Portable Chamber System for Measuring Chloroform Fluxes from

Oct 24, 2013 - ... the concentration increase during a deployment was assumed to be linear. ... Science of The Total Environment 2016 554-555, 203-210...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/est

Portable Chamber System for Measuring Chloroform Fluxes from Terrestrial Environments − Methodological Challenges Lauren Pickering,† T. Andrew Black,† Chanelle Gilbert,† Matthew Jeronimo,† Zoran Nesic,† Juergen Pilz,‡ Teresia Svensson,§ and Gunilla Ö berg*,† †

University of British Columbia, Vancouver, Canada University of Klagenfurt, Klagenfurt, Austria § Linköping University, Linköping, Sweden ‡

S Supporting Information *

ABSTRACT: This study describes a system designed to measure chloroform flux from terrestrial systems, providing a reliable first assessment of the spatial variability of flux over an area. The study takes into account that the variability of ambient air concentrations is unknown. It includes quality assurance procedures, sensitivity assessments, and testing of materials used to ensure that the flux equation used to extrapolate from concentrations to fluxes is sound and that the system does not act as a sink or a source of chloroform. The results show that many materials and components commonly used in sampling systems designed for CO2, CH4, and N2O emit chloroform and other volatile chlorinated compounds (VOCls) and are thus unsuitable in systems designed for studies of such compounds. To handle the above-mentioned challenges, we designed a system with a non-steady-state chamber and a closed-loop air-circulation unit returning scrubbed air to the chamber. Based on empirical observations, the concentration increase during a deployment was assumed to be linear. Four samples were collected consecutively and a line was fitted to the measured concentrations. The slope of the fitted line and the y-axis intercept were input variables in the equation used to transform concentration change data to flux estimates. The soundness of the flux equation and the underlying assumptions were tested and found to be reliable by comparing modeled and measured concentrations. Fluxes of chloroform in a forest clear-cut on the east coast of Vancouver Island, BC, during the year were found to vary from −130 to 620 ng m−2 h−1. The study shows that the method can reliably detect differences of approximately 50 ng m−2 h−1 in chloroform fluxes. The statistical power of the method is still comparatively strong down to differences of 35 ng m−2 h−1, but for smaller differences, the results should be interpreted with caution.



INTRODUCTION

VOCls as well as interference from other anthropogenic sources (high ambient concentrations in relation to the fluxes), and it is not possible to follow concentration changes continuously, as done for gases that absorb light at a given wavelength (such as CO2, using near-infrared radiation). For such gases, the flux can, for example, be determined by measuring difference between the concentrations at the inlet and outlet of an open chamber system when it has reached steady state, while pumping air through the chamber.9 In theory, soil−atmosphere exchange of chloroform and other VOCls can be analyzed spectrophotometrically by the use of infrared photo acoustic spectroscopy (PAS),10,11 but the detection limit of PAS is too high for studies of VOCl in

Volatile organic chlorinated compounds (VOCls) are environmentally hazardous trace gases, participating in many atmospheric chemical processes, including stratospheric ozone destruction.1 There is evidence that terrestrial sources are major contributors to atmospheric VOCls and a number of studies confirm that VOCls, such as chloroform (CHCl3), are naturally formed in soil.2,3 Soil−atmosphere exchange of VOCls is a comparably new area of research and the number of field-based studies is limited, in total amounting to fewer than 25 publications in peer-reviewed journals.2−4 Soil−atmosphere exchange of volatiles is normally determined by the use of permanently installed chambers on a collar base inserted into the ground. Determining VOCl fluxes is a more challenging task than determining fluxes of many other greenhouse gases: they are generally present in comparably low concentrations (ppq to ppt levelspicogram to nanogram per liter), there is considerable risk of interference from equipment emitting © 2013 American Chemical Society

Received: Revised: Accepted: Published: 14298

July 10, 2013 October 21, 2013 October 24, 2013 October 24, 2013 dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305

Environmental Science & Technology

Article

Figure 1. Portable field chamber system developed in the present study. The left-hand side of the figure shows a person carrying the chamber and a case (model 1550, Pelican Products, Inc., Torrance, CA) containing pump, battery, and adsorbent tubes. The left-hand side of the schematic depicts the inside of the chamber: A, chamber air outlet; B, chamber air inlet; C, atmospheric pressure vent; D, mini-fan (model MC19623, Multicomp, USA); E, humidity and temperature probe (model HMP 45C, Vaisala Inc., USA); F, handles. The right-hand side depicts the setup of the system: G, diaphragm pump (model NMP850KNDCB, KNF Neuberger GmbH, Germany); H, 16-channel relay controller (SDM-CD16AC, Campbell Scientific Inc., Canada); I, adsorbent tubes (Carbotrap 300, Gerstel); J, two-way solenoid valves (EV-2M-12-H, Clippard Instrument Laboratory, Inc., Canada); K, exhaust tubes; L and M, inflow and outflow mass flow-meters (EM1NR4R0 V 1A, Sensirion AG, Switzerland); N, datalogger (1 Hz frequency, CR 1000, Campbell Scientific Corp., Canada); O, 12 V battery; P, needle valve. Lines show the airflow pathway (double solid blue lines), electrical connections (solid red lines), and data exchange (solid black lines). A Bev-A-Line tubing (2.5 m, 3 mm I.D. × 6 mm O.D., Thermoplastic Processes, Inc., USA) connects the chamber air in- and outlet to the pump.

