SEARCH

SEARCH BY CITATION

Keywords:

  • atmospheric chemistry transport model;
  • persistent organic pollutants

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] The POP version of the Danish Eulerian Hemispheric Model (DEHM-POP) is a further development of a 3-D dynamic atmospheric chemistry transport model covering the Northern Hemisphere, which was originally developed to study atmospheric transport of conventional air pollutants and other atmospheric constituents (e.g., SOX, heavy metals, and CO2). Four different surface compartments (soil, ocean water, vegetation, and snow) are introduced in DEHM-POP with each compartment including the most dominant dynamic processes determining the exchange between air and the surface type to account for the consecutive cycles of deposition and reemission of persistent organic pollutants (POPs). This model setup makes it possible to study short-term atmospheric variability of POPs, which is exemplified in this paper by a study of the atmospheric variability of α-hexachlorocyclohexane (α-HCH), the major component of the worldwide most used insecticide: technical HCH. Simulated α-HCH air concentrations are evaluated against measurements from 21 monitoring stations within the model domain, and the model is able to predict the annual average concentration as well as the long-term trend for the 1990s. Significant correlations between simulated and measured short-term atmospheric concentrations of α-HCH are also found at the majority of the investigated monitoring stations, which shows that it is possible to resolve the atmospheric variability of POPs using an atmospheric chemistry transport model. Differences between simulated and measured atmospheric α-HCH variability can arise because the measurements may be influenced by local features that are not accounted for in the model with the relatively coarse horizontal resolution and surface description.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] Persistent organic pollutants (POPs) are persistent, bioaccumulating compounds with a potential for long-range transport and with harmful effects on human health and the environment. The physical-chemical properties of POPs result in partitioning to different media and in combination with the long environmental life times, continuous repartitioning of POPs between media occurs through intermedia exchange processes. These exchange processes are highly dynamic and depend on environmental conditions such as temperature and humidity and also on the concentration of POPs and the capacity of a medium to retain or store POPs.

[3] The environmental exchange processes and the distribution of POPs between different environmental media have been studied with experimental tools as well as in model studies. The spatially resolved distribution of POPs has been the subject of model studies with different types of models such as multicompartment mass balance models also called “box models” [e.g., Strand and Hov, 1996; Wania et al., 1999; Scheringer et al., 2000; Prevedouros et al., 2004; MacLeod et al., 2005] and with atmospheric chemistry transport models [e.g., Koziol and Pudykiewicz, 2001; Semeena and Lammel, 2003; Malanichev et al., 2004; Hansen et al., 2004].

[4] Most of the model studies have primarily focused on predicting annual averages and longer time trends of environmental POP concentrations [e.g., Wania and Mackay, 1999; Malanichev et al., 2004; MacLeod et al., 2005], whereas a few studies also have studied shorter-term air concentrations [Koziol and Pudykiewicz, 2001; Breivik and Wania, 2002] but without evaluating the simulated air concentrations against measurements as it is traditionally done for conventional air pollutants, e.g., by calculating correlation coefficients.

[5] This study presents a high-resolution 3-D dynamic atmospheric chemistry-transport model (CTM) which covers the Northern Hemisphere and was developed to study the environmental fate of POPs. The model was previously presented in a version including only two surface compartments (soil and ocean water) [Hansen et al., 2004], whereas two additional surface compartments (snow and vegetation) are included in this study. The addition of more surface compartments improves the agreement between simulated and observed α-HCH air concentrations (K. M. Hansen et al., The role of the snowpack on the fate of α-HCH in an atmospheric chemistry-transport model, submitted to Environmental Science and Technology, 2007, hereinafter referred to as Hansen et al., submitted manuscript, 2007). The applied model setup makes it possible to study the short-term atmospheric variability of POPs with a higher spatial and temporal resolution than previous POP-model studies. In this paper the model structure is shortly described and the model is tested for a single POP compound and evaluated against available air measurements within the model domain. The aim of the study is to investigate the possibility of resolving the short-term atmospheric variability of POPs by applying the high spatial and temporal resolution of an atmospheric CTM.

2. DEHM-POP Model

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information

[6] DEHM-POP is a further development of the Danish Eulerian Hemispheric Model (DEHM) a versatile atmospheric CTM. Different versions were previously applied to study atmospheric transport of sulphur and sulphate (DEHM-SOX) [Christensen, 1997], CO2 (DEHM-CO2) [Geels et al., 2004], heavy metals, e.g., mercury (DEHM-Hg) [Christensen et al., 2004; Skov et al., 2004], and a version including a chemical scheme with 63 atmospheric constituents and more than 100 chemical reactions (the DEHM-REGINA model) [Frohn et al., 2002]. DEHM is based on the full 3-D advection-diffusion equation for the modeled species and is driven by meteorological data from a numerical weather prediction model. The PSU/NCAR mesoscale model (MM5v2) [Grell et al., 1995] is applied in the DEHM-POP setup. The applied numerical schemes have all been carefully tested for the previous versions of DEHM [Christensen, 1997; Frohn et al., 2002].

[7] The DEHM-POP model domain covers the majority of the Northern Hemisphere and is in the horizontal defined on a regular 135 x 135 grid with a resolution of 150 km at 60°N. There are 20 unevenly distributed vertical layers defined on terrain following σ levels (σ = p/ps, where p, ps is the pressure at the layer and at the surface, respectively) extending up to a height of approximately 18 km. Several surface compartments have been added to the DEHM-POP model with descriptions of POP-related processes within the compartments and exchange processes between them to account for the multimedia partitioning behavior of POPs. A detailed description of all included processes can be found in [Hansen, 2006].

2.1. Air

[8] There is a continuous repartitioning between the gas and the particle phase dependent on the temperature and available particles and this process may be the most important for determining the atmospheric transport and fate of POPs [Bidleman, 1988]. POPs are lost from the atmosphere through dry deposition of particle phase compounds, through wet deposition and by chemical transformation. Dry air-surface gas exchange acts as both a source and a sink of atmospheric POP concentrations.

