Dynamics

Abstract. In TROICA (TRanscontinental Observations Into the Chemistry of the Atmosphere) campaigns (1999–2008), the simultaneous observations of near surface 222 Rn concentrations and atmospheric boundary layer thermal structure were performed along the Trans-Siberian Railway across northern Eurasia from Moscow to Vladivostok, including central, southern and far eastern parts of Russia. The data on 222 Rn and temperature vertical distribution are used to estimate 222 Rn regional scale soil fluxes based on calculations of nocturnal 222 Rn accumulation rates in the surface layer under inversion conditions. An effect of seasonal soil thawing on 2–4 times surface 222 Rn concentration increase from summer 1999 to autumn 2005 is observed. The estimated 222 Rn regional averaged fluxes vary over Russia from 29 ± 8 mBq m −2 s −1 in its so-called European territory to 95 ± 51 mBq m −2 s −1 in the southern area of Siberia. The highest 222 Rn fluxes are derived in the regions of high tectonic activity and orogenic belts of central and eastern Siberia and in far eastern Russia. The observed high 222 Rn flux variations in specific events show a strong effect of both soil and atmospheric conditions on 222 Rn near-surface abundance and the derived seasonal patterns over the continent.


Introduction
The radioactive gas radon ( 222 Rn) is one of the decay products of uranium-238 ( 238 U), the most abundant uranium isotope in the earth's crust.The main source of 222 Rn in the atmosphere is soil and its flux depends on the soil type and properties; its only sink is radioactive decay. 222Rn is a chem-ically inert gas with the half-life of 3.82 days.These features allow 222 Rn to be a useful tracer to study air transport (Prospero et al., 1970;Wilkniss et al., 1974, Dörr et al., 1983;Lee and Larsen, 1997) as well as to derive emissions of some atmospheric gases: CH 4 and CO 2 (Dörr et al., 1983;Gaudry et al., 1990;Levin et al., 1999;Moriizumi et al., 1996;Schmidt et al., 1996;Duenas et al., 1999;Biraud et al., 2000;Hirsch, 2007), N 2 O (Biraud at al., 2000;Conen et al., 2002;Messager et al., 2008;Corazza et al., 2011), CO (Messager et al., 2008), and H 2 (Yver et al., 2009). 222Rn is also commonly used for validating transport in climate models (Rasch, 2000;Szegvary et al., 2007), with 222 Rn flux being generally assumed to be spatially uniform with a rate of 1 atom cm 2 s −1 (0.021 Bq m −2 s −1 ) from ice-free land surfaces and zero from oceans (Conen and Robertson, 2002).However, 222 Rn flux varies widely in space and in time.Therefore, the information about spatial and temporal 222 Rn flux variations over a variety of conditions is very important for correct estimation of spatial distribution and strength of natural and anthropogenic sources and sinks of greenhouse gases based on the observations of their near-surface concentrations. 222Rn flux measurements were carried out in different regions of the world (Duenas et al., 1999;Turekian et al., 1977;Somashekarappa et al., 1996;Szegvary et al., 2007;Taguchi et al., 2011), including Russia (Milin et al., 1968;Kirichenko, 1970;Yakovleva, 2003;Tarasov, 2008).However, the data reported for Russia are not sufficient to form a clear picture of 222 Rn flux variations over such a vast territory.

E. V. Berezina et al.: Estimation of nocturnal 222 Rn soil fluxes over Russia
During the last fifteen years the substantial data on 222 Rn spatial variability has been obtained with the use of a mobile carriage laboratory during international TROICA (TRanscontinental Observations Into the Chemistry of the Atmosphere) expeditions along the Trans-Siberian Railway from Moscow to Vladivostok (Elansky et al., 2009).These observations allow studying the large scale variability of near-surface atmospheric composition across extensive areas of the continent with essentially different geological, geographical and climatic features.
Preliminary results of 222 Rn flux estimation from TROICA expeditions are given in Berezina and Elansky (2009) (hereafter BE09).The method used in BE09 to calculate 222 Rn soil flux implies a uniform vertical 222 Rn distribution in a stable 100 m height surface layer based on the observations of 222 Rn vertical distribution in stable atmospheric conditions presented in some investigations (Jacobi and Andre, 1963;Servant, 1966 andKataoka et al., 1998).Although such a simplification seems to be physically reasonable when studying particular events, it could lead to significant and poorly controlled errors in 222 Rn emission calculations when considering such a long-distance route with strongly variable conditions affecting 222 Rn fluxes.
In this paper we present a more elaborate procedure to assess regional-scale 222 Rn fluxes based on simultaneous observations of surface 222 Rn and temperature vertical distribution in the atmospheric boundary layer (ABL) during six TROICA expeditions in 1999-2008.The observational data along with the description of a simple numerical procedure to calculate vertical 222 Rn distribution within the nocturnal stable ABL are presented in Sect. 2. The observed regionalscale surface 222 Rn variability and derived 222 Rn fluxes are discussed in Sect.3. Finally, the general conclusions on the results of this study are formulated in Sect. 4.

