Determination of Trace Mercury in Environmental and Foods Samples

Recently, flow injection (FI) online sorption preconcentration and separation ... for trace mercury determination in environmental samples by online c...
2 downloads 0 Views 88KB Size
Download PDF
Environ. Sci. Technol. 2002, 36, 4886-4891

Determination of Trace Mercury in Environmental and Foods Samples by Online Coupling of Flow Injection Displacement Sorption Preconcentration to Electrothermal Atomic Absorption Spectrometry YAN LI, YAN JIANG, AND XIU-PING YAN* Central Laboratory, Research Center for Analytical Sciences, College of Chemistry, Nankai University, Tianjin 300071, China ZHE-MING NI Research Center for Eco-Environmental Sciences, Academia Sinica, P.O. Box 2871, Beijing 100083, China

The toxic effects of mercury are well-known. To establish sources of mercury contamination and to evaluate levels of mercury pollution, sensitive, selective, and accurate analytical methods with excellent reproducibility are required. We have developed a novel methodology for the determination of trace mercury in environmental and foods samples by online coupling of flow injection (FI) displacement sorption preconcentration in a knotted reactor (KR) to electrothermal atomic absorption spectrometry (ETAAS). The developed methodology involved the online formation of copper pyrrolidine dithiocarbamate (Cu-PDC), presorption of the resulting Cu-PDC onto the inner walls of the KR, and selective retention of the analyte Hg(II) onto the inner walls of the KR through online displacement reaction between Hg(II) and the presorbed Cu-PDC. The retained analyte was subsequently eluted by 50 µL of ethanol and online detected by ETAAS. Interferences from coexisting heavy metal ions with lower stability of their APDC complexes relative to Cu-PDC were minimized without the need of any masking reagents. The tolerable concentrations of Cu(II), Cd(II), Fe(III), Ni(II), and Zn(II) were up to 12, 20, 16, 20, and 60 mg L-1, respectively. No additional chemical modifiers for the stabilization of mercury were required in the present system owing to the stability of Hg-PDC at the drying stage, and no pyrolysis stage was necessary due to the effective removal of the matrices. With consumption of 2.5 mL of sample solution, an enhancement factor of 91 was obtained in comparison with direct injection of 50 µL of aqueous solution. The relative detection limit (3s) was 6.2 ng L-1, corresponding to an absolute detection limit of 15.5 pg. The precision (RSD, n ) 13) was 1.1% at the 2 µg L-1 level. The method was successfully applied to the determination of mercury in several certified environmental and foods reference materials and locally collected water samples. * Corresponding author phone: (86)22-23508724; fax: (86)2223503034; e-mail: [email protected]. 4886

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 22, 2002

Introduction Mercury is an environmentally and toxicologically important element (1, 2). It is considered by the Environmental Protection Agency (EPA) as a highly dangerous element because of its accumulative and persistent character in the environment and biota (3). Although mercury is not an abundant chemical element in nature, it has become widespread as a result of many industrial and agricultural applications (3-6). Consequently, the development of reliable methods for the determination of mercury at (ultra)trace levels in environmental and biological materials is of particular significance. Among the most widely used methods for mercury determination, cold vapor atomic absorption spectrometry (CVAAS) (7) as well as cold vapor atomic fluorescence spectrometry (CVAFS) (8) has received great attention owing to the simplicity and high sensitivity (9-12). To lower the detection limit and to improve the sensitivity further, the combination of online noble metal amalgamation preconcentration with CVAAS or CVAFS has been investigated (13). However, there are two disadvantages with this amalgamation preconcentration technique. First, the efficiency of mercury collection may be impaired by moisture or other gaseous reaction products that poison the surface of the amalgamation medium, necessitating occasional cleaning. Second, during heating of the collector to release the mercury, a gas flow is used to transport the vapor to the absorption cell, which means that the sensitivity is flow-rate limited. Slight changes in the flow-rate between measurements will also impair the reproducibility (13). Electrothermal atomic absorption spectrometry (ETAAS) is one of the most sensitive techniques for trace mercury determination in addition to CVAAS and CVAFS. For direct ETAAS determination of mercury, various chemical modifiers have to be employed for mercury stabilization in the graphite furnace to ensure the efficient measurement (14, 15). Owing to matrix interference and/or insufficient detection power, however, direct determination of (ultra)trace amounts of mercury in complicated matrices is difficult, and a preliminary preconcentration and separation step is usually mandatory. To improve the sensitivity, the combination of ETAAS with cold vapor generation and in situ concentration of mercury vapor onto the pretreated graphite tubes with various metal salts have been developed (12, 16). Recently, flow injection (FI) online sorption preconcentration and separation coupled with ETAAS has been successfully applied to the determination of (ultra)trace elements in a variety of sample matrices (17-20). Online sorption separation and preconcentration overcome the insufficient detection power of the analytical techniques and the matrix effects as well as the shortcomings of batch-wise operation. However, to our knowledge, no such online coupling of FI sorption preconcentration and separation to ETAAS for the determination of (ultra)trace mercury has been reported. This work was undertaken to develop a highly sensitive and selective method with excellent precision for trace mercury determination in environmental samples by online coupling of FI displacement sorption preconcentration in a knotted reactor (KR) to ETAAS. Selective preconcentration of mercury and its separation from matrices were achieved through online displacement reaction between the Hg(II) and the presorbed copper pyrrolidine dithiocarbamate (CuPDC) on the inner walls of the KR due to the stability of Hg-PDC > Cu-PDC. Interferences from coexisting heavy metal ions with lower stability of their PDC complexes relative 10.1021/es025820l CCC: $22.00

