The Lagrangian particle dispersion model FLEXPART was used to construct a global data set of 1.4 million continuous trajectories. At the model start, particles were distributed homogeneously in the atmosphere and were then transported for 5.5 years using both resolved winds from European Centre for Medium-Range Weather Forecasts analyses and parameterized turbulent and convective transport. On the basis of this data set, a climatology of transport in and to the Arctic was developed. It was found that the time air resides continuously north of 70°N, called its Arctic age, is highest near the surface in the North American sector of the Arctic. North of 80°N and near the surface, the mean Arctic age of air is about 1 week in winter and 2 weeks in summer. It decreases rapidly with altitude to about 3 days in the upper troposphere. In the most isolated regions of the Arctic, air is exposed to continuous darkness for, on average, 10–14 days in December. Transport from the stratosphere to the lower troposphere is much slower in the Arctic than in the middle latitudes. In the central Arctic, for instance, the probability that air near the surface was transported from the stratosphere within 10 days is only about 1% in winter and 0.3% in summer. Air pollution can be transported into the Arctic along three different pathways: low-level transport followed by ascent in the Arctic, low-level transport alone, and uplift outside the Arctic, followed by descent in the Arctic. Only this last pathway is frequent for pollution originating from North America and Asia, whereas European pollution can follow all three pathways in winter, and pathways one and three in summer. Sensitivities of Arctic air masses to emissions of air pollutants, based on transport alone, were calculated for times of up to 30 days before the air masses reached the Arctic. They were highest over Siberia and Europe in winter and over the oceans in summer. Using an inventory for anthropogenic black carbon (BC) emissions, it was found that near the surface and for transport timescales of 5 and 10 days, BC source contributions from south Asia are only 1.6% and 10%, respectively, of the corresponding European values, despite much higher emissions in south Asia. Using an inventory for BC emissions from forest fires, BC source contributions to the Arctic, particularly from fires in Siberia, were larger than anthropogenic BC source contributions in summer in years of average burning.
 Because of its remoteness, the Arctic troposphere, defined for the purpose of this study as the region north of 70°N, was long believed to be extremely clean. However, in the 1950s, pilots flying over the North American Arctic discovered a strange haze [Greenaway, 1950; Mitchell, 1957], which decreased visibility significantly. The haze phenomenon, accompanied by high levels of gaseous air pollutants (e.g., hydrocarbons [Solberg et al., 1996]), was observed regularly since then especially in the lower troposphere and is a result of the special meteorological situation in the Arctic in winter and early spring [Shaw, 1995]. Temperatures at the surface become extremely low, leading to a thermally very stable stratification with frequent and persistent occurrences of surface-based inversions [Bradley et al., 1992] that reduce turbulent exchange, hence dry deposition. Turbulence intensities can be enhanced over open water [Strunin et al., 1997] but this is rare in winter. The extreme dryness of the Arctic troposphere also minimizes wet deposition, thus leading to very long aerosol lifetimes in the Arctic in winter.
 Surfaces of constant potential temperature form closed domes over the Arctic, with minimum values in the Arctic boundary layer (BL) [Klonecki et al., 2003]. This isolates the Arctic lower troposphere from the rest of the atmosphere by a transport barrier, the so-called “Arctic front.” Meteorologists realized that in order to facilitate isentropic transport, a pollution source region must have the same low potential temperatures as the Arctic Haze layers [Carlson, 1981; Iversen, 1984; Barrie, 1986]. This rules out most of the world's pollution source regions because they are too warm, and leaves northern Eurasia as the main source region for the Arctic Haze [Rahn, 1981; Barrie, 1986]. There the Arctic front can be located as far south as 40°N on average in January (see Figure 1 of Barrie ). Furthermore, northern Eurasia is on a preferred pathway into the polar dome that involves diabatic cooling of air traveling over snow-covered land. This transport is highly episodic and often related to large-scale blocking events [Raatz and Shaw, 1984; Iversen and Joranger, 1985]. In contrast, air masses leaving North America's densely populated east coast are heated diabatically [Klonecki et al., 2003] because of the frequent warm conveyor belts over the downwind North Atlantic Ocean [Stohl, 2001; Stohl et al., 2002; Eckhardt et al., 2003]. Southeast Asia is located at even higher potential temperatures than North America and so was also rejected as the source of the Arctic Haze, although an Asian desert origin was suggested for elevated haze layers [Rahn et al., 1977], which are also quite frequent [Leiterer et al., 1997].
 Pollution transport to the Arctic varies considerably. During the positive phase of the North Atlantic Oscillation (NAO), transport from all three Northern Hemisphere continents (Europe, North America and Asia in order of significance) into the Arctic is enhanced, resulting in higher Arctic pollution levels [Eckhardt et al., 2003; Duncan and Bey, 2004]. Given the long-term changes of the NAO, this must be considered when studying Arctic pollution trends [Macdonald et al., 2005].
 Recently, new issues have attracted scientific interest. Anthropogenic emissions in south and east Asia have been growing rapidly during the past decades. Especially black carbon (BC) emissions are now much larger there than in Europe, North America and Russia combined. BC is important because it absorbs solar radiation and can lead to a strong albedo reduction if deposited on snow or ice [Hansen and Nazarenko, 2004]. In a recent model study, Koch and Hansen  suggest that south Asia is now the dominant source of BC in the Arctic upper troposphere and is comparable to the European source near the Arctic surface. However, if the thermodynamic argument holds that in winter it is almost impossible for a south Asian air mass to reach the Arctic lower troposphere [Carlson, 1981; Iversen, 1984; Bowling and Shaw, 1992], it is not clear how Asian BC, given BC's short atmospheric lifetime of 6 ± 2 days [Park et al., 2005], can intrude into the polar dome in the model of Koch and Hansen .