2.1.1. Gas-Particle Partitioning

[9] The fraction of POPs sorbed to particles, ϕ, is in DEHM-POP described using the expression:

  • equation image

where TSP (μg m−3) is the total suspended particle concentration in air and Kp is the particle-air partition coefficient calculated using the Koa approach [Harner and Bidleman, 1998]:

  • equation image

where Koa is the octanol-air partition coefficient and fOM is the mass fraction of organic matter in the particles set equal to 20% in DEHM-POP, which is in the upper range expected for urban aerosols [Harner and Bidleman, 1998]. To include a seasonal variation, the TSP concentration in each grid cell is presently estimated on the basis of the monthly averaged sulphate particle concentration simulated by the DEHM-SOX model [Christensen, 1997]. The sulphate particle concentration is multiplied by a factor of seven to reach TSP concentrations within a range of observed values.

2.1.2. Dry Particle Deposition

[10] The dry deposition of particles is described as in previous versions of the DEHM model and distinguishes between dry deposition to open water, to forest-covered land surface and to non-forest-covered land surface [Christensen, 1997].

2.1.3. Wet Deposition

[11] Wet deposition of POPs is calculated in DEHM-POP using a simple parametrization based on a scavenging coefficient formulation as in earlier versions of DEHM [Christensen, 1997]. The scavenging coefficient is different for the particle phase and the gas phase and for the latter the rain and snow scavenging is also treated differently.

[12] The particle phase scavenging coefficients used in the model are 7.0 × 105 and 1.0 × 105 for in-cloud and below-cloud scavenging respectively. For the gas phase scavenging by rain it is assumed that equilibrium is attained rapidly between the gas phase POPs and the dissolved phase in a raindrop, and the scavenging coefficient is estimated using the water to air partition coefficient, Kwa [e.g., Wania et al., 1998]. The scavenging of gas phase POPs by snow is described by sorption onto the surface of the snow crystals. The snow scavenging coefficient of gas phase POPs is given by [Lei and Wania, 2004]:

  • equation image

where Kia (m) is the interface air partition coefficient, SSAini (m2 kg−1) is the specific surface area of the snow flakes and ρw (kg m−3) is the density of water [Lei and Wania, 2004]. Initial specific surface area is chosen to be: SSAini = 120 m2 kg−1, which corresponds to observed values for freshly fallen snow in calm weather conditions [Legagneux et al., 2002].

2.1.4. Chemical Transformations in the Atmosphere

[13] POPs in air can be transformed by chemical processes, which are different for the gas phase and particle phase POPs. Reaction with OH radicals is generally the most important transformation process for gas phase POPs [Atkinson et al., 1999], and the atmospheric degradation in DEHM-POP is described with this process using monthly averaged OH concentrations extracted from the DEHM-REGINA model together with literature derived reaction rates. The degradation of particle phase POPs is assumed to be considerably slower than gas phase reactions, and in the DEHM-POP model it is set equal to zero like in most other model studies [e.g., Scheringer et al., 2000; Semeena and Lammel, 2005].

2.2. Surface Compartments and Air-Surface Exchange

[14] A description of air-surface exchange processes is important when modeling the environmental fate of POPs to account for the consecutive cycles of deposition and reemission. Four surface compartments are considered in DEHM-POP: soil, surface ocean water, vegetation and snow. The processes associated with these compartments and the dry air-surface gas exchange which is the key process is shortly described in this section.

2.2.1. Soil

[15] The soil module in DEHM-POP is based on the soil module from the zonally averaged multicompartment mass balance model developed by [Strand and Hov, 1996]. The land-covered surface in the model consists of a 0.15 m thick soil layer containing a mixture of soil, air and water in fractions kept constant with time. The soil receives input from dry particle deposition and wet deposition of particle and gas phase contaminants. Dry air-soil gas exchange acts as both a source and a sink of the soil POP contamination. POPs are lost from the soil layer together with evaporation of water, through leaching of excess water to the underlying ground as well as through biodegradation in the soil [Strand and Hov, 1996]. Biodegradation of POPs in soil is not well quantified and it is described in DEHM-POP using a first-order compound specific degradation rate, ksoil:

  • equation image

where Cs (pg m−3) is the concentration of the simulated compound in soil.

2.2.2. Surface Ocean Water

[16] The ocean compartment in DEHM-POP consists of a well-mixed surface ocean water layer with a depth of 75 m kept constant throughout the year, similarly to the surface layer of the Strand and Hov model [Strand and Hov, 1996]. The deep ocean and the sediments are not taken into account in DEHM-POP and the model does not distinguish between fresh water and ocean water. The surface ocean water receives input from dry particle deposition and wet particle and gas phase deposition. Dry air-water gas exchange acts as both a source and a sink to the surface ocean water concentration of POPs. Several processes contribute to the loss of POPs from the surface ocean water, e.g., biodegradation, hydrolysis, and sedimentation of POPs sorbed to particulate organic carbon. These processes are combined to a compound specific first-order loss rate, kwater, in the DEHM-POP model:

  • equation image

where Cw (pg m−3) is the concentration of the simulated compound in water.

2.2.3. Sea Ice

[17] POPs are not accumulating in the sea ice in DEHM-POP and the effect of the sea ice is thus to reduce air-water exchange and to form a platform for the formation of a snowpack. Data on the extent of sea ice within the model domain are extracted from the numerical weather prediction model MM5v2 together with the meteorological data. It is assumed that there is 5% open water in sea ice covered surface ocean water grid cells to account for leads and polynias. The air-water exchange is thus reduced to 5% in sea ice covered grid cells. As there are no ocean currents included in the model, the sea ice drift and concurrent redistribution of the POPs is not taken into account.

2.2.4. Vegetation

[18] A simple vegetation module with only one type of vegetation is introduced in DEHM-POP. The module describes the absorption of gas phase POPs by the cuticle of leafs. The vegetation coverage is calculated from a data set of global monthly average leaf area index (LAI) with a 0.5° × 0.5° resolution [Myneni et al., 1997]. These data are redistributed to the DEHM-POP model grid and it is assumed that there is a vegetation cover in grid cells with positive LAI values.

