Empirical Investigation of the Junge ... - ACS Publications

Mar 13, 2009 - Lancaster Environment Centre, Lancaster University,. Lancaster LA1 4YQ ... of variability in long-term monitoring data, using the techn...
2 downloads 4 Views 192KB Size
Environ. Sci. Technol. 2009, 43, 2746–2752

Empirical Investigation of the Junge Variability-Lifetime Relationship Using Long-Term Monitoring Data on Polychlorinated Biphenyl Concentrations in Air S A R A B E C K E R , † C R I S P I N J . H A L S A L L , * ,† MATTHEW MACLEOD,‡ MARTIN SCHERINGER,‡ KEVIN C. JONES,† AND ¨ HLER‡ KONRAD HUNGERBU Lancaster Environment Centre, Lancaster University, Lancaster LA1 4YQ, United Kingdom, and Swiss Federal Institute of Technology, ETH Zu ¨ rich, CH-8093 Zu ¨ rich, Switzerland

Received February 3, 2009. Accepted February 13, 2009.

In 1974, Junge derived an empirical relationship between the variability of concentrations of volatile trace gases in air at remote locations and their atmospheric residence time. Here, the Junge relationship is adapted to incorporate the deposition and revolatilization of semivolatile chemicals and applied to interpret nearly a decade of data on polychlorinated biphenyl (PCB) concentrations in air. A multimedia fate model, which accounts for deposition and revolatilization, is used to estimate the characteristic travel distance (CTD) for PCBs, where CTD serves as a measure of the effective atmospheric lifetime for semivolatile organic chemicals. Data are taken from sites of the Integrated Atmospheric Deposition Network in the North American Great Lakes and the Alert monitoring station in the Arctic, which is operated by the Canadian Northern Contaminants Program. Five factors that may introduce variability into measured concentrations are defined. By suppressing the effect of three of these factors in the data analysis, we identified variability consistent with the Junge relationship in many of the annual data sets (62%), with the relationship showing statistical significance (p < 0.05) in 23% of these annual data sets. The more remote monitoring sites from the Great Lakes region display the highest number of statistically significant Junge-type relationships between the variability in concentrations in air and estimated long-range transport potential in air. At sites in close proximity to areas of high population density, variability in PCB concentrations in air displays patterns that are consistent with primary or secondary temperature-driven volatilization sources. Analysis of variability in long-term monitoring data, using the techniques developed and illustrated here, provides useful insights into the factors that control the behavior of persistent semivolatile chemicals in the environment.

* Corresponding author phone: +44-1524-594330; e-mail: [email protected]. † Lancaster University. ‡ ETH Zu ¨ rich. 2746

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009

Introduction Trace gases with sources concentrated in certain areas and very long atmospheric lifetimes become homogenously distributed in the earth’s atmosphere, and concentrations measured at locations remote from source areas show very little variability over time. Gases with a similar source distribution but very short atmospheric lifetimes will not travel as far as their longer-lived counterparts and will only arrive at locations remote from source areas as a result of specific, transient wind patterns. Therefore, the variability in atmospheric concentrations of gases with shorter atmospheric residence times, measured at remote locations, will be greater than that for gases with longer atmospheric residence times. In 1974, Junge (1) derived an empirical relationship between the variability in the concentration of a trace gas in the atmosphere at a remote location and the atmospheric lifetime (ta, years) of the gas RSD ) 0.14/ta

(1)

where RSD is the relative standard deviation of the mean concentration. This approach has been subsequently applied, with varying degrees of success, to interpret monitoring data and results from models for a range of trace volatile organic compounds, notably hydrocarbons and halogenated hydrocarbons (2-4). In general, the Junge relationship is most useful for interpreting data collected from remote regions over a period of a year or longer, where atmospheric sinks for the chemical are uniformly distributed, and sources are concentrated at one or a limited number of locations (1). Data collected in the direct vicinity of known sources cannot be interpreted in this way because they do not represent the chemical’s potential for transport in the atmosphere according to its lifetime. Following the original work by Junge, many researchers have reported relationships between variability in the concentration of volatile gases and atmospheric lifetime. These usually have the general form RSD ) a/tab