estimates; (iii) test the chamber material and other parts of the sampling system to ensure the system does not act as a sink or a source of VOCls, focusing on chloroform (CHCl3), which is known to be formed naturally, and two other VOCls, carbon tetrachloride (CCl4), and methyl chloroform (CH3CCl3) that are believed to be mainly of anthropogenic origin; (iv) identify a suitable sampling regime (e.g., optimal sampling volume and sampling time); and (v) determine the resolution of the method.

ambient air, and can thus not be used for detection of VOCls in the field. At present, the only option to measure VOCls in the field is to withdraw a sufficient volume of concentrated air followed by analysis using gas chromatography.12,13 A recent study suggests that the emission of chloroform varies spatially, with local hot-spots emitting an order of magnitude or more compared to areas only a few meters away.4 The findings are not surprising, as it is well-known that the emission of other biogenically produced gases, such as CO2, N2O, and CH4, shows a strong spatial variability.5−8 A reliable flux estimate of a compound that exhibits an unknown spatial flux distribution requires a large number of sampling points, which in most situations makes permanent chambers logistically or economically unfeasible. The alternative is to carry out a first assessment to identify spatial patterns, including high and low flux locations, and then place a more feasible number of permanent chambers strategically. For the reasons given above, available methods do not, however, allow a first assessment of the spatial variability of VOCl fluxes. The fact that data is scarce further complicates matters: the uncertainty in measured concentrations is not known, nor has it been studied regarding how much the concentrations vary in ambient-air concentrations in time or space. This makes it impossible to reliably compare ambient air data from different locations or different times. The aim of the present study was to develop a robust method of measuring VOCl soil-to-atmosphere flux measurement that allows a reliable first assessment of the spatial variability of an area to facilitate identification of high- and low-flux areas, and to facilitate analyses of such fluxes in light of short- and longterm temporal patterns. We set out to develop a portable chamber system that could be handled by one person and permit the collection of up to 100 samples at a field site within a reasonable time frame (1−2 days). The objectives of our study were to (i) identify a general design that would meet our aim, (ii) develop quality assurance procedures for flux



SYSTEM DESIGN AND SAMPLING PROCEDURE The experiments presented below led us to a final design consisting of a cylindrical chamber with handles (500 mm inner diameter × 290 mm height and four mm wall thickness, corresponding to a volume of 57 L and with a weight of 5 kg) and a side-unit containing pump, batteries, and sampling tubes, fitted into a case (model 1550, Pelican Products, Inc., Torrance, CA; 520 mm L × 430 mm W × 210 mm H, total weight 14 kg; Figure 1). For reasons outlined below, we decided to develop a non-steady-state system comprising of an aluminum chamber, a closed-loop air-circulation unit with a diaphragm pump, and a VOCl sampling unit with carbon-based adsorbent tubes, thus continuously diluting the air in the chamber. During a deployment, the sampling system collects four sequential air samples: a chamber deployment starts when the chamber is set on the ground and the pump is turned on, and lasts until four samples have been collected in the adsorbent tubes (ca. 16 min). The air returned to the chamber is scrubbed to remove compounds being sampled. The rate of change of the concentration (α) is determined by fitting a straight line through the four points, using the method of least-squares and extrapolating the line to t = 0 to obtain the ambient air concentration (Co) (Figure 2). Data quality in each deployment is assessed by determining the variation from linearity by calculating the root-mean-square error (RMSE). Both of these variables (α and C0) are input data in the equation used to 14299

dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305

Environmental Science & Technology

Article

where V is the chamber volume (0.057 m3), A is the area of the soil covered by the chamber (0.196 m2), F is the flux of compound coming from the soil (ng m−2 h−1), and Q is the sampling flow rate (∼0.042 m3 h−1). Both F and Q are assumed to be constant. At the time of the chamber closure t0 = 0, the initial compound concentration is C0. Since the system time constant (V/Q) is ∼83 min, the concentration increase with time was nearly linear for the duration of chamber deployment (∼16 min, see Supporting Information). Care was taken to properly align the four measured average concentrations with midpoints of each sample cycle. The time tx was the midpoint of the last (fourth) sample period. The rate of change of the concentration (α) was determined by fitting a straight line through the four points, using the method of least-squares and extrapolating the line to t = 0 to obtain the ambient air concentration (Co) (see Figure 2). Collars, which require insertion into the ground, were not used in our study because the disturbance caused by insertion can significantly affect fluxes (e.g., Norman et al. 199718). This means that there might be an exchange of air between the chamber and the ambient air. An intermittent influx of ambient air would cause deviations from the model, which in turn would result in a large RMSE; however, a steady influx of ambient air throughout a deployment would impact the flux but have little impact on the RMSE. To minimize the risk of introducing such an error, the inflow and outflow of air through the system was carefully monitored, and they matched within 2−5 mL min−1. We modeled the impact of an inflow of 5 mL min−1, assuming ambient air concentrations and fluxes corresponding to those estimated in the deployments made in May 2012 and found that an inflow of this size would have a negligible impact on flux estimates. Reliability and Resolution of the Method. Reliable assessment of spatial and temporal variability of an area requires that the method can detect whether (1) flux at location A and location B differ at time t1 and whether (2) the flux at location A at time t1 and t2 differ. Complex data rarely allows simple true−false answers and various statistical methods have been developed to determine the likelihood that a conclusion is true or false. To reliably answer the two questions above, it is necessary to know the variability of the method, a fact that unfortunately is overlooked in many studies. When the variability of the method is known, it is not only possible to answer the two questions, but it is also possible to determine the reliability of the answers by carrying out a statistical power analysis (see, for example, ref 19). A power analysis is a method that helps assess the reliability of the results, i.e., in our case, at what resolution the developed method can reliably determine whether or not the observed fluxes differ. Two different types of error must be kept in mind: (1) the risk of saying that there is a difference, even though there is none (type I error), and (2) the risk of saying that there is no difference, although there is in fact a difference (type II error). In the first case we wrongly reject the null hypothesis (no difference) and incur a type I error; in the second case we accept the null hypothesis although the alternative is true, which causes a type II error. When carrying out a power analysis, one starts by determining how large the type I error is that one is willing to accept (the significance level), which commonly is set to 0.01, 0.05, or 0.1 (i.e., willingness to accept a 1%, 5%, or 10% risk of committing a type I error). The power of a method is the probability of not encountering a type II error (i.e., a. false negative), i.e., the likelihood of correctly

Figure 2. Measured and modeled concentrations of chloroform during a chamber closure using the closed-loop portable chamber system developed in the present study (Figure 1). Four consecutive air samples were collected on carbon-based adsorbents, and analyzed by gas-chromatography. The least-squares method was used to create a linear fit through the four measured points (dotted line), which was used to estimate the initial concentration C0 (y-intercept) and the slope of the line (α). These were input variables in the equation used to calculate the flux (eq 1).

calculate the flux (eq 1 and Figure 3). The ability of the method to detect a specified difference in flux between locations or time is determined by a statistical power analysis (Table 1).



THEORY Closed systems are commonly used in soil−atmosphere exchange studies.14 In a closed system, the concentration inside the chamber is allowed to increase (if there is a net flux from soil to air) or decrease (if there is a net flux from air to soil) and the flux is calculated from the concentration change over time. It is known that the concentration of VOCls in ambient air normally is at the ppq to ppt level (picogram to nanogram per liter), or lower.15,16 The detection limit for VOCls on commercially available GC-ECD systems varies from 0.01 to 24 ng L−1.17 Taken together, calculating fluxes based on concentration measurements requires high precision, which is why several liters of air must be collected to achieve reliable flux estimates, even when using a GC-ECD system (gas chromatograph with an electronic capture detector). Therefore, it is not possible to construct a chamber, which is small enough to be handled by one person, while at the same time allowing collection of a sufficiently large volume without risking adverse effects caused by inducing under-pressure. For this reason, closed chambers were considered unsuitable. Instead, we chose to design a chamber system, which makes it possible to keep the pressure steady by replacing the sampled air (Figure 1).18 Flux Equation. The mass of a certain compound in each sample was used to calculate the average concentration of that compound in the chamber air during a particular sampling period, and the rate of concentration change over time was used to calculate the flux as follows (see supplement for details): αtx Q + C0 F= A −Q / Vtx A (1 − e ) Q

(1)

14300

dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305

Environmental Science & Technology

Article

Figure 3. Chloroform concentrations vs time for twelve deployments made in May 2012 at a site located 10 km SW of Campbell River on the east coast of Vancouver Island, BC, Canada (49°52′ N, 125°20′ W). Also shown are the corresponding fluxes and values of RMSE. The flux estimates are based on four consecutively collected samples with air being circulated through each adsorbent tube for approximately five minutes (exact time and volume recorded). A line was fitted through the measured concentrations to determine the slope (α and the initial concentration at time 0 (C0)). Fluxes (F) were obtained using eq 1 with the values of V and A 0.057 m3, and 0.196 m2, respectively. The value of Q used in each calculation is the average flow rate for the particular deployment. The overall average value for these twelve deployments was 0.041 m3 h−1.

