Evaluation of Several Field Test Kits for Determining Concentrations of Arsenic in Drinking Water

Transcription

1 Evaluation of Several Field Test Kits for Determining Concentrations of Arsenic in Drinking Water J. Mitchell Spear Environmental Programs Penn State Harrisburg Middletown, PA You ( Mark ) Zhou Environmental Engineer Brinjac Engineering Harrisburg, PA Charles A. Cole Environmental Programs Penn State Harrisburg Middletown, PA Yuefeng F. Xie Environmental Programs Penn State Harrisburg Middletown, PA Abstract The performance of several commercially available field test kits for arsenic was evaluated and compared to standard graphite furnace atomic absorption spectrophotometry (GFAA). Performance was determined by precision, accuracy, linearity, operator bias, matrix effects and ease of use. Precision and accuracy were determined for arsenic III, arsenic V, and a combination of species. Precision was relatively good for all test kits as determined by standard deviation. Accuracy however, calculated through percent recoveries, varied greatly among test kits. Three test kits produced percent recoveries of greater than 75%. However, the other four test kits produced poor percent recoveries as defined by the authors (50 66%). Matrix effects were evaluated using additions of known interferences of sulfide and antimony in reagent water. Field samples were also tested at various concentrations to determine performance throughout the working range of the field test kits, defined as linearity. All field kits were relatively free from operator bias. Four test kits were determined to be relatively easy to use, or have a high ease of use score. Two of the field test kits met all the following criteria; easy to use, no operator bias, inexpensive, accurate and precise. Introduction The U.S. Environmental Protection Agency published the final arsenic rule on January 22, The rule specifies an effective compliance date of January 23, This rule 1

2 lowers the arsenic Maximum Contaminant Level (MCL) from 50 micrograms per liter (µg/l) to 10 µg/l. It is estimated that nearly 4000 water utilities in the United States are affected, with nearly 97% of them serving fewer than 10,000 people (1). To comply with the new arsenic MCL (10 µg/l), water systems will need to remove arsenic using either Best Available Technologies (BAT) or Small System Compliance Technologies (SSCT), for systems less than 10,000. Under the SSCT small water systems may choose point-ofuse (POU) devices (e.g. activated alumina, ion exchange, iron-based sorption media or reverse osmosis). POU treatment is often the most cost effective system for arsenic removal, especially for very small systems serving less than 200 connections (1). Although, many of these POU devices have low capital cost and require little maintenance, monitoring their treatment effectiveness is a major obstacle. Testing effluent arsenic concentrations of every customer s POU device with US EPA approved analytical methods will be very expensive. Each commercial analysis costs between $15 and $50 (2). These laboratory techniques include Graphite Furnace Atomic Absorption (GFAA), Inductively Coupled Plasma Mass Spectroscopy (ICP-MS), Inductively Coupled Plasma Emissions Spectroscopy (ICP-AES) and Hydride Generation Atomic Adsorption (HGAA) (3). If a system decides to purchase the analytical equipment rather than pay for external analysis, they will find that these instruments are expensive (between $30,000 and $150,000) and require a highly trained technician (4). In addition to the greater cost of using a contracted laboratory, the time interval between sending the sample for analysis and receiving the results makes it difficult for the water utility to test different arsenic removal strategies or determine the removal efficiencies of POU devices in a time efficient manner. Owing to this need for affordable, quick, easy and reliable testing of arsenic, several field test kits for arsenic have recently become commercially available. Many of these field kits are simple to use and therefore do not require a highly skilled technician. Several are relatively inexpensive (less than $5 per analysis); however most kits provide discrete (non-continuous) concentration values. Because they are relatively inexpensive, easy to use, and portable for field use, test kits seem to be a probable alternative to expensive commercial analyses for routine operational (noncompliance) testing. Recently however, the reliability of these field test kits has been questioned. Many previously evaluated field kits produced weak correlations with approved analytical methods (5). Our purpose was to evaluate several commercially available field test kits and determine their reliability and applicability to water utilities currently using POU devices, testing possible treatment strategies, or simply having interest in conducting noncompliance arsenic analysis on their raw and treated waters. Methods (Experimental) Test kits Seven different field test kits were evaluated. Two to four kits of different lot numbers were chosen for each of the six field test kits evaluated. The exception was the field test kit that uses anode stripping voltammetry (ASV), which we used the same electrodes 2