(2)

where a and b take on a variety of values in different empirical studies, depending on the chemical investigated. The Junge relationship has also been applied to interpret measurements of semivolatile organic compounds (SVOCs) in the atmosphere. For example, Panshin and Hites (5, 6), using atmospheric concentrations of polychlorinated biphenyls (PCBs) measured in Bermuda, found an inverse relationship between ta and the increasing degree of chlorination for PCBs, where ta was derived from eq 2, using the original parameters from the Junge equation (a ) 0.14 year-b, b ) 1). Axelman and Gustafsson (7) applied the Junge theory to examine the influence of sinks, particularly hydroxyl radical degradation, on different PCB homologues in different geographical regions. More recently, Stroebe et al. (4) argued that the original Junge relationship cannot be generally applied to SVOCs because the fitting coefficients (a and b) are not universal but depend on the type of chemical investigated and the distance of the monitoring location from sources. They used a multimedia environmental fate model to demonstrate that SVOCs will deviate from the Junge relationship for volatile gases because they can undergo cycles of deposition and revolatilization and because they can be transported in ocean water as well as air. Stroebe et al. argue that the appropriate measure of an atmospheric lifetime for SVOCs in a Jungestyle relationship is the “effective” residence time in air that 10.1021/es900336y CCC: $40.75

 2009 American Chemical Society

Published on Web 03/13/2009

considers the effects of degradation and net deposition in combination. There are many possible causes of variability in atmospheric concentrations of SVOCs such as PCBs observed at a given monitoring site. Here, five factors that could determine the variability in concentrations observed at longterm monitoring stations are distinguished: (i) Sampling and analysissthe variability introduced in sampling and analysis procedures due to the determination of concentrations below the detection limit of analytical techniques, i.e., the inclusion of concentrations below detection limits can artificially increase variability due to the uncertainty associated with these values; (ii) Active point sourcessvariability in ongoing active point sources may directly influence the monitoring station if it is not sufficiently remote from these sources. Here we use the term “active” to denote sources that release pollutants as a result of a specific activity. For example, emissions of pesticides vary throughout the year due to farming practices; (iii) Temperaturesthe most readily observed effect of temperature on SVOCs is the variability in concentrations within a year (intra-annual). This is referred to here as “seasonality”. In general, seasonality is used to cover a wide range of factors; however, here it is used to describe the temperature-driven dynamic repartitioning between the atmosphere and surface media undergone by SVOCs, which affects the variability in concentration and composition of chemicals in the atmosphere (8-10); (iv) Yearto-year (interannual) changessthe variability that can be seen over longer periods of time due to changes in source strength and/or type, and changes in the equilibrium status of different environmental reservoirs (11). For example, concentrations of PCBs in the atmosphere have been declining year after year at many monitoring stations due to the implementation of controls on emission sources (11); and, finally, (v) Random, noncyclical eventssthe variability which is governed by loss mechanisms from the atmosphere such as hydroxyl radical concentrations and precipitation rates/types and the incursion of polluted air masses brought about by changes in wind direction and speed. In other words, the random, noncyclical events affecting the composition and concentration of semivolatile chemicals seen in the atmosphere, which may go through many phases of deposition and revolatilization before reaching their ultimate fate in the environment. The last factor (factor v) describes the variability that results in a Junge-type relationship (eq 2) and can be related to the effective atmospheric lifetime of semivolatile chemicals, which takes into account deposition and revolatilization processes. In this study, three of these five factors were suppressed by (a) only considering the annual data set for a given PCB congener, where concentrations in air were above the method detection limits for more than 10 months of the year, to remove the uncertainty surrounding measurements falling below these limits (factor i); (b) selecting a group of chemicals (PCBs) which have been banned since the 1970s and have no known “active” sources to the atmosphere (factor ii); and (c) utilizing data collected in a single year to reduce influences from interannual variability caused by declining long-term trends in air concentrations (factor iv). This isolates the two remaining factors of seasonal variability driven by temperature (factor iii) and the random variability such as travel times and atmospheric sinks that are expected to produce a Junge-type variability-lifetime relationship (factor v). The objective is to use field measurements from monitoring programs to test whether the observed variability of concentrations of PCBs in air shows a Junge-type relationship. For this purpose, monitoring data from the Integrated Atmospheric Deposition Network (IADN) and the Canadian Arctic Northern Contaminants Program (NCP) are used. Effective atmospheric lifetimes for PCBs are estimated using

