Envlron. Scl. Technol. 1004, 18, 535-539
Discussion and Conclusions For all classes of data evaluated, trends were most effectively detected in uncensored data as compared to censored data even when the data censored were highly unreliable. Censoring data at any concentration level, especially given our lack of knowledge about prevailing concentrations, may eliminate valuable information. The more reliable the data censored, the greater the information lost and the more detrimental the effects of censoring. Whether or not valuable information for trend analysis is in fact eliminated by censoring of actual rather than simulated data depends mainly on whether bias is predictable for a particular type of chemical analyses and the measurement process is in statistical control. To be in statistical control, the location and shape of the measurement error distribution must be reasonably constant a t each concentration. If there are changes in bias over time that are unknown and cannot be compensated for, then valid trend tests are not possible. The use of simulated data in this study assured statistical control and a constant and predictable relation between bias and concentration. How well and how commonly this condition would be met for actual trace-level measurements is presently difficult to assess except on a case-by-casebasis. In general, the condition probably holds true for measurements made by one laboratory by a constant measurement process. If the measurement process changes over time, or if measurements from more than one laboratory are to be used, then the required condition of a constant relation between bias and concentration may well not be met. In either of these cases, only careful evaluation of and correction for differences in bias would allow valid use of the data for trend analysis. Note that the need for such evaluations applies equally to data that
are above present data reporting limits-these reporting limits are not magic boundaries for the absence or presence of statistical control. The results of the present study lend strong support to the argument, earlier advanced by Rhodes (1) and generally followed in new ASTM standards (7), that data should not be routinely censored by laboratories. Uncensored data should always be retained in permanent records available to data users even if policy makers of a laboratory decide that some censoring or other form of qualification is necessary before general public release of data. Measurement data should not be discarded unless the lack of statistical control in the measurement process is clearly demonstrated. Registry No. Water, 7732-18-5. Literature Cited (1) Rhodes, R. C. In "Environmetrics 81: Selected Papers"; SIAM Institute for Mathematics and Society: Philadelphia, 1981; pp 157-162. (2) Mann, H. B. Econornetrica 1945, 13, 245-249. (3) Kendall, M. G. "Rank Correlation Methods"; Charles Griffin: London, 1975. (4) Hirsch, R. M.; Slack, J. R.; Smith, R. A. Water Resourc. Res. 1982, 18, 107-121. (5) Aitchison, J.; Brown, J. A. C. "The Log Normal Distribution"; University Press: Cambridge, 1957. (6) Eisenhart, C. Science (Washington, D.C.)1968, 160, 1201-1204. (7) ASTM, Subcommittee 019.02. Annu. Book ASTM Stand. 1983, 11.01, Chapter D, 4210-4283. (8) Gilliom, R. J. U.S. Geological Survey, Reston, VA, unpublished data, 1983.
Received for review July 27, 1983. Accepted January 27,1984.
Determination of Subnanogram per Liter Levels of Earthy-Musty Odorants in Water by the Salted Closed-Loop Stripping Method Cordella J. Hwang," Stuart W. Krasner, Michael J. McGulre, Margaret S. Moylan, and Melissa S. Dale The Metropolitan Water District of Southern California, Water Quality Branch, La Verne, California 9 1750
Grob closed-loop stripping analysis (CLSA), coupled with gas chromatography/mass spectrometry (GC/MS), has been used to analyze for low levels of earthy-musty odorants in potable water supply systems. An increase in both sensitivity and stripping rate has been achieved by raising the ionic strength of the water samples with sodium sulfate (NazS04)before stripping. Concentrations from 0 to 300 g/L NaZSO4were studied. A practical condition of 80 g/L NaZSO4was combined with a shorter 1.5-h stripping time to double the sensitivity of the unsalted 2-h CLSA. A detection limit of 0.8 ng/L was achieved for each of the following earthy-musty compounds: 2-isopropyl3-methoxypyrazine, 2-isobutyl-3-methoxypyrazine,2methylisoboneol, geosmin, and 2,3,6-trichloroanisole. The method was validated by comparison, precision, and accuracy studies using natural source-water samples. Recoveries of dosed geosmin and 2-methylisoborneolwere 105 f 15% and 106 f 15%, respectively. This salted procedure is a valuable modification of the CLSA when higher sensitivity or reduced stripping time is required. Introduction Water utilities have long been plagued by recurring earthy-musty taste and odor problems. Two such odor0013-936X/84/0918-0535$01.50/0
ants, geosmin and 2-methylisoborneol (MIB), have been implicated in incidents throughout the world (1-3). They are secondary metabolites of certain blue-green algae and actinomycetes (4-7). These odorants, which have been found periodically in source-water reservoirs of The Metropolitan Water District of Southern California (Metropolitan) during the past 4 years, were the primary target of this study. Detection of concentrations below the odor threshold can act as an early warning procedure, permitting utilities to take action to avoid or mitigate an odor problem. Three other earthy-musty compounds, 2-isopropyl-3-methoxypyrazine(IPMP), 2-isobutyl-3methoxypyrazine (IBMP), and 2,3,6-trichloroanisole (TCA), were also studied. These five earthy-musty odorants have odor thresholds at low nanogram per liter levels (8-12) which are well below the detection limits of classical analytical techniques. Two methods developed for concentrating trace organics in water are the purge-and-trap technique (13) and closedloop stripping analysis (CLSA) ( 1 4 , 1 5 ) . The purge-andtrap technique has been used to detect relatively insoluble, volatile organic compounds at low microgram per liter levels. In 1980, the Master Analytical Scheme (16)presented a "salted-out" modification of the purge-and-trap
0 1984 American Chemical Society
Environ. Scl. Technol., Vol. 18, No. 7, 1984
535
I
Flgure 1. Closed-loop stripping apparatus.