 Another large episodic source of BC and other pollutants are wildfires [Lavoué et al., 2000]. It was realized only recently that boreal wildfire emissions influence the atmospheric composition on a hemispheric scale [Wotawa et al., 2001]. Pollution plumes originating from these fires have been observed over downwind continents [Forster et al., 2001] and can circle the entire Northern Hemisphere [Damoah et al., 2004]. An aircraft campaign frequently sampled aerosol plumes from Alaskan and maybe also Siberian forest fires over the Alaskan Arctic [Shipham et al., 1992], and a Ph.D. thesis suggests a link between BC observations in Greenland and at other Arctic sites to boreal forest fires [Lavoué, 2000]. During summer 2004, severe forest fires burning in Alaska and Canada led to strong increases in BC concentrations at four stations located in different parts of the Arctic (A. Stohl et al., Pan-Arctic enhancements of light-absorbing aerosol concentrations due to North American boreal forest fires during summer 2004, submitted to Journal of Geophysical Research, 2006) (hereinafter referred to as Stohl et al., submitted manuscript, 2006). Furthermore, the positive trend in areas burned over recent decades [Lavoué et al., 2000; Kasischke et al., 2005], likely because of a warming in the boreal region, is a matter of concern. Already, there are speculations that deposition of BC from boreal forest fires could contribute to the melting of Arctic glaciers and sea ice [Kim et al., 2005]. Boreal fires are a summertime phenomenon when removal mechanisms (wet and dry deposition) are relatively efficient and the Arctic troposphere is generally much cleaner than in winter. However, precipitation, and hence wet removal, is suppressed by the high particle numbers in the vicinity of the fires [Andreae et al., 2004], such that much of the BC can possibly reach the Arctic and be deposited there.
 Transport from the stratosphere to the Arctic troposphere is also discussed controversially. It is known that tropopause folds, which are frequent at the polar front, occur also at the Arctic front [Rao and Kirkwood, 2005]. Chemical measurements suggest a strong influence of transport from the stratosphere on the ozone concentrations in the Arctic free troposphere [Dibb et al., 2003; Allen et al., 2003]. At the same time, the frequency of stratosphere-troposphere-transport events reaching the lower troposphere in the Arctic is low [James et al., 2003; Sprenger and Wernli, 2003], and their influence on surface ozone at Alert is also small [Dibb et al., 1994]. In fact, potential temperatures in the lower stratosphere are higher than in pollution source regions in the middle latitudes, and thus even more diabatic cooling is required for stratospheric air to penetrate the polar dome.
 This paper revisits the role of transport into the Arctic troposphere. In contrast to most previous papers, this study considers both winter and summer. It also compares timescales for transport from the stratosphere with timescales for transport from the midlatitude troposphere, and it reports the first calculations of how long air typically remains in darkness. Finally, it discusses pollution transport in the context of BC, whose emission distribution is different from those species that have been used in most previous studies of transport to the Arctic. All this is done with a single consistent data set.
 FLEXPART calculates trajectories using the mean winds interpolated from the ECMWF analyses plus random motions that account for turbulence. The theory of stochastic particle models is described, for instance, in the monograph by Rodean . In the BL, random motions are calculated by solving Langevin equations for Gaussian turbulence [Stohl and Thomson, 1999], using parameterizations of Hanna . BL heights are calculated using a combined Richardson number and lifting parcel technique [Vogelezang and Holtslag, 1996]. Turbulence is assumed to be small in the troposphere, and even smaller in the stratosphere using a diffusion coefficient derived by Legras et al. . For moist convective transport, FLEXPART uses the scheme of Emanuel and Živković-Rothman , as described and tested by C. Forster et al. (Parameterization of convective transport in a Lagrangian particle dispersion model and its evaluation, submitted to Journal of Applied Meteorology, 2006). In order to maintain high accuracy near the pole, FLEXPART advects particles on a polar stereographic projection poleward of 75°.
 At the start of the FLEXPART simulation, the global atmosphere was “filled” homogeneously with 1.4 million particles. Particles were then allowed to move freely and their positions, and meteorological data interpolated from the ECMWF analyses (temperature T, specific humidity q, pressure p, potential temperature Θ, tropopause heights, etc.) were recorded in output files every 6 hours. As particles were not repositioned during the simulation, the model must not violate the so-called well-mixed criterion [Thomson, 1987], i.e., accumulate particles or leave voids in certain regions. For monitoring the particle distribution, the surface pressure and air density were reconstructed by counting the particle masses on a coarse output grid. Reconstructed surface pressure was found to be systematically too high by about 5–10% in the tropics, compensated by particle losses in the middle latitudes. This feature developed during the first year of the simulation and remained stable thereafter. The largest deviation from well-mixedness was an about 20% too high particle density in the winter-time Antarctic lower stratosphere, compensated by particle losses in the southern midlatitude stratosphere. The reason for these deviations from well-mixedness is not clear and needs further study. They are small enough, however, to be negligible for this study.