[19] The vegetation receives input from wet deposition, dry air-vegetation gas exchange acts as both input and loss and POPs are also lost from the vegetation by degradation and by leaf litter fall. The air-vegetation gas exchange flux in DEHM-POP is calculated from the equation:

  • equation image

where Cg (pg m−3) is the gas phase air concentration, Cv (pg m−3) is the concentration in the vegetation, Kva is the vegetation-air partition coefficient, and vv (m s −1) is the exchange velocity calculated using the Whitman two-layer resistance method [e.g., Mackay and Leinonen, 1975]:

  • equation image

[20] The air-side exchange velocity, vair is calculated from standard boundary layer meteorological parameters [Seinfeld and Pandis, 1998]. The vegetation-side resistance is derived on basis of expressions from the review by Barber et al. [2004] and calculated as:

  • equation image

where Pcut (m s−1) is the cuticle permeance, which is estimated experimentally from one plant species [Schönherr and Riederer, 1989] as cited in [Barber et al., 2004]:

  • equation image

The vegetation-air partition coefficient can be described using the octanol-air partition coefficient [e.g., McLachlan and Horstmann, 1998]:

  • equation image

where m and n are plant specific coefficients that can be determined experimentally. In the DEHM-POP vegetation module m = 26 and n = 0.72 are chosen as the average value between the values determined for coniferous and deciduous forest by McLachlan and Horstmann [1998].

[21] The canopy is intercepting the falling precipitation and a fraction of it is absorbed by the vegetation, whereas the rest is assumed to drip or flow off the vegetation and is thus transferred to the underlaying soil. The vegetation interception fraction, fvege, is in DEHM-POP scaled with the LAI and attains values from 0 to 0.3. When there is a snow cover, it is assumed that the air-vegetation exchange is “shut off.”

[22] The vegetation cover reduces the air-soil gas exchange and when dead leaves fall upon the ground the leaf litter is incorporated into the soil and the vegetation thus acts as an input of organic material to the soil. The decrease in air-soil gas exchange is scaled with the LAI value in each grid cell, so that a maximum LAI value reduces the air-soil exchange to zero. The input of chemical components to soil from leaf litter is in DEHM-POP calculated at the beginning of each month. If the monthly average LAI in a grid cell is smaller than the LAI from the previous month, it is assumed that leafs are lost as litter fall. The fraction of contamination corresponding to the difference between the old and new LAI value is transferred from the vegetation into the underlaying soil compartment, where it is assumed to be immediately incorporated into the soil.

2.2.5. Snow

[23] Snow crystals are potentially very efficient scavengers of both gas and particle phase POPs because of large surface areas of snow crystals [Halsall, 2004]. As snow falls on the ground it forms a snowpack, which is a highly dynamic medium with potentially large and rapid changes in both horizontal and vertical extent.

[24] A snow model describing the exchange of gas phase POPs was developed to be included in DEHM-POP [Hansen et al., 2006]. The snowpack module describes the physical evolution of a homogeneous single-layer snowpack and comprises the three components of snow accumulation, settling and melting with the total surface area available for sorption of gas phase POPs as a key physical parameter. The module was expanded to also handle particle phase POPs and implemented in DEHM-POP [Hansen, 2006; Hansen et al., submitted manuscript, 2007]. The snow receives input from dry particle deposition and wet (snow) deposition of particle and gas phase POPs. Dry air-snow gas exchange acts as both a source and a sink to the snow concentration of POPs, and POPs are furthermore lost from the snowpack by degradation and by meltwater drainage when the snowpack melts.

[25] The particle concentration in the snowpack (TSPsnow) is estimated from the monthly mean deposition of sulphate particles as calculated by the DEHM-SOX model. The particle concentration is assumed to increase linearly through the month, with the increase rate for each time step equal to the fraction of the total deposition that month. Repartitioning of POPs between the particle and the gas phase is calculated by equations (1) and (2) as for the gas-particle partitioning in air but using the TSPsnow concentration instead of TSP. The particles are assumed to remain in the snowpack until it is completely melted at which time they are transferred to the underlying surface compartments.

[26] The horizontal as well as seasonal extent of the snowpack was found to be too large in the DEHM-POP model simulations because of a too inefficient melting process [Hansen, 2006]. In an attempt to counteract this the threshold temperature for the onset of the formation of the snowpack as well as the onset of the melting is adjusted to T = −2°C and the melt rate is increased to 10 cm day−1 K−1 compared to the original snowpack module [Hansen et al., 2006]. The meltwater is assumed to be lost from the snowpack as runoff with the amount of meltwater within the snowpack reduced to 1/e after 24 hours. The runoff, including eventual contamination, is transferred to the underlying compartment. The meltwater refreezes if T < −2°C and the refreeze rate is determined by the amount of heat supplied by the temperature difference between the meltwater and the snowpack. Instantaneous mixing is enforced, where the refrozen melt layer is assumed to have an SSA = 0.

[27] During melt, the meltwater rapidly fills the pore spaces in the snowpack thereby “shutting off” chemical exchange between air and the snow surface. It is assumed that the air-snow gas exchange during periods of melting is replaced by an air-meltwater gas exchange. This exchange is calculated using the same expression as for the air-water interface.

[28] The chemical transformation processes in snow and ice are not well known. POPs in snow are not likely to undergo biodegradation, but there are possibilities of photochemical reactions. As a first approximation of degradation rates in snow, air degradation rates are thus assumed to apply in the snowpack module in DEHM-POP. Degradation of the particle-sorbed fraction of POPs is not included similarly to the assumptions in air.

2.3. Simulations

[29] The atmospheric transport and environmental fate was studied for a full decade by making a simulation covering the period from 1991 to 2000.