Data and methodology
The TROICA observational expeditions have been carried out on a regular basis since 1995 (Elansky et al., 2009) (Table 1).In this study we use the data from six expeditions in which the simultaneous measurements of 222 Rn and vertical temperature profiles in ABL were performed.The complete description of the measurement technique, data quality assessment and the data set obtained from the measurements from the railroad mobile laboratory are presented in Elansky et al. (2009).
The route of the TROICA expeditions overlaid on the radon risk map of Russia (Map of radon risk of Russia, 1996) (see discussion below in Sect.3.1.1)is shown in Fig. 1.The total length of the route from Moscow to Vladivostok (9288 km) is covered for approximately 6 days, so the total duration of a single campaign (forward and return paths) is about two weeks.The strength of 222 Rn natural sources varies strongly along the route due to essentially different ge-ological settings over the territory crossed by the railway.The significant part of the route is located in the mountain areas of the southern Urals and southern part of central and eastern Siberia where the 222 Rn surface fluxes are known to be elevated (see Fig. 1).As the railway runs along the most densely populated and industrial regions of the European part of Russia and southern area of Siberia, the anthropogenic origin of the measured 222 Rn concentrations (uranium deposits exploration and exploitation, mining and chemical combines, coalfired power plants, nuclear power cycle factories) can also be important.We expect, however, that the relative effect of this signal is substantially diminished when inverting the radon flux values since the most of the data in each radon accumulation event is obtained either upwind of possible anthropogenic sources or in low wind synoptic conditions, so the characteristic time of advection from such sources is comparable with the 222 Rn lifetime.

222 Rn measurement technique
Surface 222 Rn concentration was measured by the analyzer of 222 Rn decay products LLRDM (Low Level Radon Daughters Measurement) produced by Tracer Lab (Germany).The air intake of the instrument is placed at the front side of the carriage roof at a 4 m height a.g.l.The measurements are founded in sampling of aerosol-attached radon daughters ( 218 Po, t 1/2 = 3.05 min; 214 Pb, t 1/2 = 26.8min; 214 Bi, t 1/2 = 19.7 min) through the quartz fiber filter ribbon of the instrument using air flow (150 L h −1 ) driven by a pump.An integrated multi-channel analyzer of the LLRDM collects the alpha-spectrum of 222 Rn and its decay products in 256 channels.A microcomputer analyzes the spectrum and directly prints potential alpha energy concentration (PAEC) as well as activity concentration of single nuclides.PAEC describes the resulting α-energy of all decay products in a distinct volume until their total decay and common units are MeV L −1 or working level (1 WL = 130 000 MeV L −1 ).In the case of equilibrium between radon and its progenies, a stationary relation between radon gas concentration and the PAEC exists.The PAEC is often expressed as equivalent equilibrium concentration (EC or EEC).The EC value is in most cases lower than the real radon gas concentration in the air.The relation between EC and the radon gas concentration C Rn is called equilibrium factor F = EC / C Rn .The measured spectra as well as the calculated concentrations are saved in ASCII-files.The measurement range of the instrument is 0.1-100 Bq m −3 .The result is available after 10 min and the instrument systematic errors are about 15 %.
The instrument uses a mathematical calibration method.Therefore, no radon chamber or comparison with other WL (working level) meters is necessary.There is no explicit mathematical formula available because an iterative method is used.The collection and the decay of the filter activity are simulated during the measurement.The microcomputer of the LLRDM integrates in real time the differential equations where k is changed by using of new calibration factors.The activity of the nuclides at the surface of the sampling filter increases during sampling.Therefore, the calibration factor of the instrument changes continuously too.In this paper the radon gas concentration data are analyzed.