 Transport analyses were done on a three-dimensional grid with 3° longitude × 2° latitude and in nine layers. For displaying the pathways of air masses to the Arctic, first a particular set of Arctic particles was defined (see later). These particles were traced back in time to calculate a so-called potential emission sensitivity (PES) function, as described by Seibert and Frank  and Stohl et al. . The word “potential” here indicates that this sensitivity is based on transport alone, ignoring removal processes that otherwise would reduce the sensitivity. The value of the PES function (in units of s kg−1) in a particular grid cell is proportional to the residence time particles spend in that cell and is a measure for the mixing ratio at the receptor (i.e., in the Arctic volume defined by the selected particles) that a source of unit strength (1 kg s−1) in the respective grid cell would produce. Of special interest is the PES distribution close to the ground because most sources are located there. Thus, in this paper, PES values are reported only for the layer 0–500 m above ground. Folding (i.e., multiplying) the PES distribution in this layer with the distribution of actual emission flux densities (in units of kg m−2 s−1) from an emission inventory yields a so-called potential source contribution (PSC) map, which indicates the potential relevance of different regions as sources for the Arctic. Spatial integration of the PSC map finally gives the mass mixing ratio of the emitted species in the selected air masses.
 Emission distributions for different chemical species differ from each other. In this study, only BC is considered which is perhaps the species whose emissions are most heavily weighted toward south and east Asia, and at the same time an important species for the Arctic, as described in the previous section. BC emission data, shown in Figure 1, were taken from Bond et al.  for combined fossil fuel and biofuel emissions for the year 1996, and from Lavoué et al.  for boreal and temperate wildfires for an average year in the 1980s. Both data sets are available on a 1° × 1° grid, and the Lavoué data set also provides monthly information. Anthropogenic BC emissions from Asia are a factor of 2.5 higher than those from Europe and North America combined (continents are defined here according to their geographical definition). Annual BC emissions from boreal forest fires are more than 1 order of magnitude smaller than anthropogenic emissions but they are centered much further north. Furthermore, the emission flux in June and July is about 3 times the annual average reported in Figure 1.
 The major limitation of using FLEXPART for this study is that no chemical processes are accounted for, whereas concentration levels of actual chemical species are affected also by transformation and removal mechanisms. Thus simulated concentrations of BC are an upper estimate and must be seen against the “transport only” assumption of this analysis. However, exploring the role of transport is the only purpose of this paper.
3.1. Arctic Age of Air
 A good measure of the isolation of the Arctic troposphere is the time it takes for air to leave it. The time a particle has spent continuously north of 70°N shall thus be called its Arctic age. The Arctic age was calculated for every particle and averages were mapped on the analysis grid. Figure 2 shows the average Arctic age of air from the surface to 100 m above it, for January (Figure 2a) and July (Figure 2b), the months with the lowest and highest mean values. During both months, the air above the high topography of Greenland is the least aged, which reflects the higher wind speeds but also stronger meridional motions west of Greenland due to the frequent northward-traveling cyclones over Baffin Bay. In January, the most aged air (about 8–10 days, north of 75°N) is found in the North American sector of the Arctic whereas the least aged air (about 6 days at 80°N) resides north of Eurasia. This is consistent with the average wind patterns [Macdonald et al., 2005] and a mainly Eurasian source of low-level Arctic air pollution in winter [Barrie, 1986; Eckhardt et al., 2003]. In July, the age of air distribution is more symmetric with a rather homogeneous average value of about 13–17 days north of 75°N, with the exception of the Atlantic sector where less aged air marks a major entry route to the Arctic in summer. These results are only moderately sensitive to changes of the definition of the Arctic region. Analyses using the polar circle as the southern boundary yielded maximum ages in winter about 20% higher than those in Figure 2 but unchanged patterns. As pollution emissions occur mainly close to the surface, the time was also counted since the air was last south of 70°N and at the same time below 1000 m. This leads to about 50% higher values than shown in Figure 2, but again with very similar patterns.
Figure 3 shows time series of the monthly mean Arctic age of air, averaged over the central Arctic (north of 80°N), and for four different altitude ranges. The Arctic age of air decreases strongly with altitude, and also the amplitude of the seasonal cycle decreases, giving a rather constant mean value of about 3 days between 5 and 8 km. The seasonal minimum near the surface is generally reached in January whereas the maximum can occur from May to October. The interannual variability is less than about 20% and does not appear to be related closely to the Arctic Oscillation but the time period is too short for this to be studied properly.
3.2. Time in Darkness
 Sunlight fuels photolysis reactions and normally plays an important role in atmospheric chemistry. In the Arctic winter, however, its absence shuts down photochemistry. The duration of the polar night at a given location follows from astronomical principles. However, how long Arctic air is exposed to continuous darkness also depends on how frequently it travels far enough south to escape the polar night. Therefore it was determined when the solar zenith angle at a particle's position was less than 90° for the last time. As calculations were done once every hour, shorter periods with sunlight may have been missed. However, the sun rises very slowly in the Arctic and photolysis is insignificant for large solar zenith angles.
Figure 4 shows maps of the mean time that air in the lowest 100 m of the atmosphere has spent in darkness, for December. As for the Arctic age of air, the time in darkness maximizes in the North American sector of the Arctic with about 10–14 days in December. Time in darkness also decreases with altitude. In December, as an average for the region north of 80°N (Figure 5), it is about 10 days in the lowest 100 m, but only 4 days between 5 and 8 km.