detecting a real difference, for a given significance level. For a treatment of the calculation of power and variability from given sample sizes and significance levels, see Section 2.2 in Rasch et al.19

Table 1. Results of the Analysis of the Statistical Power of the Method with Regard to Its Ability to Detect Whether or Not Fluxes of Chloroform Differ (in Time or Space)a detected difference (ng m‑2 h‑1)

power (type I error = 0.01)

power (type I error = 0.05)

power (type I error = 0.10)

5 10 15 20 25 30 35 40 45 50

0.014 0.034 0.073 0.141 0.241 0.371 0.518 0.663 0.787 0.879

0.057 0.114 0.203 0.324 0.468 0.617 0.749 0.853 0.923 0.964

0.102 0.186 0.303 0.445 0.594 0.731 0.839 0.914 0.923 0.983



METHODS General Considerations. A reliable geospatial analysis requires sampling of at least 60−80 locations distributed over the study area (in the present study ∼50 × 50 m), which is why permanently installed chamber collars with a top closure and canister sampling followed by GC-ECD analysis3,20,21 is not a logistically viable solution. To reduce both the volume and the weight of the samples, we choose preconcentration in the field using adsorbent-based tubes.22−24 As a basis for the method development, we used the flux chamber system developed by the Biometeorology and Soil Physics Group at the University of British Columbia, Canada, for CO2 flux estimates.25 The original system has been used for field flux measurements since 2000, and its reliability has been carefully evaluated.25,26 Resolution of the Method. The ability of the method to detect a specified difference in flux between two samples (in time or space) was determined by a statistical power analysis of 89 deployments made in May 2012 at a site located 10 km SW

a

The second column gives the power (probability of not encountering a type II error, i.e., the likelihood of correctly detecting a real difference) for a fixed probability of 0.01 for a type I error (the risk of saying that there is difference, even though there is none). Similarly, the third and fourth columns provide the powers for type I error probabilities (significance levels) of 0.05 and 0.1, respectively. The analysis is based on 89 deployments, May 2012.

14301

dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305

Environmental Science & Technology

Article

Bev-A-Line (Licor Bev-a-line IV, 1/4) connected to a Swagelok connector inside the chamber (C in Figure 1), buffering any potential under- or overpressure without allowing air to enter the chamber. After closing the chamber, five air samples (40 L each) were withdrawn consecutively from the chamber at an approximate flow rate of 700 mL min−1 (the exact flow rate was recorded). The air, which was withdrawn from the chamber, flowed into an adsorbent tube and the scrubbed air was returned to the chamber. VOCl concentrations in the adsorbent tubes were measured by GC-ECD with thermal desorption autosampler as described in the Chemical Analysis section of the SI. Effect of Storage. The potential effect of storage was tested as follows: Storage on dry ice in a metal cooler was tested by spiking 35 tubes with a known amount of chloroform, carbon tetrachloride, and methyl chloroform. Ten tubes were analyzed directly after spiking, while the remaining tubes were placed in a metal cooler with 20 kg of dry ice. Three tubes were withdrawn and analyzed every 12 h until 72 h had passed, at which point the remaining ten tubes were analyzed. Storage in freezer at −5 °C was tested by spiking 32 tubes with a known amount of chloroform, carbon tetrachloride, and methyl chloroform. Ten tubes were analyzed directly after spiking, while the remaining tubes were placed in a freezer at −5 °C. Two tubes were removed from the freezer every 48 h and analyzed. After two weeks the remaining ten tubes were removed and analyzed.

