preparation approaches are presented, one for samples with low total sulphate inventories (up to 350 mg) and one for samples with elevated total sulphate inventories (350-1500 mg). LSC setup optimization aiming at the measurement of the two resulting types of samples is described for three LSC devices, namely Quantulus GCT, TriCarb 3170 TR/SL, and Quantulus LB 1220.

The inflow of groundwater into a large river can contribute significantly to the total discharge of the river. Especially at low river water level, the inflowing groundwater can strongly influence the river water quality. Therefore, both the localisation and quantification of groundwater inflow into rivers is essential for their sustainable management. Corresponding field investigations usually rely on tracer applications based on stable water isotopes (δ2H/δ18O), naturally occurring radioisotopes (e.g. 222Rn) and standard water parameters (T, pH, EC). However, an extension of this tracer toolbox is desirable. In recent studies, the idea of using anthropogenic tritium that is released from a nuclear power plant (NPP) at an upstream location as an additional tracer has been pursued. The concept assumes that the inflow of (practically tritium-free) groundwater and tributary water dilutes the anthropogenic tritium inventory of the river and thus enables quantification of the total groundwater/tributary inflow ("tritium dilution concept"). Since the inflow of tributaries is usually easy to quantify, it is suggested that the inflow of groundwater can be derived from the tritium data. In the presented pilot study, we investigated the applicability of this "tritium dilution concept" and evaluated the general challenges associated to the uncertainty of input parameters. The study was executed exemplarily at an extended river section affected by the release of NPP process water (Vltava/Elbe River, Czech Republic). The evaluation of two 1-month tritium timeseries recorded at two monitoring/gauging stations 168 km apart allowed quantifying the total groundwater/tributary inflow into the river within this section based on the tritium data with a precision that was close to the gauged increase in river discharge (+72 % and +77 %, respectively). On the other hand, the groundwater inflow that was derived after subtracting the gauged tributary inflow from the quantified total water inflow was comparably low and therefore within the approaches' range of uncertainty. From this it can be concluded that the "tritium dilution concept" is applicable if the required input parameters, namely the natural tritium groundwater/tributary endmember and the tributary discharge, are known with sufficient precision. This outcome is promising for studies that focus on less extended river sections with input parameters that are easier to define.

Fundamental approaches to the study of groundwater rely on investigating the spatial and temporal distribution of stable and radioactive isotopes and other anthropogenic compounds in natural waterbodies. The most often used tracers for estimating groundwater flow paths and residence times, groundwater/surface water interaction as well as tracing chemical (contamination) sources include stable isotopes of water (δ 18O and δ 2H), radiocarbon (14C; t1/2 = 5730 a), tritium (3H; t1/2 = 12.43 a) as well as unreactive fluorine-containing gases (e.g., chlorofluorocarbons CCl3F or CFC-11; CCl2F3 or CFC-12; C2Cl3F3 or CFC-113; and SF6). While gas tracers are usually referred to as transient tracers and are appropriate for investigating modern flow systems, the isotopic tracers are often used to investigated paleo or regional flow systems. Stable isotopes of water can also be used to investigate groundwater/surface water interactions. Another, thus far been less frequently used group of groundwater tracers, are cosmo- and geo- genic short-lived radioisotopes. These isotopes are uniquely suited for studying a wide range of groundwater problems that have short time scales including high aquifer vulnerability to quantitative and qualitative impacts and groundwater discharge to surface waters. Here, we discuss and compare the applications of radio‑sulphur (35S; half-life t1/2 = 87 d), radio‑beryllium (7Be; t1/2 = 53 d), radio‑phosphorus (32/33P; combined t1/2 = 33 d), natural tritium (3H; t1/2 = 12.43 a), radon (222Rn; t1/2 = 3.8 d) and short-lived radium (224/223Ra; combined t1/2 = 5.2 d). The paper discusses the principles of the individual tracer methods, focusing on the isotopes' input functions or values, on sampling techniques, and on methods of analyses. Case studies that applied a combined use of the tracers are referred to for readers who wish to learn more about the application of the so far underused cosmo- and geo- genic radioisotopes as aquatic tracers.