 2002 American Chemical Society Published on Web 10/08/2002

TABLE 1. Graphite Furnace Temperature Program for Determination of Mercury step

temperature/°C

1 (drying) 2 (atomization) 3 (cleaning)

50-100 450 1800

time/s

argon flow/mL min-1

40 5 3

150 0 150

to Cu-PDC, which was encountered in conventional FI sorption preconcentration systems, was effectively minimized. No additional chemical modifiers for the stabilization of mercury were required in the proposed system owing to the stability of Hg-PDC at the drying stage, and no pyrolysis stage was necessary due to the effective removal of the matrices. The proposed method was successfully applied to the determination of trace mercury in environmental and foods samples.

Experimental Section Instrumentation. All measurements were performed on a Hitachi Model 180-80 polarized Zeeman atomic absorption spectrometer fitted with a Hitachi autosampler (Part No. 1709608-1) and a Hitachi graphite cup atomizer (Part No. 1807402). A high-intensity mercury hollow cathode lamp (Ningqiang Light Sources Co. Ltd., Hengshui, China) was used as the radiation source at the 253.7 nm wavelength with a slit width of 2.6 nm. Integrated absorbance (peak area) was used for quantitation. Table 1 shows the graphite furnace temperature program used for the determination of mercury. A Model FIA-3100 flow injection system (Vital Instruments Co. Ltd, Beijing, China) was connected to the atomic absorption spectrometer with the shortest possible length of 0.35-mm i.d. PTFE tubing (ca. 20 cm). Ismaprene pump tubes were used to deliver the samples and reagents. Small-bore (0.35-mm i.d.) PTFE tubing were used for all connections, which were kept as short as possible to minimize the dead volumes. The FIA-3100 consists of two peristaltic pumps and a standard rotary injection valve (eight ports on the rotor, and eight ports on the stator). The rotation speed of the two peristaltic pumps, their stop and go intervals, and the actuation of the injection valve were programmed (see Table 2). The KRs used for the preconcentration were laboratorymade of 0.5-mm i.d. PTFE tubing by tying interlaced knots (21, 23). Reagents. All the chemicals were of the highest purity available. Doubly deionized water (DDW) was employed exclusively. A standard stock solution of 10 mg L-1 of Hg(II) was the chloride (The Second Chemical Co., Beijing, China). Working standard solutions were prepared daily by appropriate dilution of the stock solution with diluted H2SO4 (Tianjin Chemicals Co., Tianjin, China). The complexing agent solution was prepared by dissolving ammonium pyrrolidine dithiocarbamate (APDC) (Sigma-Aldrich, Ontario, Canada) in DDW. The copper solution used to react with