3.3. Transport From the Stratosphere
 Time was counted since particles last crossed the thermal tropopause [World Meteorological Organization, 1986]. In the lower troposphere, the average times since a particle left the stratosphere are shortest in the middle latitudes and longest in the tropics and in the Arctic. However, average times are on the order of 100 days and show relatively little spatial or seasonal variability. As a better measure, it was calculated how many of the particles in a grid cell have left the stratosphere within a given time period (from 1 to 20 days). This can also be interpreted as the probability a particle in this grid cell originated in the stratosphere. For the lowest 3 km, a timescale of 4 days, and for winter, the results are very similar to the corresponding ones from James et al.  and Figure 5 of Sprenger and Wernli , with probability maxima over the eastern Pacific/western North America and just off the eastern seaboard of North America. Here the focus is on the Arctic, and these dominant midlatitude features shall not be discussed.
 As the probabilities that particles are transported from the stratosphere to the Arctic lower troposphere within 4 days are extremely low, Figure 6 shows the results for a timescale of 10 days, for the lowest 500 m of the atmosphere. In winter (Figure 6, top), compared to the highest values of more than 20% at about 20°N–30°N over the eastern Pacific and western North America (not shown), probabilities in the entire Arctic, except for over the high topography of Greenland, are very low: 1%, or less. Low values extend also to the sub-Arctic over eastern North America and over Siberia. In summer (Figure 6, bottom), probabilities around the pole are even less than 0.1%. Interestingly, in summer similarly low values are also found over the Bering Sea at latitudes of 50°N–60°N, whereas otherwise probabilities in the middle latitudes are more typically on the order of 1–10%.
Figure 7 shows time series of the transport probabilities of originally stratospheric air masses for the central Arctic for different altitude ranges, for a timescale of 10 days. Similarly to what has been reported for deep stratosphere-troposphere transport events in general [James et al., 2003; Sprenger and Wernli, 2003], transport probabilities are highest in winter and lowest in summer. Probabilities increase dramatically with altitude, being higher by a factor of 2 already for the 0.5- to 1.5-km layer than near the surface. This reflects the stable layering of the polar dome both in winter and in summer. Transport probabilities in the Arctic are much lower than in the middle latitudes up to about 3 km but are similar or even higher in the upper troposphere.
 Using a longer timescale of 20 days, the spatial patterns of stratospheric transport probabilities remain similar but gradients to the middle latitudes, as well as the seasonal cycle, are reduced. The probabilities for this timescale are also much higher, about 5% in winter and 2–3% in summer, for the central Arctic near the surface.
3.4. Tropospheric Transport to the Arctic
 To shed light on pollution transport to the Arctic, in a first step particles residing for 5 days or more north of 70°N were identified. The 5-day residence time criterion was used in order to select only particles north of, or above, the Arctic front. In some regions, especially over the eastern North Atlantic Ocean, the Arctic front can be located almost permanently north of 70°N, but air would typically spend less than 5 days there. Using the selected particles, the PES values (see section 2) in the lowest 500 m of the atmosphere were calculated for the last 3, 10, and 30 days before the particles crossed 70°N. As time was counted from the crossing of latitude 70°N, particles are required to leave the Arctic (backward in time) at time zero but are allowed to reenter it (also backward in time). For winter, the results for the last 3 days (Figure 8, top) show indeed that some of the particles have left and reentered the Arctic but the maximum PES is south of 70°N. The preferred pathway to the Arctic is via high-latitude Eurasia, whereas there is a minimum over the relatively warm North Atlantic Ocean. Those few particles that do arrive from over the oceans, can come from far south (to 30°N), reflecting upward transport and high wind speeds in the storm tracks [Stohl, 2001]. Transport is slower over the continents where PES values south of 50°N are low.
 Considering the last 10 days (Figure 8, middle), high PES extends from high-latitude Eurasia to the industrialized regions of Europe. The PES values over the densely populated eastern seaboards of North America (Asia) are about a factor of 4 (10) lower than over central Europe. Integrating over 30 days (Figure 8, bottom), the zonal variability in PES values is small but the meridional variation from 10°N to 60°N still spans 3 orders of magnitude. This means that a source at 10°N needs to be 3 orders of magnitude stronger than one at 60°N for impacting the Arctic equally strong over 30 days of transport. The PES values are still higher by a factor of 2 (4) over Europe than over eastern North America (eastern Asia).
 In order to investigate the emission sensitivity also for air masses in the Arctic lower troposphere only, the calculations were repeated for a subset of particles that reached a minimum altitude below 1000 m in the Arctic (Figure 9). In this case, PES values are higher, particularly for the first few days of transport, because of the generally lower altitude of these particles. They are also more concentrated at higher latitudes and peak even more clearly over northern Eurasia than when all Arctic particles are used.
Figure 10 shows BC PSC maps for the last 3 and 30 days before crossing 70°N, respectively, for winter. For all Arctic particles and the last 3 days (Figure 10, top), sources in northern Europe dominate the BC signal. For 10 days (not shown), PSCs are large all over Europe, and there is also a smaller signal over the North American east coast and a weak but widespread signal over east Asia. For 30 days (Figure 10, middle), BC PSCs in east Asia and Europe are of similar magnitude but are lower in North America. For the subset of particles reaching altitudes below 1 km in the Arctic and for 30 days (Figure 10, bottom), American and Asian PSCs are almost the same as for all particles. However, the European PSC is doubled, a result of the increased PES in this region (see Figure 9).