Cosmogenic radio-sulphur (35S) is applicable as aqueous environmental tracer for investigating sub-yearly groundwater residence times. For the purpose 35SO4 has to be extracted from large-volume water samples (ca. 20 l) and measured by liquid scintillation counting (LSC). Publications that discuss sample preparation approaches focus on waters with low or moderate sulphate concentrations based on sulphate extraction using an ion-exchange resin. However, sulphate extraction by ion-exchange is not feasible for water samples that contain total sulphate loads of over about 1500 mg. Our paper presents an approach for the preparation of water samples with sulphate loads of up to 6000 mg based on BaSO4 precipitation directly from the sample. The key challenge of this approach is the co-precipitation of interfering natural radionuclides (226Ra, short-lived progeny of 222Rn, 210Pb + progeny, 32/33P), which complicates LSC measurement of 35S. Two options are discussed to allow either pre-precipitating the unwanted radionuclides before, or keeping them in solution during Ba35SO4 precipitation.

Soluble reactive phosphorus (SRP) concentrations in agricultural headwaters can display pronounced seasonal variability at low flow, often with the highest concentrations occurring in summer. These SRP concentrations often exceed eutrophication levels, but their main sources, spatial distribution, and temporal dynamics are often unknown. The purpose of this study is therefore to differentiate between potential SRP losses and releases from soil drainage, anoxic riparian wetlands, and stream sediments in an agricultural headwater catchment. To identify the dominant SRP sources, we carried out three longitudinal stream sampling campaigns for SRP concentrations and fluxes. We used salt dilution tests and natural 222Rn to determine water fluxes in different sections of the stream, and we sampled for SRP, Fe, and 14C dissolved organic carbon (DOC) to examine possible redox-mediated mobilization from riparian wetlands and stream sediments. The results indicate that a single short section in the upper headwater reach was responsible for most of the SRP fluxes to the stream. Analysis of samples taken under summer low-flow conditions revealed that the stream water SRP concentrations, the fraction of SRP within total dissolved P (TDP), and DOC radiocarbon ages matched those in the groundwater entering the gaining section. Pore water from the stream sediment showed evidence of reductive mobilization of SRP, but the exchange fluxes were probably too small to contribute substantially to SRP stream concentrations. We also found no evidence that shallow flow paths from riparian wetlands contributed to the observed SRP loads in the stream. Combined, the results of this campaign and previous monitoring suggest that groundwater is the main long-term contributor of SRP at low flow, and agricultural phosphorus is largely buffered in the soil zone. We argue that the seasonal variation of SRP concentrations was mainly caused by variations in the proportion of groundwater present in the streamflow, which was highest during summer low-flow periods. Accurate knowledge of the various input pathways is important for choosing effective management measures in a given catchment, as it is also possible that observations of seasonal SRP dilution patterns stem from increased mobilization in riparian zones or from point sources.

Radium-226 detection in sediment samples is generally executed by means of gamma-ray spectrometry. Data evaluation relies (besides the 186.2 keV Ra-226 gamma peak) on the combined analysis of major gamma peaks that are produced by the short-lived radon (Rn-222) daughters Pb-214 and Bi-214. Precondition for this detection approach is equilibrium decay of all members of the decay chain between Ra-226 and Bi-214. In closed systems, this equilibrium is reached after about five half-lives of Rn-222 (19 days). However, a closed system can only be guaranteed if the capsule which contains the sample prevents diffusive escape of radon. Such radon-tightness cannot be guaranteed for a wide range of plastic materials. Due to its polymer structure, plastic material generally tends to allow radon diffusion and hence radon loss from the sample resulting in a disturbance of the required decay equilibrium. The paper introduces an approach that allows quantifying radon loss from sample capsules by direct radon measurements using mobile radon detection equipment. The experimental findings are supported by theoretical considerations. An examined alternative approach based on the offset of the 186.2 keV data point from an efficiency function that is calculated exclusively from short-lived radon progeny peaks in the gamma-ray spectrum did not prove to be applicable due to a lack of supporting peaks in the low-energy section of the spectrum.