Temperature profiles measurements
Vertical temperature profiles were measured with the use of the MTP-5 microwave temperature profiler (ATTEX, Russia) from the level of the carriage roof (4 m a.g.l.) up to the 600 m height (the in situ outdoor temperature measurements at 4 m a.g.l. were also conducted independently by standard meteorological thermometer).The MTP-5 measures the atmospheric thermal radiation in the center of the molecularoxygen absorption band at around 56 GHz at different zenith  angles.The brightness temperature is then retrieved from the measurements (Kadygrov and Pick, 1998) to obtain a vertical temperature profile in a range 0-600 m a.g.l. with 50 m vertical resolution.To minimize the effect of the electric locomotive and the short-term influence of different objects located near the railway on the instrument operation, zenith angle scanning was carried out at a 10 • angle relative to the direction of the motion.The resolution of the retrieved temperature data is 5 min and the overall instrument accuracy is about 0.2 • C. Some relevant parameters of the observed nearsurface inversions (outside the large towns and their suburbs) are summarized in Tables 2 and 4.

Theoretical considerations
In the present study we use a simple numerical procedure to calculate 222 Rn accumulation rates in the stable nocturnal ABL for a number of specific accumulation events observed during the TROICA observations.For each event, we define t 1 the time of the beginning of surface inversion formation and t 2 the time of the observed maximum 222 Rn concentration, with the latter corresponding commonly to the time when the inversion starts to collapse.The time of a particular event varies from 3 to 13 h.Since the typical movement velocity of the mobile laboratory amounts to 50-70 km h −1 , a characteristic spatial scale L for an individual event is within the range of 150-1000 km.Further, it seems to be appropriate to use the following major assumptions: i. during each event 222 Rn surface flux can be set to some constant value representing space and time averaged 222 Rn emission rate over L; ii. at the time of inversion onset t 1 the surface 222 Rn concentration field is assumed to be spatially homogeneous over L; iii. radon vertical transport due to diffusion is limited by the height of the inversion layer; iv.any changes in local 222 Rn concentrations in the nearsurface layer below the inversion due to wind advection can be neglected compared to its vertical transport by eddy diffusivity.
The latter assumption is substantiated by the fact that during the observed strong surface temperature inversions horizontal air movement in ABL is generally very weak, so we do not consider air advection from any particular anthropogenic 222 Rn source and assume the main origin of 222 Rn under the inversion layer to be its soil flux.Hence, temporal evolution of 222 Rn vertical distribution under the inversion layer of the height H allows us to calculate the accumulation rate Q [Bq s −1 ] for its total amount below H , which gives an estimate for 222 Rn soil flux as far as the assumptions (i to iv) hold.In this case, the general problem of atmospheric 222 Rn vertical and temporal variations reduces to the solution of a non-stationary diffusion equation: where is the radon decay constant, and z 0 (= 4 m a.g.l.) is the time independent measurement height at which c 0 ≡ c(z 0 , t ≥ t 1 ) is the known function represented by the actually measured 222 Rn concentrations.The appropriate initial and boundary conditions for Eq. ( 1) are Thus, according to Eq. ( 2) at the start time t 1 , 222 Rn concentration is equal to its value measured prior to the inversion formation and assumed to be uniformly distributed with height due to active daytime vertical mixing.A simple explicit time-forward second order space-centered scheme was used to solve Eqs. ( 1)-( 3) on a 1-D grid with z = 1 m grid spacing between adjacent vertical levels and with 6 s time step to satisfy general stability requirements for a chosen K(z) profile.Once vertical distribution of radon is known, the total 222 Rn abundance M and accumulation rate Q within a layer 0 ≤ z ≤ H at a time t i can be calculated: where the summation is performed over the computational cells and a horizontal bar denotes time averaging.
In the case of strong inversion, the diffusion coefficient K near the earth's surface is known to be very weak, yet being quite variable with z depending on the vertical variations of wind velocity and stability.Following Cohen et al. (1972), we assume in the present study a linear dependence of K on z, with the upper-layer K being independent of height and in the surface layer below 100 m being given as where K(z 1 ) is some known diffusivity rate at a reference level.In our calculations the value of H is set to be constant and was chosen from numerical expeditions to be so high (∼ 600 m) that it does not affect at any appreciable rate the final estimates of radon fluxes.We used the vertical diffusivity profiles from 1 m height above ground given by Jacobi and Andre (1963) (their curves WNW and IWN in Fig. 1) used in the relevant studies on 222 Rn distribution (Beck and Gogolak, 1979;Moses et al., 1960).We derive a plausible range for warm-season K(z 1 ) diffusivities along the TROICA route basing on the corresponding model values at heights 50-100 m a.g.l.from NOAA ARL Archived Meteorology database (http://ready.arl.noaa.gov/READYamet.php).We chose K(z) profiles characteristic of two stability classes of ABL: T 100 > 4.0 • C -extremely stable (G), and T 100 = 1.5-4.0• C -moderately stable (F), according to the common classification of Pasquill (1961) where T 100 is a temperature change in the near-surface 100 m layer.Table 2 shows the surface temperature inversion characteristics from the TROICA data set averaged for different seasons with the strongest positive temperature gradients observed in spring and autumn expeditions (TROICA-8 and 9), owing to anticyclonic weather conditions over the majority of the route.Hence, the selected classes G and F completely cover the range of T 100 values observed during the TROICA expeditions for nocturnal surface inversions.We apply which are also in a good agreement with the results presented in Bezuglaya (1983) for Russian regions and with the average K values in a 90 m depth surface layer proposed in Hosler et al. (1983) for the F stability class.Since a particular value of the diffusivity rate has a first-order influence on the final estimates of 222 Rn fluxes, two series of the calculations with K(z 1 ) value given by Eq. ( 7) were carried out to assess a plausible range of radon soil fluxes for each observational episode.