APDC was prepared by proper dilution of the stock solution of 1000 mg L-1 of Cu(II) (National Research Center for Standard Materials, Beijing, China). Samples. The following certified reference materials (CRMs) (National Research Center for Standard Materials, Beijing, China) were analyzed to demonstrate the accuracy of the present method: GBW 07310 (marine sediment), GBW 08572 (prawn), GBW 08508 (rice flour), and GBW(E) 080392 (simulated natural water). Two river water, two lake water, and a tap water samples were collected locally. Immediately after sampling, all water samples were filtered through 0.45-µm Supor filters (Gelman Sciences). The filtered samples were then acidified to the optimal pH range for the displacement sorption preconcentration with diluted sulfuric acid, and the resultant samples were stored at 4 °C in low-density polyethylene (LDPE) bottles, which were precleaned with sub-boiling 8 mol L-1 HNO3 and rinsing with DDW. Sample Pretreatment. The apparatus for wet decomposition was a Kjeldahl flask with a reflex condenser system, as used in a British standard method (BS EN 1122:2001) (23). A certain amount of the sample (0.2500 g for the marine sediment and prawn, 1.000 g for the rice flour) was weighed accurately and mixed with 5 mL of sulfuric acid and 5 mL of nitric acid in a Kjeldahl flask. The mixture in the flask was gently heated on a hot plate and kept to sub-boiling for 30 min. After cooling the solution, the contents were transferred into a 50-mL calibrated flask and diluted to volume with DDW. Procedures. The developed FI manifold for the two different valve positions is shown in Figure 1. Details of the duration and function of each step are given in Table 2. A complete cycle of the separation and preconcentration required 164 s with a sample loading time of 40 s. In step 1 (Figure 1a), the injector valve was in the inject position and pump 2 was activated, so that the Cu-PDC was formed online and sorbed onto the inner walls of the KR. In step 2 (Figure 1b), the injector valve was turned to the fill position, but pump 1 was activated while the electromagnetic valve was turned off to introduce an air segment. In step 3 (Figure 1b), the injector valve position and the status of pump 1 were the same as in step 2, whereas the electromagnetic valve was turned on to load the sample or the standard solution into the KR. In this step, the Hg(II) replaced the Cu (II) from the presorbed Cu-PDC and retained on the inner walls of KR. In step 4 (Figure 1b), pump 2 was activated but pump 1 was stopped so that the KR was rinsed with 0.001% sulfuric acid in the counter direction to remove residual matrix from the KR. In step 5 (Figure 1b), as in step 2, air was again introduced to remove residual solution from the KR, the eluent loop (EL), and the delivery tube (DT). The ethanol drawn in this step flew back to the eluent container (EC) to minimize the reagent consumption. In step 6 (Figure 1a), the status of the two pumps and the electromagnetic valve was the same as in step 5, but the injector valve was turned to the inject

TABLE 2. Operational Sequence of FI Online Displacement Sorption of Preconcentration System Coupled with ETAAS for Determination of Trace Mercury flow rate/mL min-1

valve position step

function

1 (Figure 1a) 2 (Figure 1b) 3 (Figure 1b) 4 (Figure 1b) 5 (Figure 1b) 6 (Figure 1a) 7 (Figure 1b) 8 (Figure 1b) 9 (Figure 1a)

presorption air segmentation displacement sorption rinse KR remove residue rinser fill the eluent loop capillary into the graphite tube elution and eluent introduction capillary into the waste

injector electromagnetic duration/s inject fill fill fill fill inject fill fill inject

off off on off off off off off off

20 5 40 20 25 6 4 40 4

medium pump 0.5 mg L-1 Cu 0.03% m/m APDC air sample 0.001% v/v H2SO4 air ethanol air

pump 1

pump 2

off 3.3 3.6 off 3.3 3.3 off 2.7 off

VOL. 36, NO. 22, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

3.8 off off 3.6 off off off off off

9

4887

FIGURE 2. Effect of Cu concentration on the integrated absorbance of 2 µg L-1 Hg. All other conditions as in Figure 1 and Table 2.