of Campbell River on the east coast of Vancouver Island, British Columbia, Canada (49°52′ N, 125°20′ W), at an elevation of 300 m above sea level. It is part of the seasonal dry variety of the temperate rain forest that covers much of North America’s Pacific Northwest. Its biogeoclimatic characterization is dry maritime with Douglas fir succeeding to coastal western hemlock (for a more detailed description of the site, see: field sites/Campbell River at http://www.landfood.ubc.ca/swal/). The analyses were carried out at three significance levels, 0.01, 0.05, and 0.10, which means that there is a 1%, 5%, and 10% risk that we conclude that two fluxes do not differ when they actually do. Usually, the power cannot be controlled for a given probability α of the error of the first kind (i.e., the significance level). However, the observed flux data were nearly symmetric around its median (as confirmed by a Wilcoxon test with a pvalue greater than 0.9) and, after removal of a single outlying value, the values approximately follow a normal distribution, leading us to undertake a t test to control β and the power, respectively. Material Testing. System Components. To test if any VOCls escaped through the adsorbent tubes, two tubes were placed in series connected by a Swagelok union and sealed with a graphite ferrule, similar to the experiment described by Gallego et al.27 A series of tests were conducted with different sampling volumes (for additional details, see System Components in the SI and Figure S4). Potential VOCl contamination in the exhaust air from the pump outlet was tested by passing scrubbed air (ca. 200 L) through two adsorbent tubes placed in parallel after the pump in the setup described in Figure 1. The test was repeated 10 times for two different pumps (total 20 exhaust tubes), and the tubes were analyzed by GC-ECD (Agilent 7890) with thermal desorption. Potential contamination from the system’s internal components, such as the flow-meters, the solenoid valves, and the connecting tubing, was tested by disconnecting the system in Figure 1 from the chamber and passing ultrapure nitrogen gas (40 L, 5.0 grade) through a single adsorbent tube. The test was repeated in five replicates each at a flow rate of 700 mL min−1 (same as in the field situation) and the tubes were analyzed by GC-ECD with thermal desorption. Two separate experiments were conducted to investigate if the tubing (a) adsorbed or (b) emitted VOCls (for additional details, see System Components in the SI and Figures S5 and S6). Chambers. Potential contamination from the chamber was studied through concentration depletion tests, which involved circulating air through a sealed chamber for an extended period of time to reduce the concentration of VOCls until it approached zero. Initially, tests were conducted with the acrylic/ABS chambers used by the UBC Biometeorology and Soil Physics Group for standard measurements of CO2 flux.26 We found that these types of chambers contaminated the air with VOCls, which is why testing proceeded with an aluminum chamber. Tests were performed with a system identical to the chamber design given in Figure 1, but with a fitted aluminum bottom to prevent ambient air from entering the chamber. Tests were conducted with empty chambers (plus fan) and with chambers fitted with temperature and relative humidity probes, debris screen, and an atmospheric pressure vent, sealed and unsealed. The atmospheric pressure vent was included to reduce any pressure differentials between the chamber air and the soil or surrounding air. The vent consisted of a piece of



RESULTS AND DISCUSSION All results based on GC-ECD analysis take into account the calculated limit of detection (LOD) and limit of quantification (LOQ) presented in Table S1. Flux Calculations. The soundness of the assumptions and equations given in the Theory section were tested and found reliable by plotting modeled concentrations against measured concentrations for a series of measurements (see Figure 2). When designing the sampling procedure during one deployment we considered the following parameters: the number of samples to be collected, the length of time for each sample, the sampling flow rate, and the volume. We wanted the number of samples to be large enough to reliably note whether errors were occurring and not creating false positives/negatives/nondetects. From a statistical point of view, this means that the number of samples during a deployment must be larger than three. We wanted the time of sampling to be short enough to reduce any impacts on the fluxes themselves and maintain the linearity, but long enough to collect a sufficient amount of volume to analyze in lab. Our tests suggested that the concentration increase is generally linear for at least 30 min. Various volumes were tested, and we found that 2 L was not sufficient to reliably determine the concentrations in the air, as they occasionally approached the detection limit, whereas 3 L consistently gave sufficient amounts for reliable analysis on the GC-ECD. We tested high and low flow-rates and found that the mechanics of the system design worked best at 700 cm3 per minute or lower. The length of time is directly related to the volume and the flow-rate, and we wanted to minimize the sampling time as much as possible (to minimize the risk of collecting beyond the linear concentration increase and maximize the number of samples collected in the field). The reasoning above led us to a design with four five-minute samples collected consecutively at a flow rate of 700 cc min−1. 14302

dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305

Environmental Science & Technology

Article

returned to the chamber (Figure 1), and routinely replaced with conditioned tubes (300 °C for 60 min) after sampling 200 L of air. The tests with the Bev-A-Line tubing did not provide evidence that the tubing adsorbed any VOCls. Continued testing suggested, however, that the components in the system (fan, probe), as well as the Bev-A-Line tubing, emitted minute amounts of VOCls (see details in Material Testing − System Components in SI). For example, letting 500 L of air pass through the chamber with fan and probes produced a contamination of 0.3 ng chloroform, and running 30 L of air through 25 m of Bev-A-Line tubing gave similar results. These results were followed up with depletion tests with empty chambers as well as with probes inside the chamber. The measured values for the empty chambers followed the modeled values as these asymptotically approached zero, implying that emissions of VOCls from empty chambers were negligible. Comparison with modeled values suggests that the probes did not emit any detectable amounts of carbon tetrachloride, but that they did emit small amounts of chloroform, and methyl chloroform (for details: see Figure S7). These results indicated that contamination from the fan, probe, tubing, and other components might pose a problem when analyzing samples with low concentrations of VOCls. However, the tests in the present study were conducted with considerably larger volumes of air (500 L and 30 L), longer Bev-A-Line tubing (25 m), and for longer periods of time (4 h or more) than what is used under normal sampling conditions (3 L, 2.5 m, 4 min). Overall, these results suggest that internally caused contamination must have been negligible during a normal deployment (less than 30 pg), even when analyzing air samples with concentration levels near the detection limit. This conclusion was confirmed by various tests carried out under conditions similar to those during a normal sampling round, which gave nondetectable results (data not shown). There are indications that plants might emit VOCls,28 but it is unknown whether their formation is related to photosynthesis, and tests suggest that light does not impact the emission of VOCls from soil.2 Still, it should be noted that the chambers used here are nontransparent, and we cannot exclude the possibility that this might impact the flux. Effect of Storage. The experiments carried out suggest that the storage procedure did not have an impact on the amount of VOCls in the samples: no statistically significant trend was observed after 72 h of storage on dry ice or after two weeks of storage in a −5 °C freezer (Mann-Kendall trend test, α = 0.05), implying that the observed variability was caused by other factors than storage. Given the sample size of eight samples, the variability allows us to detect differences of about twice the magnitude of the standard deviations (with significance level 0.05 and a power 0.9). This allows us to conclude that we do not run the risk of overlooking a trend, since the observed differences in the means before and after storage are smaller than one standard deviation, for the samples stored on dry ice as well as for the samples stored in the freezer. Utility and Reliability of Method. The present study is the first systematic evaluation of the reliability of all steps involved in flux estimates of chloroform and other VOCls in the field, including assumptions made regarding the potential impact of variable ambient air concentrations, the flux equation used, the statistical power of the method, and potential errors introduced by the various physical components in the system. It is also the first presentation of a portable system that can be

As indicated earlier, the reliability of the measurements was tested for each deployment, using the least-squares method to estimate the slope and variability expressed by the root-meansquare error (RMSE). The larger the RMSE, the lower the power of the method is (see below under Resolution of Method). A sample of typical measurements with the chosen design is given in Figure 3. As mentioned earlier, the fact that we did not use collars inserted into the ground means that there might have been some exchange of air between inside and outside the chamber at the ground surface. The fact that an influx of ambient air at the base of the chamber could cause an increase in RMSE implies that the method cannot be used to determine whether a large RMSE is due to a leakage at the base of the chamber. However, the fact that we were able to match in- and outflow with an accuracy of 2−5 mL min−1, in combination with the fact that an inflow of that size was estimated to have a negligible effect, suggests that exchange of air at the base of the chamber did not interfere with the flux estimates. A detailed geostatistical analysis of the data from the May 2012 deployments will be reported in a forthcoming paper (Ö berg et al., in prep). Resolution of the Method. The power analysis based on 89 deployments shows that for a difference of 50 ng m−2 h−1, the power of the method is 88%, when choosing a significance level of 0.01, 96% for a significance level of 0.05, and 98% for a significance level of 0.1 (Table 1). For example, the detected flux at location F was 48 ng m−2 h−1 larger than the flux detected at location D (Figure 3). From Table 1 we can conclude that the likelihood that this difference is real is 88% to 98% (depending on which significance level one chooses). This example illustrates that the power analysis demonstrates that the method is quite robust when it comes to detecting differences around 50 ng m−2 h−1. The power is still comparably strong down to differences of 35 ng m−2 h−1, but for smaller differences, the results should be interpreted with caution. For example, for differences of 5 ng m−2 h−1, the power of the method is as low as 10%, even when choosing a significance level of 0.1. By using Figure 3 in combination with using Table 1, we can conclude that the likelihood that the observed difference in flux measured at locations G and H is real is only 10% (and the likelihood that the detected difference is false is consequently 90%). Material Testing. System Components. The breakthrough tests gave a loss of less than 0.03 ng for all three compounds with chloroform showing the highest amounts, even when passing air samples that were 10 and 60 times larger than the normal sampling size. This suggests that loss through breakthrough is not a concern for the proposed method (for details, see Material Testing − System Components in SI). Initially, a commercially available sampling system was used, but a custom built sampling system was eventually constructed, due to intermittent malfunctions and breakdowns of the commercially available pump when operated under field conditions, in combination with contamination of the samples from the exhaust. Exhaust tests were also carried out with the custom built system, and these tests suggested that the pump emitted low, yet detectable amounts of VOCls: 200 L of air passing through the pump gave 0.21 ± 0.09 ng, 0.14 ± 0.23 ng, and 0.12 ± 0.17 ng for CHCl3, CCl4, and CH3CCl3, respectively. To minimize potential contamination from the pump, two carbon based tubes of the same type as those used for sampling were placed in the exhaust line before the air is 14303

dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305

Environmental Science & Technology

Article

(4) Albers, C. N.; Jacobsen, O. S.; Flores, É. M. M.; Pereira, J. S. F.; Laier, T. Spatial Variation in Natural Formation of Chloroform in the Soils of Four Coniferous Forests. Biogeochemistry 2010, 103 (1−3), 317−334. (5) Groffman, P. M.; Butterbach-Bahl, K.; Fulweiler, R. W.; Gold, A. J.; Morse, J. L.; Stander, E. K.; Tague, C.; Tonitto, C.; Vidon, P. Challenges to Incorporating Spatially and Temporally Explicit Phenomena (Hotspots and Hot Moments) In Denitrification Models. Biogeochemistry 2009, 93 (1−2), 49−77. (6) Parkin, T. B. Soil Microsites As a Source of Denitrification Variability. Soil Sci. Soc. Am. J. 1987, 51, 1194−1199. (7) McClain, M. E.; Boyer, E. W.; Dent, C. L.; Gergel, S. E.; Grimm, N. B.; Groffman, P. M.; Hart, S. C.; Harvey, J. W.; Johnston, C. A.; Mayorga, E.; McDowell, W. H.; Pinay, G. Biogeochemical Hot Spots and Hot Moments at the Interface of Terrestrial and Aquatic Ecosystems. Ecosystems 2003, 6 (4), 301−312. (8) Rayment, M. B.; Jarvis, P. G. Temporal and Spatial Variation of Soil CO2 Flux in a Canadian Boreal forest. Soil Biol. Biochem. 2000, 32, 35−45. (9) Griffith, D. W. T.; Leuning, R.; Denmead, O. T.; Jamie, I. M. Air−land Exchanges of CO2, CH4 and N2O Measured by FTIR Spectrometry and Micrometeorological Techniques. Atmos. Environ. 2002, 36, 1833−1842. (10) Dumitras, D. C.; Dutu, D. C.; Matei, C.; Magureanu, A. M.; Petrus, M.; Popa, C. Laser Photoacoustic Spectroscopy: Principles, Instrumentation, And Characterization. J. Optoelectron. Adv. Mater. 2007, 9 (12), 3655−3701. (11) Bozóki, Z.; Pogány, A.; Szabó, G. Photoacoustic Instruments for Practical Applications: Present, Potentials, and Future Challenges. Appl. Spectrosc. Rev. 2011, 46 (1), 1−37. (12) Liu, X.; Pawliszyn, J. On-Site Environmental Analysis by Membrane Extraction with a Sorbent Interface Combined with a Portable Gas Chromatograph System. Int. J. Environ. Anal. Chem. 2005, 85 (15), 1189−1200. (13) Wolfe, W. J.; Williams, S. D. Soil Gas Screening for Chlorinated Solvents at Three Contaminated Karst Sites in Tennessee. Ground Water Monit. Rem. 2002, 22 (4), 91−99. (14) Rochette, P.; Bertrand, N. Soil-surface gas emissions. In Soil sampling and methods of analysis; Carter, M., Gregorich, E. G., Eds.; CRC Press: Boca Raton, 2007; pp 851−861. (15) Zoccolillo, L.; Amendola, L.; Insogna, S. Comparison of Atmosphere/Aquatic Environment Concentration Ratio of Volatile Chlorinated Hydrocarbons between Temperate Regions and Antarctica. Chemosphere 2009, 76 (11), 1525−1532. (16) Artuso, F.; Chamard, P.; Chiavarini, S.; diSarra, A.; Meloni, D.; Piacentino, S.; Sferlazzo, M. D. Tropospheric Halocompounds and Nitrous Oxide Monitored at a Remote Site in the Mediterranean. Atmos. Environ. 2010, 44 (38), 4944−4953. (17) Dobrzyńska, E.; Pośniak, M.; Szewczyńska, M.; Buszewski, B. Chlorinated Volatile Organic Compounds - Old, However, Actual Analytical and Toxicological Problem. Crit. Rev. Anal. Chem. 2010, 40 (1), 41−57. (18) Norman, J. M.; Kucharik, C. J.; Gower, S. T.; Baldocchi, D. D.; Crill, P. M.; Rayment, M.; Savage, K.; Striegl, R. G. A Comparison of Six Methods for Measuring Soil Surface Carbon Dioxide Fluxes. J. Geophys. Res. 1997, 102 (D24), 771−777. (19) Rasch, D.; Pilz, J.; Verdooren, R.; Gebhardt, A. Optimal Design of Experiments with R.; Chapman and Hall/CRC: Boca Raton, 2011. (20) Teh, Y. A.; Mazeas, O.; Atwood, A. R.; Abel, T.; Rhew, R. C. Hydrologic Regulation of Gross Methyl Chloride and Methyl Bromide Uptake from Alaskan Arctic Tundra. Global Change Biology 2009, 15 (2), 330−345. (21) Rhew, R. C.; Miller, B. R.; Weiss, R. F. Natural Methyl Bromide and Methyl Chloride Emissions from Coastal Salt Marshes. Nature 2000, 403, 292−295. (22) Noy, T.; Fabian, P.; Borchers, R.; Janssen, F.; Cramers, C.; Rijks, J. Trace Analysis of Halogentated Hydrocarbons in Gaseous Samples by on-Line Enrichment in an Adsorption Trap, On-Column Cold-