3   3 and  4, respectively.The spring inversion characteristics were not presented in Table 4 because of a lack of data for spatial averaging.The table shows that in most cases the diurnal mean 222 Rn concentrations are significantly higher for all seasons and regions than the daytime ones due to the nighttime accumulation effect (discussed in detail in Sect.3.1.2.).However, seasonal 222 Rn concentrations spatially averaged for the different studied Russian regions (see Table 3) show high daytime mean radon concentrations in spring for the European territory and Ural region and in autumn for the European territory (ETR) and far eastern Russia region.Furthermore, in the Eastern Siberia and far eastern Russia regions, spring 222 Rn concentrations are higher than the summer ones (about 2 times higher for far eastern Russia region).We suppose that prolonged temperature inversions existed up to midday hours in these observational periods in the regions under consideration (Tables 2 and 4) could contribute to such high daytime radon concentrations.The limited number of the expeditions performed in each season (see Table 1) does not allow us to completely filter out the nighttime accumulation signal from the observational data.We therefore describe here some large-scale features of spatial 222 Rn distribution based on daytime statistics (listed in Table 3) rather than on diurnal means, assuming the former to be more representative of background atmospheric radon levels over the continent.Nevertheless, spring daytime mean radon concentrations in the Ural and Western Siberia regions are also higher than summer ones.We suppose that it can be due to both snow melting in the regions mentioned during the TROICA campaigns, resulting in sharp radon increase in the air and the effect of regional 222 Rn sources.Both factors need to be investigated in detail in a special study.
According to Table 3, the highest daytime mean radon concentrations were observed in the far eastern Russia region (12.4 and 7.3 Bq m −3 in autumn and summer, respectively) and the central Siberia region (7.8 and 6.8 Bq m −3 in autumn and summer, respectively).According to the summer and spring expeditions, low daytime mean 222 Rn concentrations, 2-5 Bq m −3 , are typical for the ETR and Western Siberian regions characterized by flat terrain with low absolute elevations.However, in the autumn 2005 expedition, high daytime mean 222 Rn concentrations, up to 13 and 18 Bq m −3 in the ETR and Western Siberia, respectively, were observed (see Fig. 2 and Table 3).The probable reason of such radon increase is a cumulative effect of two factors: steady anticyclonic conditions with strong and prolonged (up to 16 h) surface temperature inversions and the existence of significant regional 222 Rn sources (mining operations, uranium deposits exposure and geological faults).On the whole, 222 Rn concentrations are higher in autumn compared to other seasons in the studied Russian regions (see Fig. 2 and Table 5).The factors which can determine such seasonal 222 Rn variations will be discussed further in Sects. 3.1.3 and 3.1.4.Table 3 shows that there exists some negative correlation in near-surface radon abundances between the western (ETR to the western Siberia region) and eastern (central and eastern Siberia regions) parts of the continental areas of northern Eurasia.This feature was earlier observed in the seasonal variability of surface air abundances of other trace gases as well (Elansky et al., 2009) and can be most probably connected to a long-wave trough/ridge system that commonly persists over continental areas of northern Eurasia, including during the periods of the TROICA expeditions.
We compared 222 Rn concentrations from the TROICA expeditions with the map of radon risk of Russia (Maximovsky et al., 1996) compiled on the basis of the generalized analytical data on radiogeochemistry, radiometric investigation and other materials obtained from long-term research of different Russian scientific organizations.The authors of the map divided Russian territory into geographical areas according to the degree of radon risk, as shown in Fig. 2. According to the TROICA data, radon concentrations in the areas of elevated radon risk shown on the map are commonly lower than that measured in the dangerous areas (see Table 4 and Fig. 2).The observed high 222 Rn concentrations between Magdagachi and Arkhara (Fig. 2) cover both the radon dangerous area and the "radon Clarke" area (the area where 222 Rn concentration is equal or below its average in the earth's crust) shown on the map of radon risk immediately to the west, which is likely to be due to the prevailing effect of the local observation times as discussed above.Generally, 222 Rn concentrations measured in the TROICA expeditions (Fig. 2) are found to be in a good agreement with the earlier studies on the radon risk areas (Maximovsky et al., 1996) as well as spatial locations of tectonic faults, which confirms our general notion about the reliability of the obtained 222 Rn data set and its applicability to invert radon soil fluxes at a regional basis.