2.3.1. Modeled Compound

[30] DEHM-POP is developed using α-hexachlorocyclohexane (α-HCH) as chemical tracer. α-HCH is the major component of the insecticide Technical HCH, which is the most used insecticide worldwide [Li et al., 2000], with an estimated usage of 9.4 × 106 tonnes between 1947 and 1997 and maximum usage around 1980 [Li et al., 2003]. α-HCH is one of the most volatile POPs with a relatively high vapor pressure compared to other POPs. In general it partitions more readily into water because of its low Kaw value. Its relatively low Koa value results in low partitioning to particles in the atmosphere.

[31] To calculate the atmospheric processes and the air-surface exchange processes physical-chemical properties of the studied compound are required. Temperature-dependent Kwa, Koa, and Kow values for α-HCH as derived by Xiao et al. [2004] are used in this study. Temperature-dependent Kia values are taken from Hoff et al. [1995].

[32] The degradation of gas phase α-HCH in air is described using experimentally derived temperature-dependent reaction rates with OH radicals [Brubaker and Hites, 1998]. The degradation rates in vegetation and snow are assumed to be equal to the degradation in air, i.e., dependent on the temperature and OH concentration. In soil the first-order degradation rate for α-HCH is assumed to be: ksoil = 1/(1 year) [Strand and Hov, 1996]. In surface ocean water the first-order loss rate α-HCH is assumed to be: kwater = 1/(10 years), which approximates the biodegradation half life calculated by Harner et al. [1999]. The soil degradation rate and the surface ocean loss rate are assumed to be the same in all grid cells.

2.3.2. Emissions

[33] Monthly averaged emissions spanning the simulated period were used as input to the model simulation. The emissions were prepared on the basis of compiled year by year usage data for each country, distributed on a 1° × 1° gridded cropland as surrogate data [Li et al., 2000]. Emissions into air were then calculated by estimating emission factors from the two events spraying and tiling, and not only fresh usage but also usage from the previous up to 15 years are taken into account [Li et al., 2000].

2.3.3. Initial Environmental Concentrations

[34] Most of the applied technical-HCH was used prior to the start of the model simulation in 1991, and because of the great environmental persistence of α-HCH, residues of previous emissions are still cycling the environment. Initial concentrations in soil are estimated by spinning-up the model for the years 1945 to 1990 using monthly averaged emission estimates for the period as input. Initial concentrations in surface ocean water could be estimated from the same simulation, however the lack of oceanic transport in DEHM-POP gave an erroneous distribution [Hansen, 2006] and the initial surface ocean concentrations are instead estimated from measurements [Hansen et al., 2004]. The initial concentrations in air are set equal to zero because of the rapid mixing of the hemispheric atmosphere (∼1 month). The initial concentrations in snow and vegetation are also set equal to zero because of the highly dynamic structure of these compartments. The concentrations from the first year of the simulation are not used for the model evaluation.

3. Results and Model Evaluation

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information

[35] The α-HCH air concentrations are generally high over areas with primary emissions such as India, southeast Asia, southern Europe, Mexico, and western Africa (Figure 1). Relatively high air concentrations are also seen over areas without primary emissions such as the Atlantic and Pacific Oceans and the Arctic, indicating atmospheric transport from source areas and/or revolatilization of α-HCH previously deposited to these areas. The α-HCH air concentrations decrease through the simulated period, following decreasing emissions (Figure 1). The concentration of α-HCH sorbed to particles is generally less than ∼5%, which is also in the range of observations [e.g., Halsall et al., 1998]. The sum of the simulated gas and particle phase concentrations was evaluated against measurements from 21 monitoring stations within the model domain. Seven of the stations are from the Integrated Atmospheric Deposition Network (IADN) [Buehler and Hites, 2002], three from a long-term measurement campaign in Québec, Canada [Aulagnier and Poissant, 2005], six from the EMEP monitoring network in Europe [e.g., Aas et al., 2003] and five from a multiyear systematic air sampling study established in the framework of the Northern Contaminants Program [Fellin et al., 1996; Hung et al., 2005].

image

Figure 1. Annual average α-HCH gas phase concentrations in the lowermost atmospheric layer for (a) 1992 and (b) 1998. Numbers refer to sampling sites applied for model evaluation (see Table 1).

Download figure to PowerPoint

image

Figure 1. (continued)

Download figure to PowerPoint

3.1. Annual Averages

[36] To evaluate the simulated long-term average α-HCH air concentrations, the measured concentrations and the corresponding simulated concentrations are averaged within each year. At some sites the annually averaged air concentrations appear to be consistently underestimated by the model, namely the five stations from the United States (Sleeping Bear Dunes, Sturgeon Point, Eagle Harbor, Chicago, and Brule River) and the three stations from Québec, Canada (Villeroy, St. Anicet, and Mingan), see Figure 2.

image

Figure 2. Annual averaged measured concentrations plotted against the annual averaged modeled concentrations for the stations in United States and Québec, Canada. The station codes are as follows: −XX, year; CA6, Villeroy; CA7, St. Anicet; CA8, Mingan; US1, Sleeping Bear Dunes; US2, Sturgeon Point; US3, Eagle Harbor; US4, Chicago; US5, Brule River.

Download figure to PowerPoint

[37] The simulated mean concentration of the other 13 sites for all years are almost the same as the mean measured concentration and there is a high correlation (r = 0.68, p < 0.1%). The difference between measured and simulated concentrations are larger than a factor of two for only a few data points (Figure 3). There is also a high correlation of the annual averaged concentrations from the United States and Québec stations (r = 0.85, p < 0.001), however, the simulated mean of all stations and all years is more than a factor of two lower than the measured mean. The possible origin of the discrepancy between measured and predicted concentrations from the U.S. and Québec stations and from the other stations is discussed in detail by K. M. Hansen et al. (Variability of α-HCH in Great Lakes air, manuscript in preparation, 2008, hereinafter referred to as Hansen et al., manuscript in preparation, 2008).

image

Figure 3. Annual averaged measured concentrations plotted against the annual averaged modeled concentrations for the stations other than the United States and Québec, Canada. The station codes are as follows: −XX, year; CA1, Alert; CA2, Kinngait; CA3, Tagish; CA4, Point Petre; CA5, Burnt Island; CZ, Košetice; FI, Pallas; NO1, Spitsbergen; NO2, Lista; SE1, Rörvik; SE2, Aspvreten; RU1, Amderma; RU2, Dunai Island.

