Environ. Sci. Technol. 2005, 39, 8411-8419
Evaluation of Novel Techniques for Measurement of Air-Water Exchange of Persistent Bioaccumulative Toxicants in Lake Superior J U D I T H A . P E R L I N G E R , * ,† DAVID E. TOBIAS,† PATRICK S. MORROW,† AND PAUL V. DOSKEY‡ Department of Civil & Environmental Engineering, Michigan Technological University, 1400 Townsend Drive, Houghton, Michigan 49931, and Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439
We report initial measurements of concentrations and net air-water exchange fluxes of target persistent bioaccumulative toxicants (PBTs) in Lake Superior utilizing techniques not previously applied for this purpose. Gaseous PBTs are collected in diffusion denuders containing sections of commercial chromatography columns and subsequently thermally extracted into the cooled injection inlet of a high-resolution gas chromatograph. The PBT sampling/analytical methods enable accurate determination of gas-phase PBT concentration and micrometeorological measurement of fluxes to be carried out. PBT fluxes are measured by the modified Bowen ratio technique in which sensible heat flux is related to PBT flux, with the assumption of identical transfer velocities of heat and PBTs between two heights in the atmospheric surface layer. Micrometeorological measurement of flux accounts for all sources of resistance to mass transfer, including atmospheric stability effects, surface films, waves, sea spray, and bubbles. The sensible heat flux, PBT concentration, and PBT flux measurements carried out in 14 2- or 3-h periods during seven sampling events in Lake Superior in summer and fall 2002 and spring 2003 demonstrate advantages under the constraints of the techniques. The uncertainty of the flux measurements was typically in the range from 1% to 160%. Gaseous concentrations of R-hexachlorocyclohexane (R-HCH) and hexachlorobenzene (HCB) over Lake Superior were in the range from 6 to 170 and 12-95 pg/m3, respectively. Fluxes out of Lake Superior were measurable in 75% of the cases in which a concentration gradient was measured, and were in the range from -0.17 to +0.064 ng/m2‚h for R-HCH and from -0.60 to -0.093 ng/m2‚h for HCB.
Introduction Numerous methods exist for determination of air-surface exchange fluxes of gases (1-3), however the low ambient air * Corresponding author phone: (906) 487-3641; fax: (906) 4872943; e-mail:
[email protected]. † Michigan Technological University. ‡ Argonne National Laboratory. 10.1021/es050899q CCC: $30.25 Published on Web 09/30/2005
2005 American Chemical Society
concentrations of persistent bioaccumulative toxicants (PBTs; picograms per cubic meter concentrations; mixing ratios of parts per quadrillion (ppq) to sub-ppq) has typically limited the methods that can be used for these chemicals. For this and other reasons, methods to measure gaseous PBT airwater exchange fluxes are few, although this flux is often the largest component of atmospheric deposition of PBTs to surface water (4). The common method utilized to determine net air-water exchange fluxes of PBTs is to apply the Whitman two-film model (5). By use of this model for diffusive exchange in the air-water interface, PBT flux is estimated from gaseous concentration determined by high-volume air sampling, aqueous concentration, the temperature-corrected Henry’s law constant for a given compound, and parametrizations of the overall transfer coefficient, kol, based on wind speed and molecular properties. The known uncertainty of this method has been reported to be 50-7400% (6-9). The accuracy of the approach has not been assessed because direct methods to measure PBT flux have not been applied to measure air-water gas exchange of PBTs. However, the accuracy of the approach suffers from lack of consideration of atmospheric stability effects (10) and an inability to quantify effects on kol of waves, surface films, bubbles, and sea spray (11). Micrometeorological methods have been employed for decades to measure air-surface fluxes of trace gases, towers, ships, and buoys. However, they have seldom been employed for measurement of PBT fluxes, both because no sensors exist that are capable of measuring concentrations of PBTs at the frequency needed for direct covariance measurement and