The Magnitude and Spatial Range of Current-Use ... - ACS Publications

Dec 16, 2013 - In this study we quantified and assessed the magnitude and spatial extent of ... We then discuss model estimates of the fate and urban ...
0 downloads 3 Views 3MB Size
Article pubs.acs.org/est

The Magnitude and Spatial Range of Current-Use Urban PCB and PBDE Emissions Estimated Using a Coupled Multimedia and Air Transport Model Susan A. Csiszar,† Miriam L. Diamond,*,†,‡ and Sreerama M. Daggupaty§ †

Department of Chemical Engineering and Applied Chemistry, University of Toronto, 200 College Street, Toronto, Ontario, Canada M5S 3E5 ‡ Department of Earth Science, University of Toronto, 22 Russell Street, Toronto, Ontario, Canada M5S 3B1 § Air Quality Research Division, Environment Canada, 4905 Dufferin Street, Toronto, Ontario, Canada M3H 5T4 S Supporting Information *

ABSTRACT: SO-MUM, a coupled atmospheric transport and multimedia urban model, was used to estimate spatially resolved (5 × 5 km2) air emissions and chemical fate based on measured air concentrations and chemical mass inventories within Toronto, Canada. Approximately 95% and 70% of Σ5PCBs (CB-28, -52, -101, -153, and -180) and Σ5PBDEs (BDE-28, -47, -100, -154, and -183) emissions of 17 (2−36) and 18 (3−42) kg y−1, respectively, undergo atmospheric transport from the city, which is partly over Lake Ontario. The urban air plume was found to reach about 50 km for PCBs and PBDEs, in the direction of prevailing winds which is almost twice the distance of the wind-independent plume. The distance traveled by the plume is a function of prevailing wind velocity, the geographic distribution of the chemical inventory, and gasparticle partitioning. Soil wash-off of historically accumulated Σ5PCBs to surface water contributed ∼0.4 kg y−1 (of mainly higher congeners) to near-shore Lake Ontario compared with volatilization of ∼6 kg y−1 of mainly lighter congeners. Atmospheric emissions from primary sources followed by deposition to surface films and subsequent wash-off to surface water contributed ∼1 kg y−1 and was the main route of Σ5PBDE loadings to near-shore Lake Ontario which acts as a net PBDE sink. Secondary emissions of PCBs and PBDEs from at least a ∼900 000 km2 rural land area would be needed to produce the equivalent primary emissions as Toronto (∼640 km2). These results provide clear support for reducing inventories of these POPs. Ontario as a case study.1,14−17 In a companion paper, Csiszar et al.,16 developed and applied SO-MUM (Spatially Oriented Multimedia Urban Model) to estimate spatially resolved emissions (on a 5 km scale) of SVOCs from an urban area. SO-MUM was built from the Boundary Layer Forecast Model and Air Pollution Prediction System, BLFMAPS and the Multimedia Urban Model, MUM. These emissions can be viewed as aggregate emission factors for all activities in an urban area contributing to a chemical’s release. When expressed per unit of chemical inventory, these emission factors can guide our expectations for changes in primary emissions as a function of changes in a city’s chemical inventory. In this study we quantified and assessed the magnitude and spatial extent of primary and secondary urban emissions and the resulting plume to the surrounding region via air transport and surface waters. We define primary emissions as “fresh” emissions coming directly from their original sources (i.e., in-

1.0. INTRODUCTION Urban areas are generally understood to act as major sources to surrounding regions of persistent organic pollutants (POPs) such as PCBs and PBDEs, and other semivolatile organic compounds (SVOCs) such as PAHs. These compounds originate from, for example, building materials, consumer products, and vehicle exhaust, respectively.1−4 Hence, chemical use within and release from urban areas leads to an “urban plume” that atmospherically transports chemicals to surrounding regions.5,6 This phenomenon has been well documented in the Great Lakes region.7−10 The extent of the urban plume or a chemical’s spatial range, has been generally quantified based on a point source or a single urban measurement,5,9 however, intraurban measurement campaigns have shown that SVOC concentrations and, presumably emissions, can vary significantly within a city.1,11,12 A diffuse geographic distribution of emissions would logically extend the distance of the urban plume and as cities grow in spatial extent13 they will no longer act as geographic point source emitters of certain SVOCs. This paper is one of a series reporting on a multiagency program to estimate emissions from an urban area and loadings to an adjacent water body, using Toronto, Canada and Lake © 2013 American Chemical Society