Download figure to PowerPoint

3.2. Short-Term Variability

[38] The 21 monitoring stations are listed in Table 1. At ten of the stations (United States: Sleeping Bear Dunes, Sturgeon Point, Eagle Harbor, and Chicago; Canada: Point Petre and Burnt Island; Norway: Lista; Sweden: Rörvik; Russia: Dunai Island; Svalbard: Spitsbergen) the correlation coefficients (r = 0.30 − 0.72) are significant within a 0.1% significance level (p < 0.001). At a further five stations (United States: Brule River; Canada: Villeroy and Kinngait; Finland: Pallas; Czech Rep.: Košetice) the correlation coefficients are lower and less statistically significant (r = 0.18 − 0.37, p < 0.01–0.05), and at the last six stations (Canada: St. Anicet, Mingan, Tagish and Alert; Sweden: Aspvreten; Russia: Amderma) there is no correlation between observed and simulated α-HCH air concentrations.

Table 1. Correlation Coefficients and Significance of Observed and Simulated α-HCH Air Concentrations in the DEHM-POP Model Domaina
Site NumberSite NameYears of SamplingDuration and IntervalNumber of SamplesCorrelation rSignificance p<
  • a

    See Figure 1 for location of sampling sites.

1Sleeping Bear Dunes, USA1992–20001 d/12 d2570.440.001
2Sturgeon Point, USA1992–20001 d/12 d2680.590.001
3Eagle Harbor, USA1992–20001 d/12 d2780.610.001
4Chicago, USA1996–20001 d/12 d1320.300.001
5Brule River, USA1996–20001 d/12 d1420.180.05
6Point Petre, Canada1992–20001 d/12 d2970.690.001
7Burnt Island, Canada1993–20001 d/12 d2260.720.001
8Villeroy, Canada1993–19961 d/6 d1680.230.01
9St. Anicet, Canada1994–19961 d/6 d117
10Mingan, Canada1994–19951 d/12 d26
11Tagish, Canada1992–19957 d/7 d110
12Alert, Canada1992–19987 d/7 d327
13Kinngait, Canada1994–20007 d/7 d440.370.02
14Lista, Norway1992–20001 d/7 d3080.700.001
15Rörvik, Sweden1992–20007 d/30 d900.560.001
16Aspvreten, Sweden1995–20007 d/30 d52
17Pallas, Finland1996–20007 d/30 d600.340.01
18Košetice, Czech Republic1999–20001 d/7 d1010.300.01
19Dunai Island, Russia1993–19957 d/7 d880.420.001
20Amderma, Russia1999–20007 d/7 d91
21Spitsbergen, Svalbard1993–20002 d/7 d4050.370.001

[39] For most of the sites the simulated concentrations agree better with the measured concentrations toward the end of the simulated period. This is probably due to decreasing primary emissions whereby atmospheric levels approach equilibrium with the surface media. Although the dynamics of the weather system creates disequilibrium locally, it will generally result in less variability in air concentrations, which is also seen in both measured and simulated α-HCH concentrations.

[40] It is interesting to note that although the annual average concentration at the U.S. and Quebec stations are underpredicted by the model, the individual observed and simulated concentrations correlate at six of the eight stations. The model thus capture the measured variability at the Great Lakes and Québec, Canada but with a consistent bias, which can be due to erroneous emission estimates, an insufficient parametrization of the model processes or problems with the measurement technique or analytical method. The data from the Great Lakes and Quebec, Canada will be discussed in detail elsewhere (Hansen et al., manuscript in preparation, 2008).

[41] For the other regions there is no consistent bias between measured and simulated concentrations. At the two high-Arctic stations (Spitsbergen, Svalbard and Alert, Canada) the model tends to underestimates winter concentrations, whereas for three low-Arctic sites (Pallas, Finland, Dunai Island, Russia and Tagish, Canada) the model tends to overestimate the concentrations in autumn, and at Kinngait, Canada the model underestimates the autumn and winter concentration. For the European midlatitude stations the model underpredicts spring concentrations at Košetice, Czech Republic, which could be due to too low emission estimates for Eastern Europe. It is not possible to explain the discrepancies between measured and predicted concentrations at all sites, however interesting findings at some of the sites are discussed in more detail in the following.

3.2.1. Lista, Norway

[42] The sampling station of Lista is placed on a small peninsula on the coast of the North Sea at the southern tip of Norway. The air at this station is sampled over one day with samples taken once a week [Haugen et al., 1998]. Only the data from 1992–1995 and 1999–2000 were available for this study. There is a good agreement between measured and simulated concentrations at Lista with a correlation coefficient of r = 0.70, p < 0.001 (Figure 4). The model tends to underestimate the measured concentrations with ∼30%. There is a clear seasonal pattern with high concentrations in summer/early autumn and low concentrations in winter/early spring in both measured and simulated concentrations (Figure 5). The simulated variability of the air concentrations is generally in good agreement with the measured, which indicates that the model captures the individual transport episodes. The agreement between measured and simulated α-HCH air concentrations increase toward the end of the simulated period. These results indicate that the sources of the measured α-HCH air concentrations at Lista are well described in the model. Back trajectory calculations show that about 2/3 of the measured air concentrations at Lista have passed over the North Sea [Haugen et al., 1998]. The good agreement between measured and simulated concentrations indicates that the simple parameterizations in the surface ocean module is sufficient to capture the dynamics of the air-water exchange in this area. The consistent underprediction by the model can be explained if the ocean concentrations in the North Sea are underestimated, but we have not been able to find measurements of α-HCH ocean concentrations in the area to test this hypothesis.

image

Figure 4. Measured versus simulated α-HCH air concentrations for Lista, Norway.