3 throughout the duration of the study. Six of the test kits use similar chemistries of hydride generation as described in Standard Method 3114 (6). These test kits quantified arsenic concentrations in the sample by generating arsine gas (AsH 3 ) and correlated the amount of arsine gas generated to the concentration of the arsenic in the sample. These test kits provided semi-qualitative incremental concentration data. One field kit used anodic stripping voltammetry in conjunction with microelectrodes. ASV has been tested by several researchers (7, 8). The arsenic is reduced and collected on the working electrode and then stripped off (oxidized) and measured similar to Standard Method ASV in conjunction with microelectrodes provides continuous data, although it requires a higher degree of analytical training. All analyses were conducted following test procedures provided with the test kits (i.e. reagent amounts, time requirements and temperature). However, no sample dilutions were performed on the waters tested. Results obtained from all test kits were compared with those obtained using EPA Method 7060A (GFAA). Method detection limit (MDL) for EPA Method 7060A (GFAA) and the field kit using ASV was calculated using the following equation: MDL = t (n-1, 1 á = 0.99) x SD (Equation 1) where t is equal to the Student s t statistic for n-1 degrees of freedom at 99% confidence level and SD is equal to the standard deviation. Laboratory performance Laboratory tests were conducted to determine the accuracy and precision of the test kits compared to known standard arsenic solutions in reagent water. A concentration of approximately 30 µg/l of arsenic was added to 18 MÙ de-ionized reagent water (Nanopure ). The concentration of 30 µg/l was chosen based on the following criteria; the value was near one of the incremental values provided by each of the kits, and was greater than the 2006 enforceable MCL of 10 µg/l, but less than the current MCL of 50 µg/l. This should assist water utilities that will be the most impacted by the upcoming regulation. Additions of arsenic III, arsenic V and equal proportions of arsenic III and V were evaluated for all test kits. Five to seven replicate tests were used to calculate accuracy (percent recoveries) and precision (standard deviation). Matrix interference and field performance Matrix interference testing was determined by additions of known interference agents (antimony and sulfide). Antimony was evaluated at three concentration levels, 0.25, 1.0 and 5.0 mg/l, spiked into the reagent water as elemental antimony containing approximately 30 µg/l of arsenic. Effects of 0.5, 5.0 and 10 mg/l of sulfide on arsenic analysis were also evaluated. In addition, controls (no antimony or sulfide additions) were used for comparison. All levels were tested in at least duplicate. Performance was determined by measuring percent recoveries. In addition, field performance testing 3

4 consisted of spiking various concentrations of arsenic III and V (1:1) into a local ground water sample. Water quality characteristics are listed in Table 1. Linear regression was used to compare test kit results to those obtained by GFAA. Table 1: Ground water characteristics Constituent or Contaminant Concentration ph 8.4 Alkalinity 230 mg/l as CaCO3 Calcium hardness 74 mg/l as CaCO3 Total Dissolved Solids 448 mg/l Fluoride 0.11 mg/l Barium mg/l Nitrate mg/l Arsenic Non Detect All other inorganic chemicals tested* Non Detect * cyanide, mercury, antimony, beryllium, cadmium, chromium, nickel, selenium, thallium. Operator performance and ease of use Operator bias was determined using four operators and comparing results of one operator against the results of the other operators, using similar protocol described by Schock and George (9). Each operator was given a large sample of known concentration. Smaller aliquots were taken from these samples and analyzed by each operator. The first operator, designated Operator 1, was experienced in the use of the test kits having performed many analyses using these test kits prior to the comparison. Operators 2, 3 and 4 had not previously used the test kits prior to the comparison. No attempt was made to determine operator bias for the test kits using ASV. Ease of Use was calculated from five operators by having each operator score the test kit on a scale of The operators determined the score of the test kits using the following criteria; how easy it was to follow the provided instructions, the method of chemical additions, operation of the testing equipment (i.e. sample bottles and caps), and the ease of interpreting the results (comparing test sample to the color chart). A score of 10 indicates the test kit was extremely easy to use, and a score of 1 indicates difficulty and the need for additional training. Results and Discussion General characteristics Concentration intervals provided by each test kit ranged from 5 intervals to continuous data (Table 2). Sample sizes ranged from 10 to 250 ml and the number of chemical reagents required ranged from 2 to 5 additions among field test kits (Table 2). Analysis time varied from 13 to 40 minutes (Table 2). Number of samples each test kit could perform was 50 or 100 for five of the six test kits evaluated. The exception, the test kit using ASV, was dependent on the life of the electrodes which will determine the total 4