method, using 60 g of anhydrous Na2S04/200mL of sample, to lower detection limits to 0.1 rg/L levels. Trussell and co-workers (17) developed a purge-and-trap gas cluomatography/maas fragmentography(GC/MF) method using 10 g of Na2S0,/25 mL of sample for determining the levels of geosmin and MIB. Independently, Yagi and coworkers (18) reported the use of a purge-and-trap GC/MF technique with the addition of 10 g of sodium chloride/100 mL of sample. Concentration of the odorants by a purge-and-trap method is less sensitive than CLSA, thus requiring MF (also called selected ion monitoring) to achieve low nanogram per liter detection levels. This limitation precludes full-scan mass spectral identification of the odorants and other compounds present in a sample. In 1980, Krasner and co-workers (19) adapted the Grob CLSA with GC/MS (full scan) for the determination of low nonogram per liter levels of geosmin, MIB, IPMP, IBMP, and TCA in water, sediments, and cultures. The purpose of this study was to investigate the impact of ionic strength adjustment on the CLSA of earthy-musty odorants in water and to define and evaluate practical modifcations to improve recoveries, lower detection limits, and shorten analysis time. Experimental Section Apparatus. Analyses were performed on a modified Grob CLSA apparatus (see Figure 1)in which semivolatile organic compounds were air stripped from a water sample and trapped on a 1.5-mg activated carbon (AC) filter (20). The system was used with the 1-L, "tall-form" stripping bottle (76 mm 0.d. X 318 mm height below ground-glass joint), which was essential to obtain the efficient stripping observed in this work. The carbon disultide filter extracts were analyzed on a Finnigan Model 4023 GC/MS. It is critical that the rate of gas (air) flow in the stripping system be consistent from run to run. Increased resistance due to a clogged filter or frit can alter recoveries significantly. Stripping-bottle frits were cleaned every 2 weeks 536
Envfron. Sci. Technol.. Vol. 18. No. 7. 1984
(about 40 runs) with chromesulfuric cleaning solutions, rinsed thoroughly with demineralized water, organic-free water, acetone, and methanol, and baked a t 180 'C for 4 h. The pump and filter-holder assembly were cleaned whenever sample carry-over was observed. The system, with the quick-connects open and the pump on, was flushed with approximately 100 mL each of organic-free water, acetone, and methanol. After the last rinse, a heat gun was used to dry the system until there was no trace of residual methanol. To clean the AC filters, the longer glass tube above the c h c o a l was fded with reagentgrade solvents, which were allowed to gravity-flow through the filter disk. The solvents were added as follows: (1)acetone, (2) organic-free water, (3) acetone, (4) carbon disulfide, (5) acetone 3 times, and (6) methylene chloride 3 times. The filter was then vacuum dried for approximately 5 min to remove residual solvent. The water rinse removed any deposited salts. The cleaning procedure, beginning with step 4, was repeated just prior to filter use. The time necessary to empty the last methylene chloride rinse (0.3-mL volume) from the top of the filter tube to the surface of the carbon was measured. This filter free-flow rate was checked before each use. Filters with flow rates lower than 0.6 mL/min were not used. Reagents. Organic-free water was made by irradiating Millipore Super-Q water with ultraviolet light for 1h in a Sybron/Barnstead ORGANICpure unit. Baker reagent-grade, granular, anhydrous sodium sulfate was baked a t 625 OC for 2 h before use. Each sodium sulfate lot was checked with a blank CLSA and pH measurement (acceptable range of water and salt pH = 6-8). For sources of other reagents, see ref 20. Procedure. Samples were collected, without headspace, in "1-L" glass bottles, with Teflon-lined caps, and stored a t 5 "C. Each sample was brought to room temperature just prior to analysis by immersing the sample bottle in a 25 "C water bath for approximately 15 min. No pH adjustment was made, since preliminary data showed no significant differences between stpndards stripped at pH 6 and pH 8, the typical range of water samples containing 80 g/L Na2SOI. An 800-mL fraction of the sample was transferred to the stripping bottle and stirred on a magnetic stirrer while baked sodium sulfate was quickly added. The bottle was covered, and mixing continued until all salt dissolved (less than 1 min). After the stirring bar was removed, the remaining 200 mL of sample was added, rinsing and wetting the inside neck of the bottle. Then the fritted stopper