because adequate mass must be acquired in a short period of time (ideally 20-30 min) such that stationary atmospheric conditions are maintained over the time period during which covariances are computed. Various research groups have utilized micrometeorological methods to measure herbicide emission fluxes during and after agricultural field application (12-14). However, in these studies very high herbicide concentrations (on the order of micrograms per cubic meter) and fluxes (over a million times higher than those reported here) were observed. High concentrations made possible the collection of adequate analyte mass in a short time period by use of low-flow (100-300 mL min-1) minitubes that were thermally extracted or high-volume XAD resin or polyurethane foam (PUF) sampling (40 or 60 L min-1, respectively) and subsequent Soxhlet extraction. To obtain adequate mass to carry out flux measurements in Lake Superior, we designed and fabricated “low-flow” (13 L min-1) diffusion denuders as traps of gaseous PBTs after those designed and tested by Krieger and Hites (15, 16). We developed analytical methods to thermally extract the trapped PBTs directly into the inlet of a high-resolution gas chromatograph (GC). The diffusion denuders are part of an energy balance platform designed to measure air-water PBT gas fluxes from a stationary research vessel using the modified Bowen ratio approach (17, 18). Gas-phase PBTs are collected in capillary tubes in diffusion denuders because they diffuse to the capillary walls rapidly under laminar flow conditions, where they sorb to the stationary coating inside the capillaries. Particles, which have much lower diffusivities, do not reach the capillary walls before they exit the denuder. A filter placed posterior to a diffusion denuder in the flow path can be used to collect particulate-phase PBTs. The diffusion denuder consists of ca. 290 sections of commercially available fusedsilica capillaries sections in a Silcosteel-coated stainless steel tube. The capillaries are fused to one another and to the VOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8411
outer tube with polyimide resin, the same resin that is used to coat the outside of most commercial fused silica capillary columns. Thus, unlike the diffusion denuder design of Krieger and Hites, who used an epoxy to fuse the column sections together that cracked when heated, this diffusion denuder can withstand the same temperature conditions as the fused-silica capillary columns from which it is constructed. The objectives of the Lake Superior study were to (1) develop methods to measure concentrations and air-water exchange rates of PBTs, (2) carry out such measurements in fall 2002 and spring 2003 and compare them with conventional concentration and flux measurements, and (3) evaluate air-water exchange models by assessing the influence of meteorological parameters on the exchange rates in Lake Superior. We report here the methods developed and utilized to carry out field collection of gas-phase PBTs and meteorological data to compute fluxes, as well as the results of our micrometeorological flux measurements. The results demonstrate the constraints and utility of the measurement techniques. Another paper reports the analytical methods we developed to measure PBT mass in the samplers (19), and a third paper describes the performance of the diffusion denuders in sampling PBTs in ambient air (20). In subsequent publications we will compare the measurements reported here with conventional concentration and Whitman twofilm flux determinations and present an evaluation of airwater exchange models.