Effect of the atmospheric stability on surface 222 Rn concentration
In TROICA expeditions, the highest 222 Rn concentrations (up to 75 Bq m −3 ) were commonly observed during the nights with strong and prolonged surface temperature inversions.Figure 3 shows the mean diurnal cycles of temperature inversion height and 222 Rn concentration in different seasons.The surface temperature inversions existed usually from 18:00-19:00 to 06:00-08:00 LT and from 17:00 to 09:00-10:00 LT in the warm and cold seasons, respectively (Fig. 3a).The highest radon concentrations, up to 30-35 Bq m −3 , were observed in the early morning (04:00-06:00 LT), being a result of nighttime accumulation below temperature inversion, prior to the beginning of inversion collapse and subsequent decrease in 222 Rn concentration as a factor of 3 to 5 on average owing to convective mixing.During the day there was an absence of significant temperature inversions (no nighttime near-surface radon accumulation episodes), meaning concentration did not change significantly during the day and for all seasons was 1.5-3.5 Bq m −3 (Fig. 3b with the caption "no inversions").Table 5 presents diurnal and daytime mean 222 Rn concentrations for the different seasons according to the TROICA measurements.The measurements performed under daytime inversion conditions were excluded from the present data to suppress the strong effect of the associated radon accumulation on the derived statistics, which resulted in daily mean 222 Rn concentrations being 1.5-2 times lower on average compared to the diurnal ones in all seasons.The highest diurnal and daytime mean 222 Rn concentrations were observed in autumn, owing to the strongest and most prolonged temperature inversions observed in this period (see Tables 2 and  4), which confirms significant influence of vertical exchange rates on surface 222 Rn variations at a seasonal scale.

Seasonal soil thawing effect on surface 222 Rn concentration
Along with vertical exchange due to the turbulent mixing, the soil (and its specific properties) is the other key factor affecting 222 Rn near-surface abundance.The soil covered with snow or ice accumulates 222 Rn, explaining its subsequent enhanced emission into the atmosphere during the first hours after snow melting (Miklyaev and Petrova, 2006).
Commonly, the diffusion equilibrium between the soil and the surface atmospheric layer is reached in several hours, after which the radon flux attains its steady-state value; however, sometimes this process can last up to several days.Glover (2006) and Glover and Blouin (2007) note that the permafrost is a barrier to 222 Rn exhalation, resulting in its 80-90 % decrease in ambient air and 10-15 times increase in its abundance in the soil.Since the major part of the Trans-Siberian Railway in Eastern Siberia goes through the permafrost area, the influence of seasonal soil thawing should be accounted for when studying seasonal aspects of the 222 Rn surface flux variations.The thawing depth was calculated in the region 52-55 • N, 105-130 • E at the time periods of the TROICA campaigns using the scheme of the heat and moisture transfer in the soil (Arzhanov et al., 2008) in the ECHAM5/MPI-OM model (SRES A1B scenario).The resulting effect of the thawing depth on the near-surface radon abundance is shown in Fig. 4. The model-predicted thaw- ing depth is approximately 1.24, 1.40 and 1.85 m for the summer TROICA-5,7,11, summer TROICA-12, and October TROICA-9 campaigns, respectively.One can see from the figure that near-surface radon concentrations increased more than 3 times (according to the daytime radon values) in this region from summer 1999 (TROICA-5) to summer 2008 (TROICA-12), reaching the highest value in autumn 2005 (TROICA-9), with the persistent increase in thawing depth being observed.To exclude the effect of the nighttime radon accumulation events, we divided nighttime and daytime data (see Fig. 4).Daytime mean 222 Rn concentrations (white squares in Fig. 4) are calculated for no temperature inversion conditions.