Download figure to PowerPoint

image

Figure 5. Measured and simulated α-HCH air concentrations at Lista, Norway.

Download figure to PowerPoint

3.2.2. Rörvik, Sweden

[43] The Rörvik sampling station is situated on the Swedish west coast in a mainly forested area (30% coniferous, 10% deciduous, and 28% grassland) with some farm land (∼20%). This site is situated only a few grid cells away from Lista in the DEHM-POP model grid. From this station there are a few data from campaigns conducted in spring 1991 and 1992. Continuous air monitoring was started in winter 1994 with one monthly sample integrated over seven days. The measured and simulated concentrations correlates (r = 0.56, p < 0.001), but the model appears to almost consistently overestimate the measured concentrations (Figure 6). The simulated concentrations are in good agreement with the measured concentrations during winter but not during the rest of the year (Figure 7). A clear seasonal pattern is seen in the simulated concentrations with higher concentrations during summer than during winter. This pattern is absent in the measured concentrations, which display no or only weak seasonal patterns. The seasonal pattern in the modeled concentrations becomes less distinct in the end of the simulated period, and in the last year there is very good agreement between the measured and simulated concentrations.

image

Figure 6. Measured versus simulated α-HCH air concentrations for Rörvik, Sweden.

Download figure to PowerPoint

image

Figure 7. Measured and simulated α-HCH air concentrations at Rörvik, Sweden, for 1994 and onward.

Download figure to PowerPoint

[44] The vegetated area surrounding the Rörvik station may act as a sink to the α-HCH air concentrations measured at the station. DEHM-POP may not be able to capture this sink fully with the simplicity of the included vegetation module and the relatively coarse horizontal resolution of the model. This suggests that the measured air concentrations at Rörvik may be more affected by the local surroundings than the ones measured at the nearby Lista station.

3.2.3. Dunai Island, Russia

[45] The station at Dunai Island in the Arctic Ocean north of Siberia was operated from March 1993 to April 1995. There is a low but statistically significant correlation between measured and simulated concentrations (r = 0.42, p < 0.001), with a tendency of higher simulated concentrations than measured (Figure 8). Highest concentrations are observed in spring, with lowest concentrations found during summer (Figure 9). The simulated concentrations also peak in spring and decrease in early summer, however, the concentrations increase through the summer to reach a second peak in autumn, after which it decreases again. Good agreement is seen between measured and simulated concentrations during winter and spring, whereas the model predicts higher concentrations than measured during summer and autumn.

image

Figure 8. Measured versus simulated α-HCH air concentrations for Dunai Island, Russia.

Download figure to PowerPoint

image

Figure 9. Measured and simulated α-HCH air concentrations at Dunai Island, Russia.

Download figure to PowerPoint

[46] The observed spring peak could indicate fresh usage connected to the agricultural season at middle to high latitudes. However, simulations with and without the snowpack module shows that the spring peak is associated with the presence of the snowpack in the model, and the peaking α-HCH air concentrations at Dunai Island is probably released from the snowpack because of increasing temperatures during spring in a large area south of the station (Hansen et al., submitted manuscript, 2007).

3.2.4. Spitsbergen, Svalbard

[47] The station at Svalbard is placed on the Zeppelin Mountain, 474 m above sea level, on the northern part of the island Spitsbergen. The surroundings consist of 100% gravel and stone. There is a low but statistically significant correlation between measured and simulated concentrations (r = 0.37, p < 0.001) (Figure 10). The model is seen to generally underpredict the measured concentrations. There is a large variability in the measured concentrations, especially in the first years, and there is apparently not any seasonal pattern, except toward the end of the simulated period, where there are higher concentrations during summer than winter (Figure 11). In the model results, there is a clear seasonal pattern with high concentrations during summer and low concentrations during winter. The simulated concentrations are in good agreement with the measured concentrations during summer, but too low for the rest of the year. This is opposite to the results from stations at lower latitudes, where the best agreement between measured and simulated concentrations is found during winter. Good agreement between measured and simulated concentrations is seen toward the end of the simulated period, especially in the last year.

image

Figure 10. Measured versus simulated α-HCH air concentrations for Spitsbergen, Svalbard.

Download figure to PowerPoint

image

Figure 11. Measured and simulated concentrations at Spitsbergen, Svalbard, Norway.

Download figure to PowerPoint

[48] The discrepancy between measured and predicted concentrations at Svalbard indicates either a source not accounted for in the model or a too strong sink in the model during winter. It is unlikely that the applied emission estimates are too low since eventual missing emissions are expected to occur during spring and summer and the simulated concentrations from the European and the Russian stations do not indicate too low emission estimates. It is possible that the modeled snowpack too efficiently retains α-HCH at low temperatures during winter, which would result in too low simulated winter α-HCH air concentrations at Svalbard. Too low simulated winter concentrations are also found at the other high-Arctic site: Alert, Canada (not shown), but not at Arctic sites from lower latitudes.

4. Concluding Remarks

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information

[49] An atmospheric chemistry-transport model covering the majority of the Northern Hemisphere was presented. The model was modified to include four different surface compartments each containing processes that describes the behavior of POPs within the compartment and the exchange between the compartments and the atmosphere. The model was evaluated against measurements from monitoring stations within the model domain and the model was able to predict annual averages as well as decadal trends in atmospheric α-HCH concentrations. Significant correlations between simulated and measured short-term atmospheric concentrations of α-HCH were also found at the majority of the investigated monitoring stations. This indicates that the most dominant dynamic processes determining the short-term environmental behavior of persistent organic pollutants are included in the model, and it shows that it is possible to resolve the short-term atmospheric variability of POPs using an atmospheric CTM. This offers an additional tool in the study of the dynamics of the environmental cycling of POPs. Discrepancies between measured and simulated concentrations cannot be explained at all sites. However, at some of the sites the discrepancies in one or more seasons may arise because the measured concentrations are influenced by local features (e.g., vegetation) that are not resolved by the relatively coarse horizontal resolution of the model or by the parameterization presently included in the model. This raise the question of the representativeness of these stations. If the measured concentrations are not representative for a large area surrounding the site the station is not very suitable for evaluating model simulations with a relatively coarse resolution. This also implies that care should be taken when interpreting the measured concentrations from these stations in a regional or global context, for example as part of a larger air monitoring strategy. This study has also highlighted the usefulness of the combined interpretation of measurements and model simulations to study atmospheric levels and variability of persistent organic pollutants. With a closer collaboration between modelers and experimentalists a discussion of the measurement techniques and analytical methods could be included in the interpretation of discrepancies between measured and simulated concentrations and between measured concentrations in some regions, which would enhance our understanding of the atmospheric levels and variability of POPs.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information