Figure 11 summarizes the BC PSC analyses by showing integrations over the areas of Europe, North America and Asia, as a function of transport time, for all Arctic particles (top) and for the low-altitude subset (bottom), in winter. In both cases, Europe makes the largest PSC and North America the smallest, regardless of the transport time. Koch and Hansen  recently reported that in their model south Asia, Russia and Europe each contribute about 20–25% BC near the surface and that south Asian BC dominates at higher altitudes. Global aerosol models suggest a BC atmospheric lifetime of 6 ± 2 days [Park et al., 2005] or 3–4 days [Liu et al., 2005]. Koch and Hansen  themselves report a lifetime of 7.3 days. In order to see how consistent the results of Koch and Hansen  are with the results reported here, the PSC from Asia south of 50°N (a region actually larger than the south Asia region defined by Koch and Hansen ) was compared to the European PSC. For the low-altitude particle subset, the ratios of PSCs integrated over south Asia and Europe, respectively are 0% (for a transport time of up to 2 days), 0.2% (3 days), 1.6% (5 days), 10% (10 days), 25% (20 days) and 40% (30 days). Thus, for BC lifetimes consistent with global aerosol model results (between 3 and 8 days), south Asian BC would contribute less than 10% of the European BC in the Arctic lower troposphere, whereas Koch and Hansen  reported an about equal contribution. For all Arctic particles, the corresponding ratios are 0% (2 days), 0.5% (3 days), 4.6% (5 days), 27% (10 days), 62% (20 days) and 86% (30 days), whereas Figure 7 of Koch and Hansen  suggests a dominance of south Asian emissions. The discrepancy is not due to different emission fields, as both studies used the same inventory.
 A reason for the discrepancy might be the particular way Arctic particles were defined here. To test this hypothesis, PES functions were also calculated on-line in additional FLEXPART backward simulations [Stohl et al., 2003], for Januaries and Februaries 2000–2005. In these simulations, all the air in nine layers north of 80°N was traced back for 30 days using 500,000 particles per month and layer. The resulting south Asian PSC below 100 m (Table 1) is only 19% of the European PSC, which actually is a factor of two smaller than the values reported earlier for the lower tropospheric particle subset after 30 days. This shows that close to the surface the European BC PSC dominates even more. The European PSC decreases above 1000 m, whereas the Asian PSC increases up to the middle troposphere. The south Asian PSC dominates over the European PSC in the Arctic upper troposphere but the European PSC to the total column is still about 20% larger than the Asian PSC after 30 days of transport, similar to the previous results. In summary, the difference between this study's results and the results of Koch and Hansen  is not due to the particular selection of particles used here.
Table 1. Average Contributions (ng m−3) From Europe and South Asia to BC Concentrations North of 80°N After 30 Days of Transport, for Januaries and Februaries in the Years 2000–2005
 Transport from most pollution source regions to the Arctic lower troposphere requires a diabatic-cooling-driven penetration of the polar dome either from above, or sideways. In the first case, aerosols would be removed efficiently by wet deposition during the ascent. In contrast, the diabatic cooling from the surface associated with a low-altitude route would stabilize the air, hence reducing dry deposition velocities. In order to distinguish between the two possibilities, meteorological parameters were recorded along all trajectories from major pollution source regions to the Arctic. Similarly sized 1000-m-high boxes over Europe (10°W–50°E, 4°N0–61°N), North America (124°W–69°W, 30°N–49°N) and east Asia (101°E–129°E, 11°N–44°N) were used as source regions. When a particle left such a box (either by ascending or laterally), the time until the particle reached the Arctic was counted, up to a maximum of 30 days. It was considered for further analyses only when it subsequently stayed in the Arctic for at least 5 days and had a minimum altitude below 1000 m there, consistent with the criteria for the low-altitude subset used previously. Trajectories were binned into one of 60 classes according to the time (from half a day to 30 days) between leaving the source box and reaching the Arctic. Meteorological parameters along the trajectories were then averaged for every transport bin from 5 days before leaving the source region until 15 days after entering the Arctic (of which particles stayed at least the first five in the Arctic). The averaging removes all information on individual trajectories but essential features are conserved as different pathways have different characteristic transport times.
Figure 12 shows plots of p, T, q and Θ versus time for particles transported from Europe to the Arctic in winter. The color indicates the relative frequency of the transport time bins, the pluses mark the time of entry to the Arctic, and the asterisks are placed 5 days later. Most particles reach the Arctic very quickly, 25% within 4 days and 50% within 9 days; only 15% take longer than 20 days. From days −5 to 0, the period before the particles are actually leaving the Europe box, p, T and q increase while Θ decreases. This is due to a general subsidence because particles must be below 1000 m at time 0. Particles reaching the Arctic within about 10 days (i.e., the majority), have very low values of Θ (≤275 K) and q (about 2.5 g kg−1) on day 0. Thus a fast passage to the Arctic lower troposphere requires the air to be already cold and dry over Europe.
 The fastest particles, reaching the Arctic within 4 days or less, are very special because they enter the Arctic at low altitudes to meet the 1000-m criterion and then ascend in the Arctic, associated with strong moisture loss. These are blocking-type situations [Iversen and Joranger, 1985] where the Arctic front retreats far north and the particles ascend to above the polar dome north of 70°N but never really enter it. These cases are important, as they allow fast low-level delivery of pollutants from Europe, followed by ascent, precipitation and wet deposition in the Arctic.