was loosely inserted. The stopper was raised momentarily while the internal-standard spiking solution was injected below the liquid level to give a concentration of 100 ng/L each of 1-chlorooctane (Cl-Cs), 1-chlorodecane (Cl-Clo), and 1-chlorododecane (C1-CI2). The bottle was tightly stoppered, placed in the 25 "C CLSA water bath, and connected to the recirculating system. The air in the headspace was flushed through an auxiliary AC filter for 10 s to remove contaminants. The sample-collection 1.5-mg AC filter w&q then placed in the filter holder, and the sample was stripped for the set time. The filter was removed and extracted with two 10-rL aliquots and one 5-rL aliquot of carbon disulfide. The extracts were analyzed by electron-impact GCjMS according to the conditions given in Table I. The closed-loop stripping procedure, including filter extraction and GC/ MS analysis, is described in detail by Krasner et al. (20). Stripping times were varied from 1 to 4 h. Samples were checked, especially at the start of stripping, for excessive foaming. If necessary, 900-mL samples were stripped to
Table I. GWMS Parameters
sample size injector GC column column temperature program carrier gas interface oven temperature electron energy ionizer temperature source pressure mass scanning program
1.6 y L 200 "C; split closed 1.5 min DB-5 fused silica capillary column; 30 m long X 0.25 mm i.d. ambient 1 min; 57-130 OC at 4 OC/min; 130-255 "C at 7 OC/min helium a t 29 cm/s linear velocity 200 "C 70 eV 270 OC -5 x 10-7 torr 41-75 and 77-400 amu" in 0.6 s
"Base peak of carbon disulfide, 76 amu, omitted. 100 I 80C
I
lPMP
100
1
I
LIBMP
I
IPMP
40
20
0
I I ( I I I I I I / I (
1001
1
- 6ok I
-
0 100,
:;I
.
40
It
-
0
0
20
*Ot-
IL
20
40
60
80
100 0
T-
-
20
40
60
'
80
100
Na2 SO4 CONCENTRATION (g/L) Flgure 2. Recoveries of earthy-musty odorants and internal standards with respect to sodlum sulfate concentration from 4-h stripping of standards.
prevent foam or liquid carry-over into the system.
Results and Discussion Development of the Method. Because sodium sulfate has been successfully used as the ionic strength adjustor for the purge-and-trap technique (16, 17) and is readily available and purifiable in most laboratories, ita effects on CLSA, in conconcentrationsfrom zero to almost saturation, were investigated. Large sodium sulfate additions of 300 and 100 g/L were unmanageable and marginally acceptable, respectively, because of poor rates of dissolution and an excessive tendency to contaminate and clog the stripping system. Therefore, detailed studies of the effects of ionic strength on recoveries and stripping efficiencies were confined to the range 0-100 g/L Na2S04. Standards (20 ng/L each of the five odorants and 100 ng/L each of the three internal standards) were analyzed, with a 4-h stripping time to maximize recoveries and minimize the effect of incomplete stripping. This series
00
02 04 06
08 10
12
0
02 04 06
08 10 12
AC FILTER FLOW (rnL MeCI2/min)
Figure 3. Effect of filter resistance, measured as flow, upon recovery of earthy-musty odorants and CI-C,, internal standard.
of results was then compared with direct GC/MS injections of standards of the spiking solutions in acetone, diluted with carbon disulfide to the same volume as the extracts. These standards gave values comparable to analogous standards of carbon disulfide only. Percent absolute recoveries of the five odorants and internal standards are shown in Figure 2. For IPMP, IBMP, MIB, and geosmin, approximately 30% recoveries were observed for the CLSA of unsalted samples. With the addition of 20 g/L Na2S04,the recoveries increased to approximately 50%. With higher salt concentrations, recoveries approached plateaus of 60-70%. TCA, the most easily stripped of the five compounds (21), had a 50% recovery in the unsalted CLSA, which increased to 65% with the use of salt. The internal standards were relatively unaffected by the salting out; absolute recoveries were around 80%. As shown in Figure 2, all five odorous compounds plateaued at the same percentage recovery range, regardless of solubilities, suggesting that other parameters may be limiting. One factor which influenced absolute recoveries was AC filter resistance. Figure 3 shows the variation in Environ. Scl. Technol., Vol. 18, No. 7, 1984 537
Table 11. Relative Percent Recoverya as a Function of Incremental Stripping Time and Sodium Sulfate Concentration
compound
incremental stripping time, h
2-isopropyl-3-methoxypyrazine
relative recovery, % 60 gfL
8OgfL
l0OgfL
Na2S04
Na2S04
1st
24
34
53 28 10 9
Na2S04 86
2nd 3rd
56 24 9 11
Na2S04 70 20 7 3
10 4