Experimental Section Diffusion Denuder Fabrication. Diffusion denuders consisted of a 1.6-cm o.d. × 25.5-cm stainless steel tube treated with Silcosteel (Restek Inc., Bellefonte, PA) containing 285289 25-cm sections of 30-m × 0.53-mm × 5-µm film thickness Zebron ZB-1 100% dimethylpolysiloxane capillary columns (Phenomenex Inc., Torrance, CA). Prior to cutting into sections, the 30-m capillary columns were conditioned at 330 °C for 4 h with nitrogen carrier gas flow in a GC. PI2525 Pyralin polyimide coating (HD Microsystems, Parlin, NJ) was used to cement the capillary columns together in the stainless steel tube. Following diffusion denuder assembly, the polyimide resin inside the diffusion denuder was hardened by a detailed curing process recommended by HD Microsystems. A post-cure procedure liberated compounds prior to use of diffusion denuders and resulted in clean chromatograms of thermal extractions of diffusion denuders (20). Analytical System and Methods. The analytical system and methods for thermal extraction and analysis of PBTs are described in detail in a related paper (19) and are only briefly described here. The analytical system consists of an Agilent Technologies 6890A Plus Series GC-micro-electron capture detector (microECD). The 6890A GC is equipped with a programmable temperature vaporization (PTV) inlet with liquid nitrogen cryogenic capability made by Gerstel Inc. (Baltimore, MD). The PTV inlet contains a Tenax-TA-packed inlet liner. The analytical column used for separation of target analytes was a DB-XLB 60-m × 0.25-mm i.d. × 0.25-µm film thickness (Agilent Technologies-J&W Scientific Inc., Folsom, CA) capillary GC column. The carrier gas was helium at a flow rate of approximately 2.1 mL min-1, which yielded a linear velocity of 36 cm s-1 in constant flow mode. Nitrogen at a flow rate of 60 mL min-1 was the makeup gas for the micro-ECD. Direct thermal extraction of analytes into the GC was carried out after modification of the 6890A GC by use of a thermal desorption unit (TDU) fabricated by CDS Analytical (Baltimore, MD). Analytes are swept out of the heated diffusion denuder and into the liquid nitrogen-cooled PTV inlet through a heated, deactivated fused silica capillary. A hot gas spike apparatus was fabricated and utilized to spike microliter quantities of standards in hexane into diffusion 8412
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 21, 2005
denuders for calibration curve and surrogate recovery determinations. For analysis of PBTs sampled from ambient air or spiked into diffusion denuders, the diffusion denuder is oriented in the TDU oven such that it is back-flushed during thermal desorption. It is connected at the top of the TDU oven with a custom 5/8-in. Swagelok to 1/16-in. Cajon adapter fitting by use of a Viton O-ring and at the bottom of the TDU oven with a custom 5/8- to 1/16-in. Swagelok reducing fitting. The following procedure describes the steps required for a ca. 55-min desorption of a diffusion denuder at a GC detector temperature of 300 °C, transfer line temperature of 260 °C, followed by a typical GC analysis. Initial conditions: GC oven temperature ) 100 °C, inlet mode ) split, total flow 32.1 mL min-1, PTV inlet temperature ) 200 °C; (1) heat TDU oven to 70 °C, set column flow to 0.8 mL min-1 and total flow to 500 mL min-1 to purge water from the diffusion denuder; (2) after 10 min, reduce the total flow to 32.1 mL/min and cool the PTV inlet to 5 °C; (3) initiate thermal desorption by increasing the TDU oven temperature to 230 °C; (4) when TDU oven temperature reaches 230 °C, wait 4 min to allow the temperature to equilibrate and then set total flow to 750 mL min-1; (5) after 11 min, turn off the TDU oven and transfer line heaters, set total flow to 6.0 mL/min; (6) set PTV inlet temperature to 10 °C and increase by 5 °C every 30 s to 105 °C; (7) once the PTV inlet temperature reaches 105 °C, set the column flow to 2.1 mL/min and the total flow to 32.1 mL/min and hold this condition for 10 min to allow any water in GC column to elute, then increase the PTV inlet temperature to 280 °C and begin GC analysis. The GC method used for analysis is as follows: inlet temperature ) 280 °C for duration of run; inlet mode ) splitless; split flow ) 30 mL min-1 after 6 min; column flow mode ) constant flow, 2.1 mL min-1; GC oven temperature: 100 °C for 6 min, 1 °C min-1 to 240 °C, 10 °C min-1 to 280 °C, hold at 280 °C for 20 min; detector temperature ) 300 °C; detector makeup gas ) nitrogen, 60 mL min-1. Calibration with Internal Standards and Calculation of Analyte Mass. Calibration was carried out by plotting the ratio, y:
y)
AAMIS AIS
(1)