 The lowest Θ values (<265 K) are obtained by particles that take about 10–15 days to reach the Arctic. Trajectory plots (not shown) reveal that they travel at low altitudes over north-central Siberia and thus enter the polar dome sideways. Their average Θ decrease shortly before arriving in the Arctic is about 1.5–2 K/day, more than expected for clear-sky radiational cooling. As these particles are low, the extra cooling must be due to contact with the ground.
 Particles reaching the Arctic after 15 days or more are lifted strongly after leaving the Europe box, owing to latent heat release, as the Θ increase and drop in q suggest. This is followed by a long phase of cooling. Note that during the long transport time, individual particles can actually ascend and descend several times and the slow descent and cooling is the result of averaging. These particles, on average, enter the Arctic at high altitudes where they continue to descend and cool. Thus these are cases of vertical descent into the Arctic dome.
 The meteorological parameters along trajectories from the east Asia box into the polar dome (Figure 13) are strikingly different. First of all, only 25% of the particles take less than 14 days, whereas 50% take more than 20 days. Values of Θ at the time of leaving Asia are higher than the corresponding values for the European box, especially when comparing the most frequent bins of 20–30 days for Asia (285 K) and 5–10 days for Europe (275 K). While the few rapid trajectories to the Arctic leave east Asia dry (q < 2 g kg−1), the much more frequent slow trajectories are initially relatively moist (q > 2.5 g kg−1). When arriving in the Arctic, all of the Asian trajectories are very dry, especially the rapid ones. The latent heat release is associated with a strong increase of Θ and decrease of p after leaving Asia, consistent with the fact that emissions from east Asia are subject to strong lifting [Stohl, 2001; Stohl et al., 2002]. The most rapid trajectories enter the Arctic at high altitudes of almost 600 hPa, after which they descend. This means that the uplift is complete before reaching the Arctic, precipitation occurs almost exclusively outside the Arctic and wet deposition in the Arctic itself is unlikely, in contrast to the fast European trajectories.
 The frequent slow (20–30 days) trajectories to the Arctic are also associated with strong initial uplift. Individual particles typically undergo several cycles of upward and downward transport with peak altitudes well above 600 hPa (not shown) which are averaged out in Figure 13 as they are out of phase. Species susceptible to washout would likely be removed completely during the several phases of upward transport.
 Transport routes from North America to the Arctic are faster and the uplift before reaching the Arctic is weaker than from east Asia. Nevertheless, direct low-level transport into the polar dome as well as first-time ascent north of 70°N, both of which are typical for European trajectories, are infrequent or missing also for the North American trajectories (not shown).
 The PES maps for summer (Figure 14) are very different from those for winter (compare with Figures 8 and 9). Transport to the Arctic is slower, such that high PES values remain confined to high latitudes for a longer time. The most striking difference to the winter is that the highest PES values are not found over Siberia but over the oceans. Values over the continental high-emission regions are relatively low, especially for the low-altitude subset (Figure 14, bottom). This seasonal change in transport patterns is caused by the weakening of the Icelandic and Aleutian lows in summer as compared to winter, and the replacement of the wintertime Siberian high by a low-pressure system in summer, due to the seasonally changing land-sea contrasts in surface heating.
 As a result (Figure 15), summer-time anthropogenic BC PSCs are much lower than in winter (Figure 10) both for all Arctic particles, and even more so for the low-altitude subset. For short transport times, the reduction in BC PSCs over winter values is most pronounced for North America and Asia (Figure 15, top), and for longer transport times, it is most pronounced for Europe (Figure 15, middle and bottom). Calculating again the ratio between PSCs integrated over south Asia and Europe, respectively, the corresponding values are 0% (2 days), 1.1% (5 days), 9% (10 days), 39% (20 days) and 72% (30 days) for the low-altitude particle subset, and 0% (2 days), 3.4% (5 days), 19% (10 days), 64% (20 days) and 106% (30 days) for all Arctic particles. For transport times up to 10 days, these ratios are lower even than in winter.
 Using the Lavoué et al.  inventory for July, BC PSC maps were also constructed for boreal forest fires (Figure 16). Even when integrated only over 3 days (Figure 16, top), there is a very strong signal over Siberia, due to the high latitudes of the major burn areas. The BC PSC over Siberia increases further with time and, in addition, a second strong maximum appears over Alaska and northwestern Canada for integrations over 10 (not shown) and 30 (Figure 16, middle and bottom) days.
 The July values from the Lavoué et al.  inventory were used for producing the BC PSC estimates for summer. This overestimates the total summer PSC, as the June and August emissions are lower by 15% and 37%, respectively. Another factor that could lead to an overestimate is that boreal forest fires are favored by droughts when transport to the cold Arctic may be systematically reduced. However, the months in our climatology when there was abnormally strong burning in either Siberia or North America showed only a small and unsystematic reduction of transport into the Arctic compared to the climatology.
 Boreal forest fires are known to sometimes inject emissions at very high altitudes into the atmosphere [Fromm et al., 2005]. FLEXPART's convection scheme was shown to be in principle capable of simulating such high-altitude injections [Damoah et al., 2006]. However, the convection scheme likely underestimates the frequency and severity of such events because the fire processes themselves (e.g., heat release) are not treated in the model. Therefore the 500-m layer may be inappropriate for calculating the PES function and the calculations were repeated for a 8-km-deep layer [Lavoué et al., 2000]. Under this assumption, the BC PSC for the low-altitude Arctic particle subset decreases by about 40% (12%) for 5 (30) days transport time but it increases slightly for the full set of Arctic particles, such that the main effect of assuming a deeper injection layer is a vertical redistribution in the Arctic.