In many cases, shallow coastal lagoons are, on the one hand, vulnerable habitats for birds and marine ecosystems and, on the other hand, threatened by discharging nutrient-laden surface waters and groundwater. In particular, the localization and quantification of submarine groundwater discharge (SGD) is of key concern in this regard. The presented study aimed at investigating SGD into a vulnerable coastal lagoon that is strongly impacted by evaporation applying a multi-tracer approach. The joint application of radionuclides ( 222 Rn, 223 Ra, 224 Ra), stable water isotopes (δ 18 O, δ 2 H) and the water salinity as environmental water tracers allowed evaluating the suitability of the individual parameters in this specific type of environment. Whilst stable isotope and salinity data were difficult to construe in terms of SGD occurrence due to the intense impact of evaporation, a radon mass balance allowed localising SGD areas within the lagoon and quantifying the related SGD flux rates. In addition, a 224 Ra/ 223 Ra ratio analysis revealed information on the apparent age of the discharged groundwater, and hence on the flushing intensity of the lagoon. Besides these site-specific results, the study allowed the following general conclusions regarding the suitability of the applied tracers: (i) we verified the suitability of a radon mass balance approach for proving/disproving SGD occurrence and quantifying SGD fluxes in shallow coastal lagoons strongly impacted by evaporation; (ii) we showed that the impact of evaporation may impede the use of water stable isotope and salinity data as SGD indicators in such specific environments; (iii) we demonstrated that the tidal impact on a lagoon water body during a sampling campaign can be compensated by adapting sampling schedule and cruise track to the tidal cycle.

Study region: Our study region extends over 450 stream km of the German part of the Elbe River, an ecologically and economically important first order river, between Scho center dot na and Wittenberge. Study focus: Diffuse groundwater born nutrients are major contributors to increased algae growth in rivers, leading to eutrophication with serious consequences for water quality and ecosystem health. Therefore, knowledge of the spatial and temporal dynamics of diffuse groundwater discharge are required since groundwater often remains as a 'black box' for the identification of nutrient sources by managers. The multi-method approach, based on the inverse geochemical and tritium modelling, a flux balance, a darcy approach and hydraulic gradients, showed complex spatiotemporal dynamics along the studied reach of the Elbe River. Groundwater inflow was variable but occurred along the entire river. Areas of high groundwater fluxes were located in the upstream mountainous catchment areas and decreasing downstream. New hydrological insights for the region: The multi-method approach provides a blueprint for the assessment of other large river systems. No single method was able to create conclusive results and most other approaches are only applicable in smaller stream systems. First time an estimation of groundwater flux rates, that can be used to quantify matter inputs, was made. In addition, we showed a way to detect and assess the impact of drainage channels in a heterogenous river system.

Mapping radon (222Rn) distribution patterns in the coastal sea is a widely applied method for localizing and quantifying submarine groundwater discharge (SGD). While the literature reports a wide range of successful case studies, methodical problems that might occur in shallow wind-exposed coastal settings are generally neglected. This paper evaluates causes and effects that resulted in a failure of the radon approach at a distinct shallow wind-exposed location in the Baltic Sea. Based on a simple radon mass balance model, we discuss the effect of both wind speed and wind direction as causal for this failure. We show that at coastal settings, which are dominated by gentle submarine slopes and shallow waters, both parameters have severe impact on coastal radon distribution patterns, thus impeding their use for SGD investigation. In such cases, the radon approach needs necessarily to allow for the impact of wind speed and wind direction not only during but also prior to the field campaign.

The ubiquitous presence of the radioisotope radon (222Rn) and its short-lived progeny (218Po, 214Pb, 214Bi, 214Po) is challenging in two respects: (i) Radon is a major issue regarding health-related problems due to potentially elevated radiation exposure of humans in dwellings, and (ii) due to the mobility of radon the short-lived progeny may cause complications in radionuclide detection in laboratories. Polymer membranes are an appropriate means for effectively preventing unwanted radon migration. However, most of the published literature focusses on robust membranes made for the large-scale sealing of dwelling substructures. Membranes that are suitable (at small-scale) for sealing purposes in radionuclide detection applications are only rarely discussed. In this paper, we present a straightforward practical approach that allows the effectiveness of any membrane to be assessed for any purpose related to radon sealing. Executing the approach requires only (i) a suitable container with inlet and outlet ports, (ii) a mobile radon detector, and (iii) any type of radon source material. The approach provides a tool that allows testing any available membrane for its applicability as radon barrier sheeting.