where AA ) area of analyte peak, MIS ) mass of the internal standard injected into the gas chromatograph, and AIS ) area of the internal standard peak, versus x, where x ) mass of analyte in the standard injected into the GC. In analysis of PBTs in diffusion denuders, the mass of standard injected corresponds to that injected into the diffusion denuder by use of the hot gas spike apparatus. A quadratic equation was used to fit a plot of y versus x in order to describe the nonlinear variation in detector response with analyte mass in the concentration range of interest, as recommended by the U.S. Environmental Protection Agency (EPA) (21). Mass of analyte in a sample was calculated by finding the positive x-root value of the quadratic function. Micrometeorological Measurement of Fluxes. We fabricated, tested, and utilized two energy balance sampling platforms for large and small research vessels to measure fluxes. The micrometeorological principle utilized in the flux measurement is expressed in the modified Bowen ratio equation (1), which assumes that the transfer coefficient of two scalar quantities between two heights in the atmospheric surface layer is identical. The assumption is invalid when the flux of the chemical is not conserved; for example, when the time scales of physicochemical reactions at the two measurement heights are different (a gradient in temperature changing the vapor/ particle partitioning) or the chemical is advected into the measurement domain from another source
(an air mass containing a high concentration of PBTs). It has not been disproven within the precision of simultaneous measurements of two scalars (22), and was verified for CO2 and H2O (23). In this case, the flux of sensible heat is measured by direct covariance, and the modified Bowen ratio is the product of the sensible heat flux and the ratio of the difference in PBT concentrations measured at two heights above the air-water interface divided by the difference in temperature at the two heights:
PBT air-water exchange flux )
H ∆[PBT] Fcp ∆θ
(2)
where H is sensible heat flux (watts per square meter); Fcp, the product of the air density and the specific heat capacity of air at constant pressure, is 1.216 × 103 (W m-2)/(K m s-1); θ is potential temperature and corresponds to the temperature that air at a pressure p would have if it were at pressure p0, taken to be 1000 hPa; and ∆ symbolizes the difference in measured PBT concentrations or averaged potential temperature at two heights above the water surface. The difference in concentration of a PBT in eq 2 is computed from the concentrations measured in the diffusion denuders at the two platform heights. A diffusion denuder spiked with ca. 500 pg of surrogate standard polychlorinated biphenyl (PCB) 65 and loaded into an aspirated sampler housing was installed on the upper platform for sampling on either vessel. A second diffusion denuder in an aspirated sampler housing and spiked with a similar mass of surrogate standard was installed on the lower platform located at approximately 1 m above the water surface (0.4 m in the case of the lower platform on the R/V Agassiz). Entrance of water droplets and large particles into diffusion denuders was minimized in the design of the inlet of the aspirated housing, which consists of a 1-in i.d. stainless steel tube that makes a 90 ° bend, followed by 20 in-1 stainless steel mesh and a 4.5-in gap between the mesh and the denuder. For the measurements reported here, PBT sampling was conducted at an average flow rate of 13 L min-1 for 2 or 3 h with separate Hi-Lite 30 pumps from SKC Inc. (Eighty Four, PA) located on deck and connected to each diffusion denuder by Tygon tubing with a check valve in between. Flow rate was regulated by use of a AWM5104VN flow sensor (Honeywell, Inc., Minneapolis, MN). Swagelok fittings on the ends of diffusion denuders were capped prior to storage in a refrigerator. Immediately prior to analysis of PBTs, diffusion denuders were brought to room temperature, uncapped, and spiked with internal standards. The potential temperature difference in eq 2 is computed from temperature measurements at the upper and lower heights. These temperature measurements are carried out by use of thermocouples in aspirated sampler housings (Campbell Scientific Inc., Logan, UT), the signals of which were recorded by the CR5000 data logger (Campbell Scientific, Inc.). Average θ values were computed for each sampling period from 1-min aspirated thermocouple data at each height. Outlier values excluded from the calculation of the average were determined as any value greater or less than 3 times the standard deviation of the mean computed excluding the outlier. Sensible heat flux, H, utilized in eq 2, is computed as