FIGURE 1. FI manifold for the online displacement sorption preconcentration for ETAAS. P1 and P2, peristaltic pumps; DT, delivery tubing; EC, eluent container; EL, eluent loop; KR, knotted reactor (0.35-mm i.d. × 100-cm long PTFE tubing); W, waste; EV, electromagnetic valve; “off” for air flow, and “on” for sample solution; ETA, electrothermal atomizer. Valve position: (A) inject; and (B) fill. position to fill the eluent loop with ethanol. In step 7, the autosampler arm automatically moved the tip of the delivery capillary into the dosing hole of the graphite cup. In step 8, air was drawn to propel the ethanol to elute the sorbed analyte and to transport the eluate into the graphite cup. Finally, the autosampler arm moved back to the wash position, and the graphite furnace temperature program was started for ETAAS determination.

Results and Discussion Consideration of an FI Online Displacement Sorption Preconcentration System for ETAAS. In a preliminary study of this work, a conventional FI online KR sorption preconcentration involving formation and sorption of the Hg-PDC in the KR via direct merging of the Hg(II) and APDC solutions was carried out. Soon, we found that such a system suffered serious interferences from other heavy metals owing to their competition for the complexing reagent and/or active sites on the inner walls of KR, as observed in previous FI online KR sorption preconcentration systems for ETAAS determination of antimony(III) (23), lead (24), and cobalt (25), which made it impossible to apply simple aqueous standard solution calibration for the determination of trace mercury in relatively complex matrices. To overcome the above-mentioned problem, it was decided to design an FI displacement sorption preconcentration system coupled with ETAAS for trace mercury determination. The present online displacement sorption system employed Cu-PDC instead of APDC as the reagent. Selective preconcentration of trace mercury and its separation from matrices were achieved through the online displacement reaction between the analyte Hg(II) and the presorbed Cu-PDC on the inner walls of the KR due to the stability of Hg-PDC > Cu-PDC. In this way, it is impossible for heavy metals with lower stability of their PDC complexes 4888

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 22, 2002

relative to Cu-PDC to compete for the reagent and the active sites on the inner walls of KR. To fulfill the proposed online displacement sorption preconcentration protocol, the FI manifold was designed such that the complex Cu-PDC was formed through online merging of the APDC and Cu solutions, and the resultant Cu-PDC was then presorbed onto the inner walls of the KR. After removing the residual APDC and Cu solutions in the KR with an air flow, the sample solution was loaded into the KR, and the Hg(II) in the sample solution was then retained onto the KR due to a displacement reaction between the Hg(II) and the presorbed Cu-PDC on the inner surfaces of the KR. Meanwhile, those coexisting ions that could not replace the Cu(II) from presorbed Cu-PDC passed through the KR to waste. To facilitate the online coupling of the present FI displacement sorption preconcentration system to ETAAS for determination of mercury, previously developed “airsegmented and air-transported operational sequence” (23, 24) was integrated into the present FI manifold design. As a result, the preconcentrated analyte was quantitatively eluated with 50 µL of ethanol, and all the eluate was introduced into the graphite cup completely without any problems. Online Formation and Presorption of Cu-PDC. The concentrations of Cu (II) and APDC are the two important factors controlling the presorption of Cu-PDC on the inner walls of the KR and in turn affecting the subsequent displacement sorption preconcentration of Hg. Figure 2 shows the effect of Cu concentration on the integrated absorbance of 2 µg L-1 Hg. The integrated absorbance of mercury significantly increased as the concentration of Cu increased to 0.2 mg L-1 and then leveled off until 5 mg L-1 of Cu, the maximum Cu concentration tested in this work. The effect of APDC concentration on the integrated absorbance of mercury was examined at 0.5 mg L-1 of Cu. As illustrated in Figure 3, the optimal APDC concentration ranged from 0.02 to 0.1% m/m. Below 0.02% m/m APDC, the integrated absorbance of mercury increased with increasing APDC concentration. For further experiment, 0.5 mg L-1 of Cu and 0.03% m/m APDC were employed during online formation and presorption of Cu-PDC in the KR. The effect of pH on the Cu-PDC formation and its subsequent sorption onto the inner walls of the KR was investigated with an APDC concentration of 0.03% m/m. As shown in Figure 4, the optimum pH of the effluent for presorption was quite wide, ranging from 1.4 to 7.8. For further work, a pH of 5.0 in the effluent was employed for presorption, corresponding to pH 3.8 in the copper solution. Displacement Sorption Preconcentration of Hg(II). The pH of sample solution plays an important role in the displacement sorption preconcentration because it would affect the stability of the presorbed Cu-PDC, the retention

FIGURE 3. Effect of APDC concentration on the integrated absorbance of 2 µg L-1 Hg. All other conditions as in Figure 1 and Table 2.