Figure 17 shows the BC PSC integrated over the continental areas for both anthropogenic and forest fire BC emissions as a function of transport time. The by far largest BC PSC comes from Asian (mostly Siberian) forest fires, followed by almost equal European anthropogenic and North American forest fire PSCs. For all Arctic particles, the ratios between total biomass burning and total anthropogenic BC PSCs are 20 (for 2 days), 4 (5 days), 2.2 (10 days), 1.5 (20 days), 1.15 (30 days). Thus ratios depend strongly on the lifetime of BC, but for a BC lifetime less than 10 days, forest fire emissions would dominate the BC in the Arctic in summer.
 In summer, the polar dome is less strongly stratified at low altitudes because surface inversions are weaker than in winter. However, isentropes still slope upward towards the pole, thus preventing isentropic transport into the dome and wind speeds are lower. By examining again individual trajectories from the three continental source boxes to the Arctic, it is found that transport times from Asia are very long, with 50% of the trajectories requiring more than 23 days. Like in winter, Asian air masses descend into the polar dome from above. Therefore it is even more likely than in winter that wet deposition of aerosols and water-soluble gases would be complete before Asian air masses reach the Arctic.
 Trajectories from the European box are initially drier than Asian ones, and the fastest ones can reach the Arctic with relatively little loss of moisture, followed by ascent and moisture loss in the Arctic. This favors wet deposition in the Arctic. The diabatic-surface-cooling pathway from Europe to the Arctic, however, is missing in summer such that the remaining trajectories are descending into the polar dome from above, similar to the Asian trajectories.
 Trajectories from a box over Siberia in the region of frequent boreal forest fires (60°E–140°E, 48°N–66°N) were also examined. Note that most Siberian forest fires emit at low altitudes [Lavoué et al., 2000], making this a viable approach. Owing to the proximity of the burning region to the Arctic, 25% (50%) of the trajectories enter the Arctic within 3 (10) days. These frequent fast trajectories ascend quickly, associated with strong moisture loss, most of which occurs over the Arctic. This is an even more extreme behavior than for fast trajectories from Europe to the Arctic, and it seems that much of the BC emitted by boreal forest fires would be deposited in the Arctic.
4. Discussion and Conclusions
 1. In the literature, the role of transport for causing the winter-time Arctic Haze is discussed ambiguously. On the one hand, it has been realized that there is strong south-north transport over Eurasia [Rahn, 1981; Barrie, 1986]; on the other hand it was also argued that the Arctic lower troposphere is well isolated because the isentropes form a closed dome over the Arctic that cannot easily be penetrated [Carlson, 1981; Iversen, 1984]. According to the latter picture, transport is rather a barrier to pollution inflow into the Arctic, and the main reason for the Arctic Haze is the inefficiency of removal processes, as highlighted in the cartoon shown as Figure 6 of Shaw . This study confirms that in winter the Arctic age of air, and thus the degree of isolation in the central Arctic, increases with decreasing altitude, from about 3 days at 5–8 km to about 1 week near the surface (Figure 3). Maximum values of 10 days occur in the most isolated regions in the North American sector of the Arctic (Figure 2a) where these air masses are also exposed to continuous darkness for 10–14 days in December (8–12 days in January). However, these values are comparable to the timescales of intercontinental pollution transport in the middle latitudes. From the emission in one continent, it typically takes about 1–2 weeks for a pollution plume to arrive over another continent [Stohl, 2001; Stohl et al., 2002; Holzer et al., 2005] if transport in the free troposphere, where wind speeds are fast, is involved. The Arctic troposphere is flushed on about the same relatively short timescale in winter. In summer, in contrast, the corresponding timescale is almost twice as long. While inefficient removal processes are certainly important, seasonal differences in transport alone would also cause a winter-peak of pollution levels, as already noted by Klonecki et al. . Although the polar dome creates a transport barrier, this barrier is permeable in winter as the same process, surface cooling, that builds the polar dome is also active in air masses penetrating it. In particular, air parcels traveling over snow-covered surfaces can cool quickly enough to be transported deep into the polar dome within less than a week. In contrast, in summer the Arctic lower troposphere is indeed quite isolated because diabatic surface-cooling is largely absent.
 2. The Arctic lower troposphere is even more isolated in terms of transport from the stratosphere. In summer, the probability that air in the central Arctic near the surface has been transported from the stratosphere within 10 days (and even within 20 days) is almost negligible. In winter, when deep stratosphere-troposphere-transport is generally more frequent [Sprenger and Wernli, 2003; James et al., 2003], the probabilities are still only about 1% for a timescale of 10 days. There is no direct vertical transport from the stratosphere into the polar dome. Instead, stratospheric air masses slide down the sloping isentropic surfaces at the Arctic (or polar) front into the middle latitudes, to be eventually transported into the polar dome. By the time it reaches the Arctic lower troposphere, the air would be so well mixed with midlatitude tropospheric air, that it would have lost its stratospheric character. Thus the influence of stratospheric air masses on the concentrations of chemical species (e.g., ozone, or nitrogen compounds) near the Arctic surface seems to be marginal.