Received: Revised: Accepted: Published: 1075

July 11, 2013 October 14, 2013 December 16, 2013 December 16, 2013 dx.doi.org/10.1021/es403080t | Environ. Sci. Technol. 2014, 48, 1075−1083

Environmental Science & Technology

Article

2.0. MATERIALS AND METHODS 2.1. Model Description. The unsteady-state version of MUM and SO-MUM model framework are described in detail by Csiszar et al.16,32 Briefly, the model’s geographic domain of 12350 km2 was divided into 494, 5 × 5 km2 cells where interactions between air and other media as well as multimedia transport were modeled in MUM. Air transport (in 10 vertical layers) between cells was modeled using BLFMAPS.33 The fugacity-based MUM divides the urban environment into six compartments (air, water, soil, sediment, vegetation, and surface film) with the following mass balance (MB) equations

use or in-storage PCBs or PBDE-containing products) and secondary emissions as re-emission of previously emitted and then deposited chemical (e.g., chemical volatilization from water or wash-off of soil). Analogously to the long-range transport metrics developed by Scheringer,18 Bennett et al.,19 and Beyer et al.,20 such as the Characteristic Travel Distance or CTD, we used SO-MUM to quantify the potential of a chemical to travel from an urban reference location to reach a concentration of 1/e of that reference location. We also calculated the distance that concentrations dropped to 1/10 the reference location representing background concentrations.1 We next used SO-MUM with estimates of the inventories of PCBs and PBDEs (penta- and octa-BDEs) in Toronto, Canada, to explore the implications of reducing these inventories on primary versus secondary emissions. Despite undergoing regulatory control in the 1970s, the major inventory of PCBs has only recently been slated for removal. PCBs have been in use since the 1930s and in 1977 Canada banned production (which never occurred in the country) and importation.21,22 PCB regulations that passed into law in 2008 and were amended in December 2011, specify that equipment may not be used past December 31, 2009 if it contains PCBs at concentrations greater than 500 mg kg−1 or between 50 and 500 mg kg−1 if the equipment is located in a “sensitive” location such as nearby a drinking water treatment facility, child care facility, school or hospital. Equipment containing PCBs between 50 and 500 mg kg−1 that is not in a sensitive location, and PCBs in light ballasts or pole top transformers can be used until 2025.23 As of 2006, an estimated 437 tonnes of PCBs were in use and storage in Toronto.21 PCB emissions could temporarily increase during equipment removal due to accidental spills, and other sources could include accidental fires, waste incineration, and biomass burning.24−28 However, Melymuk et al.29 found that 75% of variability in PCB air concentrations in Toronto could be explained by the geographic distribution of PCBs in in-use capacitors, transformers, building sealants, and in storage. The penta- and octa-BDE flame retardant mixtures were declared toxic in 2008 in Canada under the Canadian Environmental Protection Act after its draft risk assessment was released in 2006.30 North American manufactures signed a voluntarily agreement with the U.S. EPA to cease production of the penta- and octa- mixtures in December 2004. Production (which never occurred in Canada) and importation of the mixtures (but not in imported products) were banned in Canada in 2008.30,31 Canadian regulations do not appear to pertain to penta- and octa-BDE contained in finished imported goods. Csiszar et al.16 estimated an inventory of ∼200 (90− 1000) tonnes of penta- and octa-BDEs in computers, printers, televisions, furniture, and cars in Toronto between ∼2005− 2008. Here we briefly present SO-MUM and its application to PCBs and penta- and octa-BDEs using Toronto, Canada as a case study. We then discuss model estimates of the fate and urban transport pathways of these compounds which leads to estimates of the magnitude of emissions leaving the city and, since Toronto is on the shores of Lake Ontario, loadings to Lake Ontario from the city. Finally, we use the model to investigate the magnitude of primary emissions from the city’s chemical inventories versus secondary emissions as the inventories decrease.