the characteristic travel distance (CTD) calculated by a multimedia fate and transport model, the OECD overall persistence (POV), and long-range transport potential (LRTP) screening tool (Tool V2.0) (12). This results in a travel distance (km), which can be used to describe the effective atmospheric lifetime of semivolatile chemicals incorporating the deposition and revolatilization processes. This work is motivated by the need to understand the impacts of persistent organic pollutants (POPs) and identify new chemicals that have a high potential for persistence and long-range atmospheric transport in the context of international agreements such as the Stockholm Convention (13) and the Convention on Long-Range Transboundary Air Pollution (14).

Theory Variability in atmospheric PCB concentration can be interpreted in two hypothetical scenarios. In scenario I, a monitoring station in close proximity to a source is considered and temperature-driven seasonal variability in primary and/ or secondary sources dominates variability in observed concentrations (factor iii). In scenario II, the monitoring station is considered to be remote from sources and random variability in atmospheric sinks and travel time from sources to monitoring stations dominates variability in observed concentrations (factor v). Scenario I. At a monitoring site that is in close proximity to a volatilization source, the dominant influence on the variability of PCB concentration in the air is expected to be seasonal changes in temperature. PCB concentrations measured over the course of a year typically display regular variability related to seasonal temperature fluctuations, marked by higher concentrations in summer and lower concentrations in winter (9). This seasonality is most evident for the heavier congeners due to their higher enthalpies of vaporization (8). Therefore, temperature-driven volatilization of PCBs from primary or secondary diffuse sources is expected to result in a positive relationship between the RSD of concentrations in air and the degree of chlorination. At monitoring stations situated within or near urban environments, volatilization from primary sources such as PCB storage sites and building materials will control variability within an annual data set. This may be the case at several sites in the Integrated Atmospheric Deposition Network (IADN), where PCB concentrations are correlated with proximity to human populations and, hence, urban sources (11). More remote sites may also be under the influence of temperature-driven secondary volatilization sources due to the alteration of water bodies, such as the Great Lakes, from historic sinks to present sources (15). Scenario II. At truly remote locations, where there are no sources (either primary or secondary), variability in PCB concentration is not determined by local conditions such as seasonal temperature fluctuations but rather random, noncyclical influences such as wind speed and direction, in combination with the atmospheric lifetime of the different PCB homologues. This leads to an inverse relationship between log RSD and log ta as implied by the Junge relationship (eq 2). As stated in the introduction, for semivolatile chemicals, effective residence time in air, considering deposition and revolatilization (ta,eff), can be estimated using a multimedia model and is often reported as a characteristic travel distance (CTD), where CTD is ta,eff multiplied by an assumed wind speed (v, m s-1). Thus, for semivolatile chemicals, it is useful to restate the Junge relationship as RSD ) A/CTDb

(3)

where A ) av . By taking the logarithm of both sides, eq 3 can be rearranged in a linear form b

VOL. 43, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2747

log RSD ) log A-b log CTD

(4)