[50] We are grateful to all the people that have supplied observational data on α-HCH air concentrations for model evaluation: Wenche Aas for data from the EMEP monitoring network, Eva Brorström-Lundén for 1991 and 1992 data from Rörvik, Pierette Blanchard for data from the IADN network, Hailey Hung and Yushan Su for data from the Canadian Northern Contaminants Program, Phil Fellin for data from the Amderma station, and Laurier Poissant for data from Québec.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information
  • Aas, W., S. Solberg, T. Berg, S. Manø, and K. E. Yttri (2003), Monitoring of long range transported air pollutants, annual report for 2002, SFT Rep. 877/2003 NILU OR 23/2003, Norw. Inst. for Air Res., Kjeller, Norway.
  • Atkinson, R., R. Guicherit, R. A. Hites, W.-U. Palm, J. N. Seiber, and P. de Voogt (1999), Transformations of pesticides in the atmosphere: A state of the art, Water Air Soil Pollut., 115, 219243.
  • Aulagnier, F., and L. Poissant (2005), Some pesticides occurrence in air and precipitation in Québec, Canada, Environ. Sci. Technol., 39(9), 29602967.
  • Barber, J. L., G. O. Thomas, G. Kerstiens, and K. C. Jones (2004), Current issues and uncertainties in the measurement and modelling of air-vegetation exchange and within-plant processing of POPs, Environ. Pollut., 128, 99138.
  • Bidleman, T. F. (1988), Atmospheric processes, Environ. Sci. Technol., 22(4), 361367.
  • Breivik, K., and F. Wania (2002), Evaluating a model of the historical behavior of two hexachlorocyclohexanes in the Baltic Sea environment, Environ. Sci. Technol., 36(5), 10141023.
  • Brubaker, W. W., and R. A. Hites (1998), OH reaction kinetics of gas-phase α- and γ-hexachlorocyclohexane and hexachlorobenzene, Environ. Sci. Technol., 32(6), 766769.
  • Buehler, S. S., and R. A. Hites (2002), The Great Lakes' integrated atmospheric deposition network, Environ. Sci. Technol., Part A, 36(17), 354359.
  • Christensen, J. H. (1997), The Danish Eulerian Hemispheric Model—A three-dimensional air pollution model used for the Arctic, Atmos. Environ., 31(24), 41694191.
  • Christensen, J. H., J. Brandt, L. M. Frohn, and H. Skov (2004), Modelling of mercury in the Arctic with the Danish Eulerian Hemispheric Model, Atmos. Chem. Phys., 4, 22512257.
  • Fellin, P., L. A. Barrie, D. Dougherty, D. Toom, D. Muir, N. Grift, L. Lockhart, and B. Billeck (1996), Air monitoring in the Arctic: Results for selected persistent organic pollutants for 1992, Environ. Toxicol. Chem., 15(3), 253261.
  • Frohn, L. M., J. H. Christensen, and J. Brandt (2002), Development of a high-resolution nested air pollution model—The numerical approach, J. Comput. Phys., 179(1), 6894.
  • Geels, C., S. C. Doney, R. Dargaville, J. Brandt, and J. H. Christensen (2004), Investigating the sources of synoptic variability in atmospheric CO2 measurements over the Northern Hemisphere continents: A regional model study, Tellus, Ser. B, 56, 3550.
  • Grell, G. A., J. Dudhia, and D. R. Stauffer (1995), A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5), NCAR Tech. Note, NCAR/TN-398+STR, Mesoscale and Microscale Meteorol. Div., Natl. Cent. for Atmos. Res., Boulder, Colo.
  • Halsall, C. J. (2004), Investigating the occurence of persistent organic pollutants (POPs) in the Arctic: Their atmospheric behaviour and interaction with the seasonal snow pack, Environ. Pollut., 128(1–2), 163175.
  • Halsall, C. J., R. Bailey, G. A. Stern, L. A. Barrie, P. Fellin, D. C. G. Muir, B. Rosenberg, Y. Rovinsky, E. Y. Kononov, and B. Pastukhov (1998), Multi-year observations of organohalogen pesticides in the Arctic atmosphere, Environ. Pollut., 102, 5162.
  • Hansen, K. M. (2006), Transport of persistent organic pollutants into the Arctic, Ph.D. thesis, 250 pp., Dep. of Atmos. Environ., Natl. Environ. Res. Inst., Roskilde, Denmark. (Available at http://www2.dmu.dk/1_viden/2_Publikationer/3_Ovrige/rapporter/phd_kmh.pdf).
  • Hansen, K. M., J. H. Christensen, J. Brandt, L. M. Frohn, and C. Geels (2004), Modelling atmospheric transport of α-hexachlorocyclohexane in the Northern Hemisphere with a 3-D dynamical model: DEHM-POP, Atmos. Chem. Phys., 4, 11251137.
  • Hansen, K. M., C. J. Halsall, and J. H. Christensen (2006), A dynamic model to study the exchange of gas-phase persistent organic pollutants between air and a seasonal snowpack, Environ. Sci. Technol., 40(8), 26442652.
  • Harner, T., and T. F. Bidleman (1998), Octanol-air partition coefficient for describing particle/gas partitioning of aromatic compounds in urban air, Environ. Sci. Technol., 32(10), 14941502.
  • Harner, T., H. Kylin, T. F. Bidleman, and W. M. J. Strachan (1999), Removal of α- and γ-hexachlorocyclohexane and enantiomers of α-hexachlorocyclohexane in the eastern Arctic Ocean, Environ. Sci. Technol., 33(8), 11571164.
  • Haugen, J.-E., F. Wania, N. Ritter, and M. Schlabach (1998), Hexachlorocyclohexanes in air in Southern Norway. Temporal variation, source allocation, and temperature dependence, Environ. Sci. Technol., 32(2), 217224.