5 number of samples it can analyze. Prices per analysis ranged from $0.30 to $4.40 per analysis (Table 2). Table 2: Major characteristics of the field test kits evaluated. Test Kit Concentration intervals Number of reagents Test time (min) BVC-100 ECO-W100 Hach 10, 25, 50, 100, , 50, 100, 150, 200, 300, 500, 750 Sample size required (ml) , 30, 50, 70, 300, LaMotte 4, 6, 8, 10, 12, 14, 16, 18, 20, 30, 40, 50, 60, 70, 80, 100, 140, 160 Merck 10, 25, 50, 100, 500 Quick II 1, 2, 2.5, 3, 4, 5, 6, 7, 10, 12, 14, 20, 30, >30, >40, >60, >80, >100 Trace Detect Continuous <10 b 50 Unit price $ / kit a Sample per kit (Price $ / test) (0.30) (0.36) (1.06) (3.06) (0.70) (4.40) a Prices based at time of purchase b It takes about 45 minutes to build calibration curve c Based on the assumption you can analyze 5000 samples with proper electrode usage and maintenance Laboratory performance The relative standard deviation (RSD) was less than 3 percent for all measurements obtained by the GFAA and the standard deviation was calculated at The calculated MDL was equal to 1.9 µg/l. Because six of the testing kits evaluated provide discrete (non continuous) concentration values, standard deviations and MDLs could not be determined in the traditions sense. Percent differences obtained between the test kits and EPA method 7060A were chosen to indicate the accuracy of the test kits (Figure 1). Standard deviation calculated for each test kit was chosen to signify the overall precision. The test kit using ASV provided continuous data. We were therefore able to calculate the MDL. RSD was less than five percent for all measurements, and standard deviation was 12, c (2.50) 5

6 equal to 0.65, corresponding to a MDL value of 2.1 µg/l. This value was similar to our calculated MDL using EPA Method 7060A. Percent recoveries on test kits ranged from 33.3 to percent for arsenic III spiked into reagent water at a level of 31.4 µg/l (Figure 1). Percent recoveries ranged for 33.3 to 97.3 percent for arsenic V spiked at a concentration level of 30.7 µg/l (Figure 1). Similar recoveries were achieved using a combination of arsenic III and V (1:1), 33.3 to 100 percent (Figure 1). No determination was made for the arsenic V for the BVC field kit because the method is for determination of arsenic III. Determination for the combination of arsenic III and V was approximately half that of the recovery for the arsenic III addition alone (Figure 1). This seems to verify that the BVC kit is only measuring the arsenic III, which was half the amount in the arsenic III + V combination addition as in the arsenic III addition alone. All of the test kits had a standard deviation of less than 10 µg/l, with many having a SD of zero, indicating a high level of repeatability - precision. This also suggests the five to seven replicates chosen in this study were sufficient to determine the percent recovery values. It has been suggested that to obtain the most accurate results, the samples be determined in triplicate whenever possible (9). When using few replicates, the median value which can eliminate either extremely high or low values is often the better predictor of the actual concentration when compared to the arithmetic mean (10). However, mean values were used in our calculations, because no extremely low or high values were obtained as indicated by the low standard deviations. 120 Percent Recovery As III As V As III + V 0 BVC ECO Hach LaMotte Merck Quick II (old) Quick II (new) Trace Detect Figure 1: Percent recoveries of arsenic field test kits in reagent water. Standard deviations indicated by error bars. Note the large difference in recoveries between the Quick II field kits designated as old versus new. During the study, the manufacturer recalled the test strips of mercuric bromide. The old packaging had an inadequate polyethylene (PE) lining which permitted 6