Within the Ararat Valley (Armenia), a continuously growing water demand (for irrigation and fish farming) and a simultaneous decline in groundwater recharge (due to climate change) result in increasing stress on the local groundwater resources. This detrimental development is reflected by groundwater-level drops and an associated reduction of the area with artesian conditions in the valley centre. This situation calls for increasing efforts aimed at more sustainable water resources management. The aim of this baseline study was the collection of data that allows for study on the origin and age distribution of the Ararat Valley groundwater based on environmental tracers, namely stable (δ 2 H, δ 18 O) and radioactive ( 35 S, 3 H) isotopes, as well as physical-chemical indicators. The results show that the Ararat Valley receives modern recharge, despite its (semi-)arid climate. While subannual groundwater residence times could be disproved ( 35 S), the detected 3 H pattern suggests groundwater ages of several decades, with the oldest waters being recharged around 60 years ago. The differing groundwater ages are reflected by varying scatter of stable isotope and hydrochemical signatures. The presence of young groundwater (i.e., younger that the 1970s), some containing nitrate, indicates groundwater vulnerability and underscores the importance of increased efforts to achieve sustainable management of this natural resource. Since stable isotope signatures indicate the recharge areas to be located in the mountains surrounding the valley, these efforts must not be limited to the central part of the valley where most of the abstraction wells are located.

Investigations in hydrology and hydrogeology are often hampered by a lack of parameters that permit direct observation or monitoring of the processes of interest [...]

The eutrophic Lake Eichbaumsee, a ~1 km long and 280 m wide (maximum water depth 16 m) dredging lake southeast of Hamburg (Germany), has been treated for water quality improvements using various techniques (i.e., aeration plants, removal of dissolved phosphorous by aluminum phosphorous precipitation, and by Bentophos® (Phoslock Environmental Technologies, Sydney, Australia), adsorption) during the past ~15 years. Despite these treatments, no long-term improvement of the water quality has been observed and the lake water phosphorous content has continued to increase by e.g., ~670 kg phosphorous between autumn 2014 and autumn 2019. As no creeks or rivers drain into the lake and hydrological groundwater models do not suggest any major groundwater discharge into the lake, sources of phosphorous (and other nutrients) are unknown. We investigated the phosphorous fluxes from sediment pore water and from groundwater in the water body of the lake. Sediment pore water was extracted from sediment cores recovered by divers in August 2018 and February 2019. Diffusive phosphorous fluxes from pore water were calculated based on phosphorus gradients. Stable water isotopes (δ2H, δ18O) were measured in the lake water, in interstitial waters in the banks surrounding the lake, in the Elbe River, and in three groundwater wells close to the lake. Stable isotope (δ2H, δ18O) water mass balance models were used to compute water inflow/outflow to/from the lake. Our results revealed pore-water borne phosphorous fluxes between 0.2 mg/m2/d and 1.9 mg/m2/d. Assuming that the measured phosphorous fluxes are temporarily and spatially representative for the whole lake, about 11 kg/a to 110 kg/a of phosphorous is released from sediments. This amount is lower than the observed lake water phosphorous increase of ~344 kg between April 2018 and November 2018. Water stable isotope (δ2H, δ18O) compositions indicate a water exchange between an aquifer and the lake water. Based on stable isotope mass balances we estimated an inflow of phosphorous from the aquifer to the lake of between ~150 kg/a and ~390 kg/a. This result suggests that groundwater-borne phosphorous is a significant phosphorous source for the Eichbaumsee and highlights the importance of groundwater for lake water phosphorous balances.