d(ZiVfi i ) dmi = = Ei + dt dt for i = a , r , s , se , v , f

∑ (Djif j ) − DTifi i≠j

(1)

where mi, Vi, Zi, f i, Ei, and DTi are the mass (mol), volume (m3), bulk Z-value (mol m−3 Pa−1), fugacity (Pa), specified emission source or sink (mol h−1), and total loss D-value (mol h−1 Pa−1) (transport and transformation), respectively, for compartment i; Dji is the intercompartmental transfer D-value from compartment j to i. The unsteady-state equations are solved using the implicit Euler approximation with a 5 min time-step corresponding to the time-step used in BLFMAPS. BLFMAPS, given a chemical emission E (mol s−1), solves for the air concentration, Ca (mol m−3), using the following air chemical transport equation ∂Ca ∂Ca d ⎛ ∂Ca ⎞ = −v ⃗∇Ca − w + k h∇2 Ca + ⎟+E ⎜k z * ∂z dz ⎝ ∂z ⎠ ∂t * * *

(2)

where t is time; w* is the vertical velocity in terrain following coordinates; z* = z − h(x,y) (where z is height and h(x,y) is surface elevation); and kh and kz are the horizontal and vertical eddy diffusivities (m2 s−1), respectively.33,34 Toronto is located on the north shore of Lake Ontario and has five tributaries running from north to south where they discharge into the lake. The model was used to estimate air− water exchange fluxes in cells that included Lake Ontario, by reducing eq 1 to include only the air and water compartments. In cells that border Lake Ontario, fugacities were solved using two sets of MB equations: those of eq 1 for the land area (excluding tributaries) and the reduced air−water mass balance for the lake area. Finally, for the cells that had tributary discharges into Lake Ontario we used two sets of mass balance equations: those of eq 1 based on the land, tributary, and lake areas within the cell and the two-compartment air−water mass balance over Lake Ontario. 2.2. Model Parameterization. Toronto is Canada’s largest city with 2.5 million people covering ∼640 km2. Bulk air emissions were estimated using the method described by Csiszar et al.16 Briefly, chemical concentrations in air measured across Toronto by Melymuk et al.1 by means of passive air samplers (PAS) were used as inputs to a steady-state version of MUM in order to back-calculate emissions necessary to support those concentrations. These emissions were converted into an emission rate per unit inventory in each of the nine cells having measured air concentrations out of 39 cells constituting Toronto. These nine emission rates per unit inventory were then averaged and that average was multiplied by the mass inventory in each of the 39 cells in Toronto to produce cellspecific aggregate emission rates (e.g., gchemical y−1) for all cells 1076

dx.doi.org/10.1021/es403080t | Environ. Sci. Technol. 2014, 48, 1075−1083

Environmental Science & Technology

Article

with geographic “point source” emissions. The model was also run under various inventory reduction scenarios in order to predict the results of reducing these inventories on primary versus secondary emissions.

in the city. We set the emission rate to zero outside of Toronto in order to assess the effects of Toronto only on surrounding regions. The effect of assuming zero emissions outside of Toronto was tested for PCBs by running SO-MUM with an emission rate necessary to support measured rural air concentrations in all cells outside of the city. The results of this model run with rural emissions did not differ significantly from the run in which rural emissions were neglected.16 We used the PCB mass inventory of Diamond et al.21 and Robson et al.35 and the PBDE mass inventory of Csiszar et al.16 to scale emissions. The model was initialized as follows. Initial air concentrations were set to zero. In order to initialize the concentrations in other media (tributaries, soil, sediment, vegetation, and film) we ran the steady-state version of MUM using average environmental properties of the simulation period, estimated emissions, as well as inflow air concentrations calculated in a manner similar to the emissions. The initial soil concentrations were adjusted so that they were close to measured surface soil concentrations in Toronto36,37 as described by Csiszar et al.16 The model was given a run-up time of 2 weeks before recording data. Physical-chemical properties used in the model for PCBs and PBDEs are listed in Csiszar et al.16,32 with partition coefficients from Schenker et al.38,39 Meteorological data were provided by BLFMAPS (which was initialized using data from the Global Environmental Multiscale Model, GEM), as well as surface water and soil temperatures. Values of OH radical concentrations, solar irradiance, and total suspended particulate concentrations were described by Csiszar et al.16 Based on the results of a sensitivity analysis,16 we used midpoint values of dry deposition velocities between 0.2 cm s−1 and up to 7.8 and 9.7 cm s−1 for PCBs and PBDEs, respectively, representing values that have been used by others40−44 (Supporting Information (SI) Table S1). Land use was extrapolated from land-use base maps45 using GIS. Tributary flow rates and depths, and inflow water concentrations into Toronto were described by Csiszar et al. 16 Air−water mass transfer coefficients were updated at each time step as they depended on temperature (in both water and air) and wind speed which were provided by the BLFMAPS part of the model with their calculation described in the SI and Csiszar et al.16 Lake Ontario surface water concentrations of PCBs and PBDEs at various locations (five for PCBs and two for PBDEs) in Lake Ontario nearby Toronto were obtained from Alice Dove (personal communication) 46 and Ueno et al., 47 respectively (SI Table S2, Figures S2 and S3). These values were used to interpolate concentrations using an exponential function to 3700 km2 of Lake Ontario extending out from the downtown cell, DT (5,2) (SI Figure S1 and text in SI). Due to the limited availability of water concentration data and since modeling Lake Ontario water concentrations was beyond the study’s scope, water concentrations were kept constant over time. 2.3. Model Runs. The model was run for spring (91 days) in 2008 corresponding to the spring air measurement campaign of Melymuk et al.1 The following congeners were modeled individually: five PCB congeners CB-28, -52, -101, -153, and -180 and five PBDE congeners BDE-28, -47, -100, -154, and -183. The model was run with emissions to all Toronto cells and from the individual cell with the highest emission only (DT for PCBs and (5,3) for PBDEs, SI Figure S1) in order to compare spatially resolved urban emissions from the entire city