7 a reaction between the mercuric bromide and the aluminum foil packaging, causing poor recoveries. The new strips dramatically increased the performance of their product (Figure 1). Percent recoveries increased from 64 to 94, 7 to 96 and from 34 to 100 percent in the arsenic III, V, and III + V addition respectively (Figure 1). Matrix interference and field performance Results of the field testing are presented in Figure 2. Percent recoveries and standard deviation were similar to that using reagent water. Recoveries ranged form 38.8 to percent (Figure 2) as compared to 33.3 to percent (Figure 1) for the reagent water. Standard deviation ranged from 0 to 7.2 for seven replicates on each test kit. A similar improvement was observed in the field testing compared to the laboratory testing for the Quick II field kit designated as old versus new. Recovery increased from 41 to 105 percent for this test (Figure 2). 120 Percent Recovery BVC ECO Hach LaMotte Merck Quick II (old) Quick II (new) Trace Detect Figure 2: Percent arsenic recovery of arsenic field test kits using a field water source. Standard deviations indicated by error bars. Matrix interference results are shown in Figures 3 and 4. Antimony would rarely be found in the levels tested, with the majority of drinking waters found in the U.S. below 1.5 µg/l (11). However, we evaluated levels up to 5.0 mg/l. Recoveries did not seem to be additionally compromised in the presence of antimony in the concentrations range of 0.0 to 5.0 mg/l (Figure 3). The percent recoveries were less than +/- 5 percent different in the control sample (0.0 mg/l of antimony) as compared to the highest concentration addition (5.0 mg/l) for all but the LaMotte and Quick II test kits (Figure 3). The Quick II and LaMotte test kits results indicate a positive interference (approximately 16 to 20% higher recovery) in the level 1.0 mg/l and 1.0 and 5.0 mg/l antimony additions, respectively. Sulfides did not seem to influence the percent 7

8 recoveries for six of the seven test kits in the range of 0.0 to 10.0 mg/l (Figure 4). Increase in sulfide concentrations however, did seem to reduce the percent recovery in the Quick II test kit from 98 percent at 0.0 mg/l sulfide addition to 66 percent in the 10.0 mg/l sulfide addition. Concentrations of hydrogen sulfides at the levels tested however would not be common for drinking waters (4) Percent Recovery mg/l Sb 0.25 mg/l Sb 1 mg/l Sb 5 mg/l Sb 0.0 BVC ECO Hach LaMotte Merck Quick II (new) Trace Detect Figure 3: Percent arsenic recovery with several concentration levels of interference agent antimony Percent Recovery mg/l Sulfide 0.5 mg/l Sulfide 5 mg/l Sulfide 10 mg/l Sulfide 0.0 BVC ECO Hach LaMotte Merck Quick II (new) Trace Detect Figure 4: Percent arsenic recovery with several concentration levels of interference agent sulfide. 8