Nocturnal 222 Rn soil flux calculation
We use the measured 222 Rn concentrations in nocturnal accumulation events to estimate associated radon surface fluxes using the numerical approach discussed in Sect.2.3.An example of 222 Rn flux calculation at the route part 1256-1076 km from Moscow 10 July 2001, 02:54-06:10 LT (TROICA-7), is presented in Fig. 6.The observed region is located in a flat area with a typical elevation from 150-200 m a.s.l.The figure shows the time series of the atmospheric temperatures at different heights a.g.l., the measured radon concentration, and the calculated total radon content varying approximately linear with time.Invoking Eq. ( 6), the regression slope of M on t gives the mean radon emission rate, which is an approximate estimate for Q.
In a particular nighttime accumulation event the atmospheric transport conditions within the surface inversion layer vary both with time and altitude.Hence, its resulting effect on radon accumulation rate can hardly be quantified at a rational basis, taking into account the lack of observational  data on the full set of parameters governing the turbulent mixing regime.In present simulations the major factor affecting the radon vertical distribution, and hence accumulation rate, is the vertical mixing rate profile controlled by the parameter K(z 1 ).Since the exact value of the temperature gradient in a particular inversion event changes within a range of G and F stability classes, two sets of calculations were performed by setting K(z 1 ) equal to 10 and 100 cm 2 s −1 according to Eq. ( 7) to obtain Q(G) and Q(F) values for radon accumulation rates for G and F stability classes, respectively.Accordingly, for each accumulation event i we define as the best estimates for Q and an estimate error for Q i , respectively.The relative estimated error is commonly a few tens of percent and reaches as much as 50 % in some events.To make our estimates be representative at a regional scale, we calculate the expected means and associated errors as and respectively where g i = σ −1 Q,i , and summation by i is performed over all accumulation events observed during the TROICA expeditions within a particular region defined according to Fig. 2. The calculated weighted-mean region averaged radon soil fluxes are summarized in Fig. 6 and Table 6.One can see that the derived 222 Rn soil flux varies significantly over Russia, from 29 ± 8 mBq m −2 s −1 to 95 ± 51 mBq m m −2 s −1 , depending on the geological features as well as the seasons.The highest 222 Rn fluxes are derived in the mountain regions of central Siberia, eastern Siberia and far eastern Russia.In these Russian regions radon soil emissions are 1.5-3 times higher than in the plains (Table 6).In spring the weighted-mean region averaged 222 Rn To observe an effect of the seasonal soil thawing on 222 Rn soil flux (see Sect. 3.1.3),we calculated the weighted mean 222 Rn flux for each expedition presented in Fig. 4. The estimated values do not show a significant increase from the summer expeditions to the autumn one.However, it is not correct to compare the mean 222 Rn fluxes calculated for each expedition because the 222 Rn flux episodes in the expeditions under consideration can differ geographically.So we chose the summer episodes (from all summer expeditions) corresponding to the autumn episodes (in the same regions) and derived some increase in autumn weighted mean 222 Rn flux (55.7 ± 2.1 mBq m −2 s −1 ) compared to the summer one (50.1 ± 2.8 mBq m −2 s −1 ).The similar increase in both the seasonal soil thawing and 222 Rn concentrations is shown in  2010) are usually 4-5 times lower than the ones from TROICA measurements.The Russian 222 Rn flux map (http://radon.unibas.ch)generated by the scientific group from the University of Basel, Switzerland (Szegvary et al., 2007), and based on a gamma-dose rate map of Russia derived from aeroradiometric measurements (Map of natural gamma radiation doses of Russia, 1996) shows the highest radon fluxes in the studied regions of central and eastern Siberia; their estimations for far eastern Russia region correspond to ours.But according to the map, 222 Rn flux varies over Russia from 0.2 to 1.2 atom cm 2 s −1 (from 4 to 24 mBq m −2 s −1 ) which is 3-7 times lower than the range we inferred.It is possible that the wide range of the K(z) values used for radon flux calculation could result in the overestimation of the mean radon flux values.
We also compared our results with the direct 222 Rn flux measurements reported for some Russian regions.According to the Perm CGMS radiation monitoring, in 2006 the mean 222 Rn flux was 40 ± 10 mBq m −2 s −1 in Perm region (http://wp.permecology.ru/report/report2006/17.html) which corresponds to 40 ± 20 mBq m −2 s −1 on average from the TROICA expeditions. 222Rn soil fluxes in the city of Krasnoyarsk and its suburb Minusinsk are reported to vary from 14 to 20 mBq m −2 s −1 and from 9 to 60 mBq m −2 s −1 , respectively (Voevodin and Kurguz, 2012;Sobyanina et al., 2012), being 40 mBq m −2 s −1 on average.These measurements are in a good agreement with our estimations for the Krasnoyarsk area (40 ± 30 mBq m −2 s −1 ).Kirichenko (1970) reported 222 Rn flux in the southern Ural region to be 11 mBq m −2 s −1 from the atmospheric radon profile studies in summer, which is lower than our estimations, 30-70 mBq m −2 s −1 .The same author reported the mean 222 Rn flux in the European territory (Leningrad, Moscow, Kaluga city areas) in summer to be 7 mBq m −2 s −1 .The mean 222 Rn www.atmos-chem-phys.net/13/11695/2013/flux from the summer-autumn 222 Rn flux measurements in the Fyodorovskoe Forest Reserve, Tver region (the European part of Russia), presented in Levin et al. (2002) is 50-130 mBq m −2 s −1 .According to Milin et al. (1967), 222 Rn flux in Moscow region in summer is 38 mBq m −2 s −1 .Our estimations for the European territory of Russia (Moscow-Perm region) give the mean 222 Rn flux to be 29 ± 8 and 62 ± 31 mBq m −2 s −1 in summer and autumn, respectively.Milin et al. (1967) reported the mean 222 Rn flux from summer measurements in Kirov to be 15 mBq m −2 s −1 , which is in agreement with our calculations (20 ± 10 mBq m −2 s −1 ).Miklyaev and Petrova (2006) measured 222 Rn flux at different sites in Moscow and reported that 222 Rn flux varies from 5 to 72 mBq m −2 s −1 (21 ± 12 mBq m −2 s −1 on average) and from 4 to 264 mBq m −2 s −1 (38.6 ± 34.4 mBq m −2 s −1 on average) in the regions with sandy and clay soils, respectively. 222Rn fluxes calculated from the observations on the mobile laboratory around Moscow (TROICA-10, 4-7 October 2006) at two observational parts of the Moscow region: (1) from Zhilino to Voskresensk, where sandy soils are spread, are 13 ± 12 mBq m −2 s −1 ; (2) from Jaganovo -Orekhovo-Zuyevo -Dmitrov, where both clay and sandy soils are spread, are 27 ± 20 mBq m −2 s −1 .