Thus, a plot of log RSD versus log CTD for a range of PCB congeners will yield a straight line with slope b and intercept log A if (i) CTD calculated by the selected multimedia model is a good estimate of the effective atmospheric residence time of PCBs and (ii) a Junge-type relationship between variability in air concentrations and effective atmospheric lifetime exists for a given set of observations. Figure 1 illustrates the trends in log RSD that are expected under scenarios I (Figure 1A) and II (Figure 1B) across the range of PCB homologues with 1-10 chlorines. Figure 1A is based on the variability in strength of a volatilization source measured at one month intervals over the course of a year in which average temperature varies more than 25 °C from winter to summer. This was estimated from variability in vapor pressure as a function of temperature using enthalpies of vaporization derived from the relationship recommended by MacLeod et al. (8) and assuming the source strength is directly proportional to vapor pressure. A variety of multimedia environmental fate models have been developed that are capable of estimating the CTD of chemicals (16). Here, we use Tool V2.0, which is described in detail by Wegmann et al. (12). The log RSD illustrated for scenario II (Figure 1B) is derived from CTDs calculated using Tool V2.0 in combination with eq 4, assuming a ) 0.14 year-b and b ) 0.27, which is the value estimated by Stroebe et al. (4) for volatile chemicals with ta between 1 and 10 days. As the CTD does not vary significantly between congeners within homologue groups, average values for the 10 homologue groups are illustrated in order to improve the clarity of the plots. The annual field data can, therefore, be tested for two types of relationships: (1) positive relationships between log RSD and the chlorine number, indicating scenario I behavior, and (2) negative relationships between log RSD and log CTD, indicating scenario II behavior.

Methods There are several monitoring stations that have accrued databases of atmospheric PCB concentrations for periods of up to 10 years or greater. Because of their availability and high temporal resolution, data from the monitoring sites of the North American Great Lakes Integrated Atmospheric Deposition Network (IADN) and Canadian Arctic Northern Contaminants Program (NCP) were selected for this study. Because years are analyzed separately, this provides 52 tests to see if the variability of PCB concentrations in air conforms to a Junge-type relationship. Sample Sites and Data Use. The locations of the IADN and Arctic sample sites, together with site descriptions, are presented in Table 1, along with population densities within a radius of 45 km, to provide an indication of remoteness. All of these sites have ongoing sampler operations, but the data examined here cover 7-10 years from 1991-2000 inclusive (exact years at each site are provided in Table 1). Air sampling methodology, chemical analysis, and quality controls can be obtained from previous publications (refs 9 and 17-19). At Alert, weekly samples are collected, aspirating ∼13 000 m3 of air per sample. The American and Canadian IADN sites take 24 h samples, aspirating air volumes of approximately 800 and 350 m3, respectively. These samples are taken on different days each week, initially every six days and then, after 1993, with the introduction of more sites, every twelve days. This has resulted in ∼52 samples per year at Alert and ∼26 samples per year at the Great Lakes Sites, although sampler “downtime” (from mechanical malfunction, etc.) often results in fewer samples. The annual data sets for 1992 and 2000 at Alert were not included in the analysis because of contamination issues in 1992 and missing values due to sampler downtime in 2000. 2748

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 8, 2009

FIGURE 1. Theoretical plots of illustrative relationships between the logarithm of relative standard deviation of concentrations of PCB homologues in air (log RSD) and (A) the chlorine number and (B) the logarithm of characteristic travel distance (log CTD). Two bounding scenarios are illustrated. Scenario I is the variability in concentration in air as determined by temperature-driven variability in a local volatilization source. Scenario II is the variability in concentration in air as described by the Junge relationship assuming a ) 0.14 yearb and b ) 0.27. Concentrations of individual PCB congeners (pg m-3), with coeluting congeners removed, were exported into Excel spreadsheets for each site, with data grouped by month. Samples falling below the method detection limits (MDLs), which were site and congener specific, were removed prior to data analysis. For the U.S. IADN sites, MDLs for each congener were calculated from the means (x) and standard deviations (SD) derived from annual averages of monthly field blanks, where MDL ) x + 3SD. Blank data were provided with the sample data for each site, except the Canadian sites where MDLs were provided. In the case of Burnt Island, field blanks have been reported to be ∼40% of the average PCB concentration (11), resulting in high MDLs for this site. A summary of the MDLs for each PCB homologue group can be found in Table S1 of the Supporting Information together with the total number of congener years for each homologue (i.e., number of congeners within a homologue multiplied by the number of sample years) and the number of congener years excluded because of missing data points or data falling below detection limits. In both cases, if this resulted in