FIGURE 4. Effect of pH of the presorption effluent on the integrated absorbance of 2 µg L-1 Hg. All other conditions as in Figure 1 and Table 2.

FIGURE 5. Effect of the pH of sample solution on the integrated absorbance of 2 µg L-1 Hg. All other conditions as in Figure 1 and Table 2. of Hg-PDC, and/or the displacement reaction. Figure 5 depicts the effect of the pH of the sample solution on the integrated absorbance of 2 µg L-1 Hg(II). Clearly, the absorbance of mercury sharply increased with increasing of the pH of sample solution up to 0.9, and remained constant until pH 3.5, but decreased with further increase of pH. The relatively wide optimal pH range (0.9-3.5) for sample solution

is an advantage of the present system for analytical application. The effect of sample loading flow rate on the integrated absorbance of mercury was investigated with a standard solution of 2 µg L-1 Hg for 40 s preconcentration. It was found that the integrated absorbance of mercury increased linearly up to a sample loading flow rate of about 4.5 mL min-1 and then leveled off with further increase of sample loading flow rate. Studies on the influence of sample loading time on the displacement sorption preconcentration show that the integrated absorbance of mercury increased linearly up to a sample loading time of 60 s, after which the slope decreased gradually, presumably as a result of an insufficient capacity of the KR and/or a chromatographic effect, with the sorbed analyte being partially leached by further sample (21). KR Rinsing. The aim of such a rinse step for KR and connecting tubing before elution is to remove nonadsorbed or weakly adsorbed concomitant elements without stripping the sorbed analyte from the KR. This step is of particular significance with samples of high matrix contents. To find out optimal conditions for rinsing the KR, the effects of the wash medium, its acidity, the duration, and flow rate for KR rinsing were examined carefully with respect to the analyte signal and background. Deionized water, diluted nitric acid, or hydrochloride acid was found be unsuitable as the wash medium because they could strip the retained analyte from the KR. However, the use of 0.001% v/v sulfuric acid as the wash medium at a flow rate of 3.0 mL min-1 for 10 s was quite effective for the removal of residual matrix without analyte losses. Analyte Elution. In this work ethanol was employed as the eluent for the reasons described in previous reports (1719). The effect of the eluent volume on the absorbance of 2 µg L-1 Hg was investigated at an elution flow rate of 3.0 mL min-1. It was found that 50 µL of ethanol was enough for quantitative elution of the sorbed analyte from a 100-cm long KR. ETAAS Parameters. Because only 50 µL of ethanolic eluate was introduced into the graphite furnace, no preheating step was required during eluate delivery. In addition, no pyrolysis stage was necessary owing to effective removal of matrices in the present FI online displacement sorption preconcentration system. Unlike conventional ETAAS for determination of mercury, which always needs chemical modifiers for the stabilization of mercury, no additional chemical modifiers were necessary in the proposed system due to the stability of the Hg-PDC at the drying stage. Moreover, the introduction of the mercury as Hg-PDC into the graphite furnace also allowed efficient atomization of the analyte at low temperatures (325-600 °C), prolonging the lifetime of the graphite furnace. Interference Evaluation. The effect of typical heavy metal ions of Cu(II), Cd(II), Fe(III), Ni(II), and Zn(II) on the determination of 2 µg L-1 Hg(II) were investigated. In a preliminary study of this work with a conventional FI online KR sorption preconcentration involving formation and sorption of the Hg-PDC in the KR via direct merging of the Hg(II) and APDC solutions, the tolerance of Cu(II), Cd(II), Fe(III), Ni(II), and Zn(II) was found rather poor, i.e., the maximum tolerable concentrations of Cu(II), Cd(II), Fe(III), Ni(II), and Zn(II) were 0.05, 0.05, 0.10, 0.05, and 0.10 mg L-1, respectively (column 4 in Table 3). The interference of these heavy metal ions likely resulted from their competition with Hg(II) for the complexing agent APDC and/or active sites on the inner walls of KR. However, in the proposed FI online displacement sorption preconcentration system the tolerance of these typical heavy metal ions was greatly improved. As shown in column 3 in Table 3, the tolerable concentrations of Cu(II), Cd(II), Fe(III), Ni(II), and Zn(II) were up to 12, 20, 16, 20, and 60 mg L-1, VOL. 36, NO. 22, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4889