Cosmogenic radiosulfur (35S; half-life: 87.4 days) is transferred with the rain to the groundwater (as 35SO42−) and can be used as residence time tracer for the detection of sub-yearly groundwater residence times. Due to the distinct but non-regular annual 35S pattern in precipitation, related data evaluation requires consideration of a35S input function that is based on representative rain samples. While minor rain events can easily be sampled quantitatively and hence representatively, a long-lasting rain event may get documented by a sample that represents only a certain sequence of the event, thus potentially resulting in a35S activity concentration that might not be representative. With the aim to examine the magnitude of temporal variations of the 35S activity concentration in rain during long-lasting rain events, we present and discuss two related exemplary 35S time series. Furthermore, we evaluate the applicability of the parameters total sulfate and electrical conductivity (EC), both detected in rainwater as easily attainable 35S proxies. The results of the study show (i) that the 35S activity concentration may vary substantially during long-lasting rain events due to atmospheric migration processes and aerosol washout and (ii) that neither sulfate nor EC are suitable as 35S proxies due to the different origin of 35S on the one hand and sulfate/EC on the other. Hence, for the determination of a35S activity concentration that is representative for a long-lasting rain event 35S analyses of an adequate number of sub-samples cannot be avoided.

Investigation of river water/groundwater interaction aims generally at: (i) localizing water migration pathways; and (ii) quantifying water and associated matter exchange between the two natural water resources. Related numerical models generally rely on model-specific parameters that represent the physical conditions of the catchment and suitable aqueous tracer data. A generally applicable approach for this purpose is based on the finite element model FINIFLUX that is using the radioactive noble gas radon-222 as naturally occurring tracer. During the study discussed in this paper, radon and physical stream data were used with the aim to localize and quantify groundwater discharge into a well-defined section of a small headwater stream. Besides site-specific results of two sampling campaigns, the outcomes of the study reveal: (i) the general difficulties of conducting river water/groundwater interaction studies in small and heterogeneous headwater catchments; and (ii) the particular challenge of defining well constrained site- and campaign-specific values for both the groundwater radon endmember and the radon degassing coefficient. It was revealed that determination of both parameters should be based on as many data sources as possible and include a critical assessment of the reasonability of the gathered and used datasets. The results of our study exposed potential limitations of the approach if executed in small and turbulent headwater streams. Hence, we want to emphasize that the project was not only executed as a case study at a distinct site but rather aimed at evaluating the applicability of the chosen approach for conducting river water/groundwater interaction studies in heterogeneous headwater catchments.

Information about groundwater residence times is essential for evaluating appropriate groundwater abstraction rates and aquifer vulnerabilities and hence for sustainable groundwater management in general. Naturally occurring radionuclides are suitable tools for related investigations. While the applicability of several long-lived radionuclides for the investigation of long-term processes has been demonstrated frequently, residence times of less than one year are only scarcely discussed in the literature. That is due to the rather small number of applicable radionuclides that show adequately short half-lives. A promising approach for investigating sub-yearly residence times applies radioactive sulphur. S-35 is continuously produced in the upper atmosphere from where it is transferred with the rain to the groundwater. As soon as the water enters the subsurface its S-35 activity concentration decreases with an 87.4 day half-life. This makes S-35 suitable for investigating sub-yearly groundwater residence times. However, the low S-35 activities in natural waters require sulphate preconcentration for S-35 detection by means of liquid scintillation counting (LSC). That is usually done by sulphate extraction from large water samples with an anion-exchange resin (Amberlite IRA400, Cl-form), elution from the resin with NaCl, and precipitation as BaSO4. Our study aimed at optimizing the standard sample preparation procedure by avoiding the laborious precipitation step. We suggest (i) sulphate extraction using the exchange resin Amberlite IRA67 (OH-form), (ii) elution with ammonium hydroxide, (iii) evaporation of the eluate and (iv) dissolving the resulting dry precipitate in 2 ml H2O. In contrast to the standard approach our method results in a final sample solution of low ionic strength, which allows applying the water miscible scintillation cocktail Hionic-Fluor. Since Hionic-Fluor accepts only aqueous solutions of low ionic strength the approach is applicable for waters with high S-35/(SO42-)-S-32 ratios, i.e., low total sulphate sample loads (e.g. rainwater).