used to reliably carry out an assessment of the spatial variability of VOCls emissions at a site. The study shows that many materials and components commonly used in sampling systems designed for CO2, CH4, and N2O emit VOCls and are thus unsuitable in systems designed for such studies. Furthermore, the study suggests that ambient air concentrations show a dramatic spatial variability. To deal with these issues in a reliable manner, we designed a system with a non-steady-state chamber and a closed-loop aircirculation unit with a pump that returns scrubbed air to the chamber. Even so, the method demonstrates some variability even with these precautions, as is often the case in field studies. The variability of the method, in combination with the large risk of contaminating the samples, requires that any system designed to reliably measure VOCls fluxes from terrestrial sources should include a quality assurance system. It is essential that the power of the method is analyzed, otherwise it is not possible to determine the resolution of the method. The method presented in the present paper can reliably detect differences of approximately 50 ng m−2 h−1 in chloroform fluxes. The statistical power of the method is still comparatively strong down to differences of 35 ng m−2 h−1, but for smaller differences, the results should be interpreted with caution. One person can carry out 20−25 deployments in a day with one system. A rigorous geospatial analysis requiring 100 samples would thus require sampling for four days with one system, two systems for a period of two days, or four systems during one day. It remains to be determined at what scale the temporal variability is likely to interfere with the spatial assessments.



ASSOCIATED CONTENT

S Supporting Information *

Additional experimental detail and graphics. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This study was made possible through funding from the Canadian Foundation for Innovation (CFI), The Canadian Natural Science and Engineering Research Council (NSERC), and B.C. Knowledge Development Fund (BCKDF), which hereby is warmly acknowledged.



REFERENCES

(1) Montzka, S. A.; Reimann, S.; Engel, A.; Krüger, K.; O’Doherty, S.; Sturges, W. T., Ozone-Depleting Substances (ODSs) and Related Chemicals. In Scientific Assessment of Ozone Depletion: 2010; World Meteorological Organization: Geneva, Switzerland, 2011; Vol. 52. (2) Rhew, R. C.; Teh, Y. A.; Abel, T.; Atwood, A.; Mazeas, O. Chloroform Emissions from the Alaskan Arctic Tundra. Geophys. Res. Lett. 2008, 35 (21), 5. (3) Dimmer, C. H.; Simmonds, P. G.; Nickless, G.; Bassford, M. R. Biogenic Fluxes of Halomethane from Irish Peatland Ecosystems. Atmos. Environ. 2001, 35, 321−330. 14304

dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305

Environmental Science & Technology

Article

Trapping and Capillary Gas Chromatography. I. Method and Instrumentation. J. Chromatogr. 1987, 393, 343−356. (23) McCaffrey, C. A.; MacLachlan, J.; Brookes, B. I. Adsorbent Tube Evaluation for the Preconcentration of Volatile Organic Compounds in Air for Analysis by Gas Chromatography-Mass Spectrometry. Analyst 1994, 119, 897−902. (24) Dewulf, J.; Vanlangenhove, H.; Huybrechts, T. Developments in the Analysis of Volatile Halogenated Compounds. TrAC Trends Anal. Chem. 2006, 25 (4), 300−309. (25) Drewitt, G. B.; Black, T. A.; Nesic, Z.; Humphreys, E. R.; Jork, E. M.; Swanson, R.; Ethier, G. J.; Griffis, T.; Morgenstern, K. Measuring Forest Floor CO2 Fluxes in a Douglas-Fir Forest. Agricultural and Forest Meteorology 2002, 110, 299−317. (26) Jassal, R. S.; Black, T. A.; Nesic, Z.; Gaumont-Guay, D. Using Automated Non-Steady-State Chamber Systems for Making Continuous Long-Term Measurements of Soil CO2 Efflux in Forest Ecosystems. Agricultural Forest Meteorology 2012, 161, 57−65. (27) Gallego, E.; Roca, F. J.; Perales, J. F.; Guardino, X. Comparative Study of the Adsorption Performance of a Multi-Sorbent Bed (Carbotrap, Carbopack X, Carboxen 569) and a Tenax TA Adsorbent Tube for the Analysis of Volatile Organic Compounds (VOCs). Talanta 2010, 81 (3), 916−924. (28) Laturnus, F.; Matucha, M. Chloride - a Precursor in the Formation of Volatile Organochlorines by Forest Plants? J. Environ. Radioact. 2008, 99 (1), 119−125.

14305

dx.doi.org/10.1021/es403062c | Environ. Sci. Technol. 2013, 47, 14298−14305