TABLE 3. Effect of Potentially Interfering Species on the Determination of Mercury at the 2 µg L-1 Level recovery (mean ( s, n ) 3)/% concn/ mg L-1

ion Cu(II)

Cd(II)

Fe(III)

Ni(II) Zn(II)

0.05 0.10 12 14 16 0.05 0.10 16 20 25 0.05 0.10 16 20 0.05 0.10 20 0.10 0.50 60 100

present displacement sorption preconcn system

conventional sorption preconcn system

100 ( 1 100 ( 2 99 ( 1 86 ( 1 76 ( 2 100 ( 1 100 ( 1 96 ( 1 93 ( 2 79 ( 1 100 ( 1 100 ( 1 105 ( 2 116 ( 2 100 ( 1 100 ( 2 102 ( 1 100 ( 1 100 ( 2 99 ( 1 88 ( 2

96 ( 1 87 ( 2

94 ( 1 75 ( 2

101 ( 1 118 ( 3 97 ( 1 89 ( 2 100 ( 1 88 ( 2

TABLE 4. Characteristic Data of the FI Online Displacement Sorption Preconcentration for ETAAS Determination of Mercury under the Conditions Given in Tables 1 and 2 working range/µg L-1 preconcentration time/s sampling frequency/samples h-1 enhancement factora sample consumption/mL reagent consumption/mL ethanol 0.5 mg L-1 Cu 0.03% m/m APDC retention efficiencyb/% detection limit (3s)c/ng L-1 precision (RSD, n ) 13)/% calibration function (11 standards, n ) 3, CHg in µg L-1) correlation coefficient

0.05-10.0 40 22 91 2.5 0.050 1.20 1.20 84 6.2 1.1 (2 µg L-1) Aint ) 0.0689CHg +0.0029 0.9986

a

Compared with direct injection of 50 µL of aqueous solution. Compared with the total mass of the analyte introduced to the KR. Calculated on the basis of three times the standard deviation for 11 replicate measurements of a standard solution of 20 ng L-1 Hg.

b c

respectively. These results clearly demonstrated the high selectivity of the proposed FI online displacement sorption preconcentration and separation for ETAAS determination of mercury. Analytical Performance of the FI Online Displacement Sorption Preconcentration for ETAAS. Characteristic data on the performance of the FI online KR displacement sorption preconcentration system for the determination of Hg are given in Table 4. With consumption of 2.5 mL of the sample solution, an enhancement factor of 91 was obtained in comparison with direct injection of 50 µL of aqueous standard solution with the use of a universal palladium modifier. The retention efficiency, relative to the total mass of the analyte introduced to the KR, was 84%. Although the retention of mercury by the KR was not quantitative, excellent precision (1.1% RSD at the 2 µg L-1 level, n ) 13) was still achieved thanks to the reproducible flow conditions. The sampling frequency for liquid samples was 22 samples h-1. The detection limit was calculated on the basis of three times standard deviation for 11 replicate determinations of 20 ng L-1 Hg. The relative detection limit was determined to be 6.2 4890

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 22, 2002

TABLE 5. Analytical Results (Mean ( s, n ) 5) for the Determination of Mercury in Certified Reference Materials (CRMs) concentration of Hg CRMs

unit

GBW 07310 marine sediment GBW 08572 prawn GBW 08508 rice flour GBW(E) 080392 simulated natural water

µg

g-1

certified

determined

0.280 ( 0.04

0.275 ( 0.002

µg 0.201 ( 0.004 0.196 ( 0.002 µg g-1 0.038 ( 0.003 0.032 ( 0.005 mg L-1 0.0100 ( 0.0005 0.0099 ( 0.0002 g-1

TABLE 6. Analytical Results for the Determination of Mercury in Water Samples sample

concentration determined (mean ( s, n ) 5)/µg L-1

recovery of 0.05 µg L-1 Hg spiking (%)

river water 1 river water 2 lake water 1 lake water 2 tap water