3.0. RESULTS AND DISCUSSION Emissions of Σ5PCBs and Σ5PBDEs from Toronto of 17 (2− 36) kg y−1 and 18 (3−42) kg y−1 were estimated using SOMUM. The emission ranges were based on the standard deviations of emission rates per mass inventory calculated in the nine cells for which we had measured air concentrations.16 The estimates were based on spring concentrations and likely overestimate yearly emissions as measured air concentrations are lower in the fall and winter.1,16 These emission estimates, when expressed on an areal basis, were similar to those estimated using a one-box version of MUM and those of others using other methods.21,42,48−50 The model estimated that 95% of Σ5PCBs emitted to air were advected out of the city with the remainder deposited to vegetation, soil, and film within the city (Figure 1a). In contrast,

Figure 1. Mass balance diagrams for the sum of 39, 5 × 5 km2 cells spanning the city of Toronto for (a) Σ5PCBs and (b) Σ5PBDEs. Units are in g d−1; solid lines refer to transport, dotted lines refer to transformation, and the curved arrow refers to emissions. Rates represent averages over the 91 day simulation period of spring 2008. Values in brackets in each compartment represent the change in mass (g d−1) over the simulation period.

∼72% of emissions to air of Σ5PBDEs were advected beyond the city (Figure 1b). The percentages of Σ5PCBs and Σ5PBDEs in the gas phase were 84 and 27%, respectively. The urban plumes of both chemicals extended over the lake to the east of the city for this spring simulation since the prevailing winds were from the southwest (Figure 2). It is important to evaluate the reliability of these and the following estimates. Csiszar et al.16 evaluated SO-MUM’s performance and discussed model sensitivities and uncertainties. Briefly, the model reproduced the geographic pattern of concentrations and estimated concentrations of Σ5PCB and 1077

dx.doi.org/10.1021/es403080t | Environ. Sci. Technol. 2014, 48, 1075−1083

Environmental Science & Technology

Article

Figure 2. Modeled air concentrations (up to a 100 m height) in the 5 × 5 km2 cells that span the model domain of 12350 km2, averaged over the spring 2008 simulation period (91 days) for a) Σ5PCBs and b) Σ5PBDEs. The domain is located at the northwestern end of Lake Ontario in easterncentral Canada and includes Toronto. See SI Figure S1 for additional details.