Conclusions
The most significant variations in surface radon concentrations along the Trans-Siberian Railway are caused by the diurnal change in the ABL stability.The highest 222 Rn concentrations (up to 75 Bq m −3 ) were usually observed during nighttime strong and prolonged temperature inversions in the mountain regions of Russia (central and eastern Siberia and far eastern Russia regions).Due to weak vertical mixing in the stable atmosphere, 222 Rn accumulates in ASL and its concentrations increased several times compared to its values during unstable atmospheric conditions.If we know the rate of 222 Rn accumulation in the nighttime stable ABL and the height of its mixing layer, we can estimate nocturnal radon soil flux.
The nocturnal 222 Rn soil flux calculated from the simultaneous measurements of 222 Rn and vertical temperature profiles in ABL in six expeditions from Moscow to Vladivostok varies over continental Russia from 29 ± 8 mBq m −2 s −1 to 95 ± 51 mBq m m −2 s −1 , depending on the geological features as well as the seasons.The highest 222 Rn soil flux values are derived for the mountain regions of central and eastern Siberia and far eastern Russia.Generally, 222 Rn concentration and flux over Russia peak in autumn and bottom out in spring.We suppose that there is a contribution of seasonal soil thawing to high radon concentrations and fluxes in the permafrost regions in autumn.However, there are some episodes with high 222 Rn concentrations (Ural and eastern Siberia regions) and fluxes (Ural, central Siberia and far eastern Russia regions) in spring.We suppose that possible snow melting in these regions during the spring TROICA campaigns or/and air advection from local and regional radon sources could result in sharp radon increase in the air.Further detailed investigations are required.
It is possible that the wide range of the K(z) values used for radon flux calculation could result in the overestimation of the mean radon flux values.As a whole, 222 Rn fluxes estimated from the expeditions on the mobile laboratory are in agreement with the direct measurements reported for Russian regions in literature.
The results presented in this paper can be important to investigate and document in detail the trends in fluxes of N 2 O, CO 2 , and CH 4 during the coming decades of global warming in the late Holocene or so-called mid-Anthropocene.