9 In addition to the interference study and the 30 µg/l spiked field samples, test kits were evaluated at several additional concentrations. Arsenic was spiked into finished water at 5, 10, 25, 50 and 75 µg/l to determine the performance of the test kits through the working range of their calibration (linearity). The values obtained by the test kits were plotted against the actual values, as determined by GFAA (Figure 5-8). Test kit response curves where plotted with the following equation: y = mx + b (Equation 2) where m = slope or test kit response, and b = y-intercept. Slopes close to one indicate no bias, slopes much less than one would indicate the potential of a negative bias (sample results lower than the actual concentration) in the higher range of the test kit and a positive bias at the lower range. Slopes greater than one likewise indicate the potential of a negative bias in the lower concentration levels of the test kit and a positive bias in the higher concentrations. A slope equal to or near 1.00 and a negative y- intercept indicates an equal negative bias throughout the range of the test kit. Therefore, a positive intercept and slope near 1 indicates a positive bias throughout the range of the test kit. Only three test kits have both a slope near one and the intercept close to zero (Figure 5-8). The Trace Detect, LaMotte and Quick II kits have calculated slopes near 1.0 (0.900, and 0.770, respectively) and an intercept less than (+/-) 2.0. Results of previous studies have shown poor correlations of arsenic test kits to flow injection hydride generation atomic absorption spectrophotometry (5). Measured arsenic in test kits (ug/l) y = 0.66x y = 0.28x BVC ECO Linear (No Bias) Linear (ECO) Linear (BVC) GFAA (ug/l) Figure 5: Comparison of the BVC-100 and ECO test kits analytical results with GFAA results. 9

10 Measured arsenic in test kits (ug/l) y = 0.85x y = 0.73x Hach LaMotte Linear (No Bias) Linear (LaMotte) Linear (Hach) GFAA (ug/l) Figure 6: Comparison of the Hach and LaMotte test kits analytical results with GFAA results. Measured arsenic in test kits (ug/l) y = 0.77x y = 0.56x Merck Quick II Linear (No Bias) Linear (Quick II) Linear (Merck) GFAA (ug/l) Figure 7: Comparison of the Merck and Quick II test kits analytical results with GFAA results. 10

11 Measured value in test kit (ug/l) y = 0.90x TraceDetect Linear (No Bias) Linear (TraceDetect) GFAA (ug/l) Figure 8: Comparison of the Trace Detect test kit analytical results with GFAA results. Operator performance and ease of use All test kits show strong correlations among operators (Table 3). Correlation coefficients ranged from 0.69 to Slopes among operators ranged from 1.0 to 0.47 (Table 3). Five of the six field kits had slopes and correlation coefficients greater than Only the Merck kit with a value of 0.47 has a slope that deviated greatly from 1.0. This however was likely caused by the relatively low recoveries found throughout the range of samples tested. Sample sizes ranged form 11 to 21 samples. Table 3: Operator bias Test kit Operator 1 Slope (intercept) BVC-100 Operator 4 ECO-W100 Operator 4 Hach Operator 2 LaMotte Operator 3 Merck Operator 3 Quick II Operator 2 * Significant to the 0.01 alpha level (0.0) 0.99 (-1.6) 0.84 (2.44) 1.0 (0.77) 0.47 (4.9) 0.90 (1.0) Correlation coefficient 0.897* 0.905* 0.829* 0.890* 0.689* 0.873* 11

12 Ease of Use scores (1-10 scale, 10 being the easiest) ranged from 2.3 to 8.4 and standard deviation was less than 1.0 for all kits (Figure 9). The Hach, LaMotte, Merck, and Quick II all scored significantly (alpha < 5%) higher than the other three kits (Figure 9), indicating they were easier to perform relative to the other field test kits evaluated. Significant differences indicated by different letters. Test kits with a letter b are significantly different from test kits with either a letter a or c and visa versa. "ease of use" b b a a a a c 0 BVC-100 ECO-W100 Hach LaMotte Merck Quick II Trace Detect Figure 9: Field test kit ease of use Conclusions The results from this study suggest that three of the field test kits produce reliable results and could be used for noncompliance sampling. The water utility should first evaluate the expected range of arsenic concentration of the system before determining which test kit is best suited to their needs. The authors suggest that the LaMotte and the Quick II will be most applicable for water utilities having a raw water arsenic concentration of less than 100 µg/l. The Trace Detect kit performed very well, but because of the high initial cost is most applicable in medium to large water utilities that have a) the capital to purchase the instrument and b) the greater number of samples to analyze, causing a reduced cost per sample scenario. Additionally, this field kit is also appropriate for contracted vendors of POU devises that are responsible for monitoring the influent / effluent concentration of the devices they install. The authors suggest for the purpose of noncompliance testing of arsenic in water systems, that the kits must be able to test below the 10 µg/l level, and have several concentration intervals (at least 8) between the 20 and 0 µg/l. This is particularly useful in testing the removal efficiencies of different POU devices on raw water. It may also be useful for 12