H ) Fcp w′θ′
(3)
where w′ is the deviation from the mean vertical wind speed measured at 20 Hz, θ′ is the deviation from the mean potential temperature measured at 20 Hz, and the over bar indicates covariance determination over the measurement period. Sensible heat fluxes were computed as 2- or 3-h averages of 10-min averages. Outlier values excluded from the calculation
FIGURE 1. Upper and lower platforms for large research vessels on the R/V Lake Guardian. of the average were determined as for potential temperature. In eq 3, wind speed and temperature can be measured by use of a model CSAT3 (Campbell Scientific, Inc.) sonic anemometer and a fine wire thermocouple (FW05, Campbell Scientific Inc.). In this study the temperature calculated internally by the sonic anemometer was utilized to compute H. All signals (three-dimensional wind speed, thermocouple signal, internal sonic anemometer temperature) are recorded on the CR5000 data logger. These signals are corrected for axial acceleration according to methods developed by Edson et al. (24) using an attitude and reference system and a Model AHRS300CB, Crossbow Technology Inc. accelerometer (San Jose, CA) that measures and transmits roll, pitch, and heading (azimuth) angles to the CR5000 data logger. The accelerometer is completely enclosed inside a plastic container and fastened to the arm of the sonic anemometer. The data logger also records relative humidity and temperature by using a shielded Vaisala Model HMP45C unit and wind speed and direction with a RM Young wind monitor (both purchased from Campbell Scientific Inc.) measured at the upper platform. Field Sampling. Field sampling to measure ambient atmospheric concentrations and fluxes of seven target PBTs was carried out in summer and fall 2002 and spring, 2003 near the Keweenaw Peninsula in Lake Superior on the U.S. EPA’s research vessel the R/V Lake Guardian on three occasions and on Michigan Tech’s research vessel the R/V Agassiz on four occasions (Table 1). Micrometeorological sampling was carried out onboard the 180-ft R/V Lake Guardian while the vessel was stationary, utilizing upper and lower energy balance platforms for large research vessels (Figure 1). On this vessel, the upper platform is attached to the mast in the ship’s bow with a steel collar and bolts that tighten the collar at a height of approximately 5.0 m above the hotel deck (ca. 8.5 m above the water surface). At this location, mean flow tilt angles are expected to be approximately 5°, which has been considered to be acceptable for similar work (25-27). The lower platform consists of a VOL. 39, NO. 21, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8413
TABLE 1. Sampling Events date
Julian day
time
site designation
latitude/longitude (deg min)
Tavg (°C)
vessela
notes
7/7/02 7/8/02 7/8/02 7/8/02 9/17/02 9/20/02 10/3/02 4/26/03 4/26/03 4/26/03 4/26/03 4/27/03 4/27/03 5/25/03
188 189 189 189 260 263 276 116 116 116 116 117 117 145
2038-2238 0010-0210 0756-1004 1355-1555 1342-1542 1312-1348 1320-1523 1010-1315 1337-1642 1700-2001 2056-0002 0107-0409 0621-0954 0851-1154
SU20 SU17
46 50.00/90 17.00 47 09.86/89 39.72
HN050 HN050 HN050 HN050
47 17.10/88 36.96 47 17.10/88 36.96 47 17.62/88 38.27 47 17.23/88 36.87
20.0 19.9 19.5 17.3 19.4 16.1 10.6 3.14 2.97 5.04 6.47 6.06 5.97 4.75
LG LG LG LG A A A LG LG LG LG LG LG A
b b b b c c c
LG ) R/V Lake Guardian; A ) R/V Agassiz. during cruise. a
EH050 b
47 30.48/88 08.59
c
c
No motion correction data. No heading, wind direction, or water temperature data collected
hollow aluminum 2 in. × 2 in. × 6 m pole that is guided toward the water surface in an L-shaped steel guide that is bolted to the ship’s bow. The pole is fastened to the steel guide during sampling and retrieved to change diffusion denuders and to stow the platform while the ship is underway. Sampling was also carried out on Michigan Tech’s 36-ft R/V Agassiz while the vessel was stationary, utilizing a 10-m telescoping mast and upper and lower energy balance platforms for small research vessels.