Σ5PBDE to within an order-of-magnitude of time integrated measured concentrations obtained with passive air samplers. The model also reproduced temporal variability in concentrations measured with a high-volume air sampler. Model estimated emission rates were most sensitive to horizontal wind speed which, fortunately, is well-known, and varied by up to an order of magnitude. Modeled deposition rates of PBDEs varied by up to an order of magnitude based on sensitivity to particle dry deposition velocities for which values have been found to range from 0.2 cm s−1 to nearly 10 cm s−1.40,43 Csiszar et al.32 investigated the sensitivity of film wash-off to impervious surface area and antecedent dry time, and found that wash-off masses were directly proportional (1:1) to impervious surface area and varied within a factor of 2 between dry times of 5 and 20 days. Taking all sources of uncertainty into account, model estimates are anticipated to be reliable to within an order-ofmagnitude. 3.1. Urban Travel Distances. The Characteristic Travel Distance, CTD, of a chemical is defined as the distance at which its air concentration is reduced to 1/e (i.e., 63%) from a point source.19,20 Here we investigated the analogous transport distance of the urban PCB and PBDE plume, which we defined as the distance at which the air concentration is reduced to 1/e or 1/10 (representing a background level) of the concentration at a reference urban location (i.e., location of highest emission). This distance differs from the CTD as it takes into account the spatial heterogeneity arising from spatially distributed urban emissions rather than treating the emissions from an urban area as a single geographic point.

We estimated the extent of the urban plume in two ways. First, we identified the distances from the reference cell at which modeled concentrations (up to a 100 m height) were below 1/e or 1/10 of that reference cell; this method accounted for prevailing wind velocities. Second, we fitted the modeled concentrations in all cells (regardless of direction) to an exponential as a function of distance from the reference cell, which is analogous to fitting a radial dilution model.51 This latter calculation yields an average travel distance as it does not account for wind velocities. Additionally, in order to assess the differences between the travel distances from spatially heterogeneous urban emissions, as described above, versus the travel distances from a geographical point source, we ran SO-MUM with emissions emanating from the reference cell only (DT for PCBs and cell (5,3) for PBDEs). The 1/e and 1/10 travel distances were ∼25 and ∼60 km for Σ5PCBs and ∼30 and ∼50 km for Σ5PBDEs, respectively, when accounting for the prevailing wind velocity. Given the uncertainties in the model as discussed above, the distances for these two compounds are virtually identical. In contrast, using concentrations from all cells and an exponential fit to these data, the 1/e and 1/10 travel distances were ∼15 and ∼34 km for Σ5PCBs and ∼13 and ∼31 km for Σ5PBDEs (SI Table S3), or nearly half the wind-dependent travel distances (Figure 3). SI Table S4 summarizes the 1/e and 1/10 travel distances of PCBs and PBDEs for dry particle deposition, wet deposition, and film concentrations which were similar to, but less than those for the plume in air. In comparison, Green et al.5 estimated, by means of an interpolation model supported by overwater air measurements 1078

dx.doi.org/10.1021/es403080t | Environ. Sci. Technol. 2014, 48, 1075−1083

Environmental Science & Technology

Article

Figure 3. Modeled air concentrations averaged over the 91 day spring simulation period for (a) Σ5PCBs and (b) Σ5PBDEs where 0 km is the cell with the highest concentration. Indicated in the figure are cell concentrations considering primary emissions from all Toronto cells (⧫), the exponential fit to modeled air concentrations in cells along the predominant wind direction (−, royal blue), the exponential fit for all cells which is referred to as “wind independent” (−), and the exponential fit for the SO-MUM model run that considered a single point source emission from the urban reference cell (DT for PCBs and (5,3) for PBDEs) (−, red). (●) and (●, green) indicate the points at which the exponential fit drops to 1/e and 1/10 the reference concentration, respectively.

explanation by running the model with emissions from downtown only. The results showed shorter 1/e and 1/10 distances for PBDEs than those using emissions from all Toronto cells due to greater deposition losses of these more particle-sorbed chemicals (Table S3, Figure 3) and vice versa for PCBs. Thus, the extent of the urban air plume is affected by chemical mobility due to the prevailing wind velocity, the geographic distribution of the inventory, and gas-particle partitioning. 3.2. Urban Chemical Pathways. Of the 5% of Σ5PCB emissions not advected from the city, 30, 52, 16 and 1% were transferred to soil, vegetation, films, and tributaries, respectively. For Σ5PBDEs, the comparable percentages were 40, 27, 32, and 0.6%, respectively, for the 28% not advected from the city. These values refer to net transfer (deposition minus volatilization) from air (Figure 1). In more detail, the model estimated that ∼0.6 g d−1 Σ5PCBs was deposited to soil with only ∼0.1 g d−1 returning via volatilization. The comparable values for Σ5PBDEs were ∼5 g d−1 and