13 developing correlations between arsenic concentrations in the treated water and reliability of a surrogate analysis such as conductivity in POU-RO devices. Although our results indicated low standard deviations, there is merit in not relying on single sample determinations, and therefore the authors agree with Shock and George (9) that samples should be analyzed in triplicate. This however will undoubtedly increase the costs for the analyses and may not be practical for all tests. One item not addressed is the waste generated and the health concerns associated with conducting these analyses. Most of these field test kits generate wastes of the following chemicals; Hydrochloric acid, zinc, potassium iodide, stannous chloride and mercuric bromide test strips. Of these wastes, the mercuric bromide test strips are the most difficult to address since they cannot be disposed of in general waste. Of greater concern for those test kits that use chemistries similar to Standard Method 3114 is the arsine gas which is generated during the analysis. This arsine gas may be dangerous to the analyst if a large number of samples are performed in a poorly ventilated area (5, 12). Despite these concerns, field test kits seem to be a likely choice for many water utilities conducting noncompliance sampling. If the correct field kit is chosen, and properly used, it can offer the water utility and operator information on arsenic concentration in a low cost approach. The results obtained under the study will help water systems make an educated decision in choosing the appropriate field test kit for conducting routine (noncompliance) analysis of arsenic. This will aid them in complying with the promulgated arsenic rule. The results may also be used by the test kit manufacturers to improve their products. One manufacturer changed their packaging of the test strips during the study, which significantly improved their results. Acknowledgments The authors would like to acknowledge the US EPA Small Public Water Systems Technology Assistance Center Grant which funded this project. The authors would also like to thank TraceDetect for loan of the Nano-Band Explorer. Mention of trade names or commercial products does not constitute endorsement. References (1) Kommineni, S., Narasimhan, R. and Durbin, H Point-of-use/point-ofentry treatment for arsenic removal: operational issues and costs. Proceedings, Water Quality Technology Conference. (2) Ray, H Arsenic in water: a brief overview. Water Conditioning and Purification. March

14 (3) USEPA, National Primary Dinking Water Regulations; Arsenic and Clarifications to Compliance and New Source Contaminants Monitoring. Federal Register, 66:14:6975. (4) Anderson, R. D., McNeill, L. S., Morton, S. C. and Edwards, M., Field Measurement Methods for Arsenic in Drinking Water. AWWA Water Quality Technology Conference. (5) Rahman, M.M., Mukherjee, D., Sengupta, M.K., Chowdhury, U.K., Lodh, D., Chanda, C. R., Roy, S., Selim, Quamruzzaman, Q., Milton, A. H., Shahidullah, S. M., Rahman, T., Chakraborti, D Effectiveness and reliability of field test kits: Are the million dollar screening projects effective or not? Environmental Science and Technology. 36(24), (6) APHA, AWWA and WEF Standard methods for the examination of water and wastewater. 20 th Edition. Washington D.C.: American Public Health Association. (7) Huang, H. L. and Dasgupta, P. K A field-deployable instrument for the measurement and speciation of arsenic in potable water. Analytica Chimica Acta, 380:1:12. (8) Feeney, R and Kounaves, S. P On-site analysis of arsenic in groundwater using microfabricated gold ultramicroelectrode array. Analytical Chemistry, 72:10:2222. (9) Schock, M. R. and George, G. K Evaluation of a field test kit for monitoring lead in drinking water. Journal of AWWA, 85(8), 90:100. (10) Dean, R. B. and Dixon, W. J Simplified statistics for small numbers of observations. Analytical Chemistry 23:4:636. (11) Frey, M. and Edwards, M Surveying arsenic occurrence in U.S. drinking water. Journal of AWWA, 89:3:105. (12) Hussam, A., Alauddin, M., Khaun, A. H., Rasul, S. B., and Munir, A. K. M Evaluation of arsine generation in arsenic field kit. Environmental Science and Technology. 33,