  • Hoff, J. T., F. Wania, D. Mackay, and R. Gillham (1995), Sorption of nonpolar organic vapors by ice and snow, Environ. Sci. Technol., 29(8), 19821989.
  • Hung, H., et al. (2005), Temporal and spatial variabilities of atmospheric polychlorinated biphenyls (PCBs), organochlorine (OC) pesticides and polycyclic aromatic hydrocarbons (PAHs) in the Canadian Arctic: Results from a decade of monitoring, Sci. Total Environ., 342, 119144.
  • Koziol, A. S., and J. A. Pudykiewicz (2001), Global-scale environmental transport of persistent organic pollutants, Chemosphere, 45, 11811200.
  • Legagneux, L., A. Cabanes, and F. Dominé (2002), Measurement of the specific surface area of 176 snow samples using methane adsorption at 77 K, J. Geophys. Res., 107(D17), 4335, doi:10.1029/2001JD001016.
  • Lei, Y. D., and F. Wania (2004), Is rain or snow a more efficient scavenger of organic chemicals? Atmos. Environ., 38, 35573571.
  • Li, Y.-F., M. T. Scholtz, and B. J. van Heyst (2000), Global gridded emission inventories of α-hexachlorocyclohexane, J. Geophys. Res., 105(D5), 66216632.
  • Li, Y.-F., M. T. Scholtz, and B. J. van Heyst (2003), Global gridded emission inventories of β-hexachlorocyclohexane, Environ. Sci. Technol., 37(16), 34933498.
  • Mackay, D., and P. J. Leinonen (1975), Rate of evaporation of low- solubility contaminants from water bodies to atmosphere, Environ. Sci. Technol., 9(13), 11781180.
  • MacLeod, M., W. J. Riley, and T. McKone (2005), Assessing the influence of climate variability on atmospheric concentrations of polychlorinated biphenyls using a global-scale mass balance model (BETR Global), Environ. Sci. Technol., 39(17), 67496756.
  • Malanichev, A., E. Mantseva, V. Shatalov, B. Strukov, and N. Vulykh (2004), Numerical evaluation of the PCBs transport over the Northern Hemisphere, Environ. Pollut., 128(1–2), 279289.
  • McLachlan, M. S., and M. Horstmann (1998), Forests as filters of airborne organic pollutants: A model, Environ. Sci. Technol., 32(3), 413420.
  • Myneni, R. B., R. R. Nemani, and S. W. Running (1997), Algorithm for the estimation of global land cover, LAI and FPAR based on radiative transfer models, IEEE Trans. Geosci. Remote Sens., 35, 13801393.
  • Prevedouros, K., M. MacLeod, K. C. Jones, and A. J. Sweetman (2004), Modelling the fate of persistent organic pollutants in Europe: Parameterisation of a gridded distribution model, Environ. Pollut., 128(1–2), 251261.
  • Scheringer, M., F. Wegmann, K. Fenner, and K. Hungerbühler (2000), Investigation of the cold condensation of persistent organic pollutants with a global multimedia fate model, Environ. Sci. Technol., 34(9), 18421850.
  • Schönherr, J., and M. Riederer (1989), Foliar penetration and accumulation of organic chemicals in plant cuticles, Rev. Environ. Contamin. Toxicol., 108, 170.
  • Seinfeld, J. H., and S. N. Pandis (1998), Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, John Wiley, Hoboken, N. J.
  • Semeena, V. S., and G. Lammel (2003), Effects of various scenarios of entry of DDT and γ-HCH on the global environmental fate as predicted by a multicompartment chemistry-transport model, Fresenius Environ. Bull., 12(8), 925939.
  • Semeena, V. S., and G. Lammel (2005), The significance of the grasshopper effect on the atmospheric distribution of persistent organic substances, Geophys. Res. Lett., 32, L07804, doi:10.1029/2004GL022229.
  • Skov, H., J. H. Christensen, M. E. Goodsite, N. Z. Heidam, B. Jensen, P. Wåhlin, and G. Geernaert (2004), Fate of elemental mercury in the Arctic during atmospheric mercury depletion episodes and the load of atmospheric mercury to the Arctic, Environ. Sci. Technol., 38(8), 23732382.
  • Strand, A., and Ø. Hov (1996), A model strategy for the simulation of chlorinated hydrocarbon distributions in the global environment, Water Air Soil Pollut., 86, 283316.
  • Wania, F., and D. Mackay (1999), Global chemical fate of α-hexachlorocyclohexane. 2. Use of a global distribution model for mass balancing, source apportionment and trend prediction, Environ. Toxicol. Chem., 18(7), 14001407.
  • Wania, F., J. Axelman, and D. Broman (1998), A review of processes involved in the exchange of persistent organic pollutants across the air-sea interface, Environ. Pollut., 102, 323.
  • Wania, F., D. Mackay, Y.-F. Li, T. F. Bidleman, and A. Strand (1999), Global chemical fate of α-hexachlorocyclohexane, 1. Evaluation of a global distribution model, Environ. Toxicol. Chem., 18(7), 13901399.
  • Xiao, H., N. Li, and F. Wania (2004), Compilation, evaluation, and selection of physical-chemical property data for α-, β-, and γ-Hexachlorocyclohexane, J. Chem. Eng. Data, 49, 173185.

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. DEHM-POP Model
  5. 3. Results and Model Evaluation
  6. 4. Concluding Remarks
  7. Acknowledgments
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
jgrd13826-sup-0001-t01.txtplain text document1KTab-delimited Table 1.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.