Fig. 1 .
Fig. 1.Map of radon risk of Russia and the TROICA expeditions route along the Trans-Siberian Railway from Moscow to Vladivostok.

Figure 2
Figure2shows the spatial distribution of original 10 min mean222 Rn concentrations and 10th, 50th and 90th percentiles calculated for 100 km parts of the route.For spring and autumn data only the 50th percentile values are considered because of a limited data set for calculations.The figure also presents the altitude a.s.l.along the Trans-Siberian Railway to demonstrate an importance of terrain elevation in the observed radon distribution.For the regional-scale representation, we divided the Russian territory along the Trans-Siberian Railway into 6 regions according to their basic geological features: ETR or European territory of Russia (Moscow-Perm), Ural (Perm-Ekaterinburg), western Siberia (Ekaterinburg-Novosibirsk), central Siberia (Novosibirsk-Irkutsk), eastern Siberia (Irkutsk-Belogorsk) and far eastern

Fig. 2 .
Fig. 2. Spatial distribution of surface 222 Rn concentration (10 min average values) and altitude a.s.l.from Moscow to Vladivostok in TROICA expeditions.The radon risk areas corresponding to the map of radon risk of Russia are presented as shaded rectangles.Percentiles are presented for 10 min radon values for each 100 km route part (for spring and autumn data only the 50th percentile values are presented because of a limited data set for calculations).

Fig. 4 .
Fig. 4. Mean 222 Rn concentrations obtained from the TROICA data and the soil thawing depth calculated for the period of the expeditions in the region 52-55 • N, 105-130 • E using ECHAM5/MPI-OM model (SRES A1B scenario).

Fig. 5 .
Fig. 5. Nocturnal 222 Rn flux calculation in stable atmospheric conditions at the route part 1256-1076 km from Vladivostok to Moscow (the East European region) 10 July 2001, 02:54-06:10 LT (the TROICA-7 expedition).Slope defines 222 Rn flux in this region.Glazov is the largest locality on the presented route.

Fig. 6 .
Fig. 6. 222 Rn soil fluxes over Russia calculated from the spring (a), summer (b) and autumn (c) expeditions.The length of each rectangle corresponds to the route part (in km) for which 222 Rn flux was estimated.The joined arrows indicate the strong faults crossing the route.

Fig. 4 .
It can confirm the possible influence of the seasonal soil thawing on 222 Rn soil exhalation.We compared 222 Rn fluxes calculated from the TROICA expeditions with the 222 Rn flux maps derived from the data modeling.Schery and Wasiolek (1998) proposed a global 222 Rn flux map based on a porous media transport theory and calibrated them with expedition 222 Rn flux data from Australia and Hawaii.The map gives 222 Rn flux for the Russian latitudes to be about 20-30 mBq m −2 s −1 but has a large uncertainty because of the lack of global data on soil moisture and 226 Ra content.Hirao et al. (2010) improved the performance of the model complimented by the soil and 226 Ra content and estimated the global 222 Rn flux density distribution for the period 1979-2007.These estimations give 222 Rn flux for Russian regions to be up to 30 mBq m −2 s −1 .The 222 Rn flux estimations by Schery and Wasiolek (1998) and Hirao et al. (

Table 2 .
Surface temperature inversion characteristics averaged in different seasons.

Table 4 .
Surface temperature inversion characteristics averaged for the studied Russian regions from the summer and autumn expeditions.