 3. By analyzing meteorological parameters along trajectories from major pollution source regions into the Arctic lower troposphere, three typical pathways were identified. (1) Rapid (about 4 days, or less) low-level transport into the Arctic followed by uplift at the Arctic front when and where it is located far north. This transport route is possible from densely populated regions in Europe but not from populated regions in North America and Asia, as only Europe is located at high enough latitudes that major emission regions can be north of the polar front. This is also the most frequent transport route from boreal forest fire regions into the Arctic, at least for smoldering fires without significant convection (see Stohl et al. (submitted manuscript, 2006) for a description of such an event). As the uplift and precipitation occurs north of 70°N, this transport route has a strong potential to deposit aerosols and water-soluble pollutants in the Arctic. (2) Low-level transport of already cold air into the polar dome, associated with further diabatic surface cooling, taking about 10–15 days. This pathway can deliver highly polluted air masses to the Arctic because of the flow's low moisture content and stable stratification. This pathway is possible only from European and high-latitude Asian sources as it involves transport over snow-covered high-latitude Siberia. It is absent in summer when the ground is a net source of heat over Eurasia [Klonecki et al., 2003]. (3) Ascent south of the Arctic, often fast and close to the source region, followed either by high-altitude transport or by several cycles of upward and downward transport, and eventually slow descent into the polar dome due to radiational cooling. This is the only frequent transport route from North America and east Asia but it is less frequent from Europe. Wet scavenging outside the Arctic is likely very efficient for this transport route.
 4. The long passage from pollution source regions in North America and, particularly, south and east Asia to the Arctic yields much lower PES values there than over Europe. Assuming a transport time of 10 days, which is longer than the BC lifetime typically reported [Park et al., 2005; Liu et al., 2005; Koch and Hansen, 2005], the south Asian BC PSC near the Arctic surface (in the total column) is about 10% (20–30%) of the European PSC. South Asia's relative contribution would be even lower for shorter BC lifetimes. This is in disagreement with the results of Koch and Hansen , who have used the same emission data but report about equal contributions from Europe and south Asia near the surface and a dominance of south Asian BC in the total column. The difference between the two studies could be explained with a factor of 2–3 longer lifetime of Asian than of European BC. Anthropogenic BC is usually hydrophobic when emitted but becomes quickly hydrophilic in the atmosphere [Seland and Iversen, 1999], with a fixed timescale of 1 day in the model of Koch and Hansen . If Asian BC is lifted more quickly than BC from other source regions, which is not unlikely, more of it can reach the upper troposphere where its lifetime is long. On the other hand, south Asian air masses are initially moister than European ones and thus experience more precipitation en route to the Arctic. Their pathway typically also involves several cycles of upward and downward transport, making precipitation likely also after the hydrophobic-to-hydrophilic conversion of BC. It seems unlikely that the combined effect would be a longer lifetime of BC from south Asia than from Europe. It is more likely that the two models also produce systematically different transport patterns. This might also have implications for the soot climate feedback calculations reported by Hansen and Nazarenko , which used the same model as Koch and Hansen . It should be noted that both this and the study of Koch and Hansen  are based on the emission inventory of Bond et al. , which seems to underestimate Asian BC emissions by about 60% [Park et al., 2005]. However, even when accounting for that, in this study Asian PSCs remain lower than European PSCs.
 5. Arctic concentrations of other species than BC would be dominated even more by European emissions, as BC emissions are most heavily weighted toward south and east Asia. Indeed, in other studies a dominance of European emissions was found for carbon monoxide [Eckhardt et al., 2003; Klonecki et al., 2003; Lamarque and Hess, 2003].
 6. For summer, when the potential for BC radiative effects is largest, this study suggests that BC in the Arctic is dominated by emissions from boreal, mostly Siberian, forest fires. For a transport time of less than 10 days, this source would be comparable to or exceed all anthropogenic contributions combined. During the past decade, annual areas burnt in the boreal forest varied by an order of magnitude [Kasischke et al., 2005], such that BC concentrations and deposition in the Arctic during years of strong burning would be dominated even more by this source. This was indeed observed in the summer of 2004 when North American forest fires enhanced BC levels throughout the Arctic (Stohl et al., submitted manuscript, 2006). Furthermore, there may also be a positive trend in annual areas burnt owing to the warming in the boreal region. The inventory by Lavoué et al.  that was used here, for instance, gives an average annual value of 3.6 million ha burnt in Russia, whereas Kasischke et al.  give an average value of 9.6 million ha for the period 1998–2002. Whether this reflects differences in methodology or a trend is unclear. It emphasizes, however, that BC emissions from boreal forest fires could be higher than assumed here. There could also be an important feedback loop between high-latitude warming, increased frequency and severity of boreal forest fires, BC deposition in the Arctic, enhanced melting of Arctic snow and ice due to the albedo decrease, and yet stronger warming of the high-latitude climate.
 I thank several of my former colleagues at the Technical University of Munich (S. Eckhardt, R. Damoah, P. James, and N. Spichtinger-Rakowsky) for retrieving the ECMWF data used here, and C. Forster, P. Seibert and others for contributing to the development of FLEXPART. BC emission data were obtained from T. Bond and D. Lavoué. I thank D. Lavoué also for sending me a copy of his Ph.D. thesis as well as for discussions. Two reviewers helped with improving the presentation of the results by suggesting a healthy cut in the length of the paper. I appreciate K. Tørseth's support of my activities at NILU.