Results and Discussion Sensible Heat Fluxes. On the basis of results for oceanic measurements (27), 180-min averaged sensible heat fluxes can be expected to exhibit an uncertainty of (10% when corrected for motion. Sensible heat fluxes could not be corrected for motion for some sampling events due to lack of required input data for these events as indicated in Table 1, footnotes b and c. Without motion correction under conditions in the open ocean (6-7° rolls, 2-3° pitch), the uncertainty in H doubles to (20% (C. Fairall, personal communication). No relationships were observed between relative standard deviations (RSDs) of accelerations or rolls and RSD of sensible heat flux for the nine periods in which these motion data are available, suggesting that motion does not significantly influence the variation in sensible heat flux. All data needed to carry out motion correction was available for the April 26-27, 2003 sampling event. During that event, platform motion was observed that is much less than that typically observed in oceanic measurement of trace gas fluxes (see, e.g., refs 28 and 29). During the six sampling periods on these dates, RSDs of three-dimensional (3-D) accelerations and rolls ranged from 0.01% to 0.03%. Because these parameters provide direct measurements of motion and they are quite invariant, sensible heat fluxes would be changed to a negligibly small extent through motion correction, and errors introduced through motion correction (27) would likely have negated improvements in sensible heat flux values through correction. Thus platform motion correction was not carried out on this data set. A conservative estimate for the uncertainty in our H values is (20%. Due in part to rainy conditions during measurements on April 27, sensible heat flux measured during the two periods on that day exhibited greater variability in 10-min sensible heat flux values. Variability during rainfall is caused by alteration of the time-of-flight of the ultrasonic signal between transducer heads. In fall 2004, improved ultrasonic signal processing software and wicks for transducer heads became available from Campbell Scientific Inc., which when installed on the sonic anemometer significantly improved performance under rainy conditions. 8414
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 21, 2005
FIGURE 2. Sensible heat flux out of Lake Superior measured during the events listed in Table 1 (filled circles; values reported in Table 4) and estimates of mean monthly average sensible heat fluxes from Lofgren and Zhu for 1992 to 1995 (asterisks; 30, 31). The values of sensible heat flux reported in Table 4 agree well with estimates of monthly average values reported for Lake Superior for 1992-1995 by Lofgren and Zhu (31, 32) (Figure 2). The estimates were based on AVHRR data and meteorological data from locations surrounding Lake Superior. Lofgren and Zhu did not constrain their calculations to produce a net zero annual total heat flux, and for Lake Superior found an imbalance in the net annual total heat flux of 21 W m-2 out of the lake. They did not account for effects of ice cover on winter flux values and suggested this could explain the imbalance. The good agreement between the measured and estimated fluxes may provide validity to Lofgren and Zhu’s estimates, or it may be a coincidence given that the estimations were carried out for a time period that begins 10 years earlier than our measurements. If sensible heat fluxes have not changed in the intervening time, the spring-summer-fall measurements reported here suggest that the corresponding estimates made by Lofgren and Zhu may be correct and that ice cover may account for the imbalance in their annual net total heat flux estimate. As pointed out by Lofgren and Zhu and observed by us, the measurements and estimations of H compared in Figure 2 are attenuated in late spring through early summer, and to a lesser degree during summer, due to an effect of static stability in the surface layer over the lake during this time period. These results demonstrate one of our motivations for carrying out over-water micrometeorological PBT flux measurements. They directly account for the effect of stability of the atmospheric surface layer, as well as all other sources of resistance, on air-water gas exchange. Over-land con-
TABLE 2. Method Detection Limits for the Target Analytes analyte
MDL (pg/m3 in 2.3 m3 of air)
n
R-HCH γ-HCH HCB PCB 18 PCB 44 PCB 52 PCB 101
1.6 14 8.5 10 9.9 28 23
3 6 8 8 3 6 6
centrations and fluxes can differ significantly from over-water values due to atmospheric stability differences. In such cases estimates of over-water fluxes based on over-land concentrations will be biased (10). Use of Diffusion Denuder Sampling and Thermal Extraction To Determine Gaseous PBT Concentrations. The performance of the diffusion denuder sampling/analysis techniques developed during this project are presented elsewhere (20); however, with respect to the flux measurements reported here it should be stated that diffusion denuders exhibited no leakage or cracking even though they were used repeatedly over the course of the three-year project, and in general, with the methods developed, the devices function well for collection, storage, and analysis of the target PBTs for micrometeorological flux measurement. Particles were not significantly retained in diffusion denuders in the size range from 0.05 to 1.0 µm. Side-by-side sampling was carried out in separate sampling trains, and differences in response were found to be insignificant, indicating that differences in concentrations for low-flow diffusion denuders at two heights are measurable. Recent modeling and experiments indicate strong temperature-dependent breakthrough behavior of target analytes. Experiments, in which pairs of denuders were connected in tandem such that the back denuder captured any breakthrough from the front denuder, indicated that no breakthrough (defined as mass collected on rear denuder less than the method detection limit or
