Intercomparison study of atmospheric 222Rn and 222Rn progeny monitors

Abstract. The use of the noble gas radon (222Rn) as a tracer for different research studies, for example observation-based estimation of greenhouse gas (GHG) fluxes, has led to the need of high-quality 222Rn activity concentration observations with high spatial and temporal resolution. So far a robust metrology chain for these measurements is not yet available. A portable direct atmospheric radon monitor (ARMON), based on electrostatic
collection of 218Po, is now running at Spanish stations. This monitor has not yet been compared with other 222Rn and 222Rn
progeny monitors commonly used at atmospheric stations. A 3-month intercomparison campaign of atmospheric 222Rn and 222Rn
progeny monitors based on different measurement techniques was realized
during the fall and winter of 2016–2017 to evaluate (i) calibration and
correction factors between monitors necessary to harmonize the atmospheric
radon observations and (ii) the dependence of each monitor's response in
relation to the sampling height and meteorological and atmospheric aerosol
conditions. Results of this study have shown the following. (i) All monitors were able to
reproduce the atmospheric radon variability on a daily basis. (ii) Linear
regression fits between the monitors exhibited slopes, representing the
correction factors, between 0.62 and 1.17 and offsets ranging between −0.85 and −0.23 Bq m−3 when sampling 2 m above ground level
(a.g.l.). Corresponding results at 100 m a.g.l. exhibited slopes of 0.94 and 1.03 with offsets of −0.13 and 0.01 Bq m−3, respectively.
(iii) No influence of atmospheric temperature and relative humidity on monitor responses was observed for unsaturated conditions at 100 m a.g.l.,
whereas slight influences (order of 10−2) of ambient temperature were
observed at 2 m a.g.l. (iv) Changes in the ratio between 222Rn progeny and 222Rn monitor responses were observed under very low atmospheric aerosol concentrations. Results also show that the new ARMON could be useful at atmospheric radon
monitoring stations with space restrictions or as a mobile reference
instrument to calibrate in situ 222Rn progeny monitors and fixed radon monitors. In the near future a long-term comparison study between ARMON, HRM, and ANSTO monitors would be useful to better evaluate (i) the uncertainties of radon measurements in the range of a few hundred millibecquerels per cubic meter to a few becquerels per cubic meter and (ii) the response time correction of the ANSTO monitor for
representing fast changes in the ambient radon concentrations.


HRM, and ANSTO monitors would be useful to better evaluate (i) the uncertainties of radon measurements in the range of a few hundred millibecquerels per cubic meter to a few becquerels per cubic meter and (ii) the response time correction of the ANSTO monitor for representing fast changes in the ambient radon concentrations.

Introduction
Over continents, the natural radioactive noble gas radon ( 222 Rn) (half-life T 1/2 = 3.8 d) is continuously generated within the soil from the decay of radium ( 226 Ra) (Nazaroff and Nero, 1988), and it can then escape into the atmosphere by diffusion, depending on soil characteristics and meteorological conditions (Grossi et al., 2011;López-Coto et al., 2013;Karstens et al., 2015). The global 222 Rn source into the atmosphere is mainly restricted to land surfaces (Szegvary et al., 2009;Karstens et al., 2015), with the 222 Rn flux from water surfaces considered negligible for most applications (Schery and Huang, 2004).
The diversity of these three aforementioned measurement techniques could introduce biases or compatibility issues that would limit the comparability of the results obtained by independent studies and the subsequent application of atmospheric radon data for regional to global investigations (e.g., Schmithüsen et al., 2017). Thus, a comparative assessment of all the experimental techniques applied for atmospheric 222 Rn activity concentration measurements and a harmonization of their datasets are needed, as suggested by the International Atomic Energy Agency (IAEA, 2012). Xia et al. (2010) carried out a comparison of the response of a dual-flow-loop two-filter detector from the Australian Nuclear Science and Technology Organisation (ANSTO, Whittlestone and Zahorowski 1998) and a one-filter monitor (α/β Monitor P3) manufactured by the Bundesamt für Strahlenschutz, Germany (BfS) (Stockburger and Sittkus, 1966), for atmospheric 222 Rn measurements under various meteorological conditions at 2.5 m above ground level (a.g.l.) over 1 year. Their results showed that both systems followed the same patterns and produced very similar results most of the time, except under specific meteorological conditions such as when precipitation or the proximity of the forest canopy could remove short-lived progeny from the air mass to be measured by the one-filter monitor. However, Xia et al. (2010) did not find a clear relationship between precipitation intensity and the ratio between progeny-derived 222 Rn and 222 Rn activity concentration to convert the progeny signal to 222 Rn activity concentration. Grossi et al. (2016) presented results from two short (about 7-9 d) comparisons between a one-filter monitor from Heidelberg University (Heidelberg Radon Monitor, HRM; Levin et al., 2002) and an atmospheric radon monitor (ARMON; Grossi et al., 2012), an electrostatic deposition monitor from the Universitat Politècnica de Catalunya (UPC). The two comparison campaigns were carried out at a coastal and a mountain site, with sampling in both cases from 10 m a.g.l. These comparisons revealed that the responses of both monitors were in agreement except for water-saturated atmospheric conditions or periods of rainfall. Again, the quantity of comparison data was not sufficient to confirm any statistical correlation.
Loss of aerosols in the air intake systems can also complicate the derivation of 222 Rn activity concentrations from one-filter systems such as the HRM. Levin et al. (2017) carried out an assessment of 222 Rn progeny loss in long tubing by laboratory and field experiments. Results of these experiments, for 8.2 mm inner diameter (i.d.) Decabon tubing, gave an empirical correction function for 222 Rn progeny measurements, which enables the correction of measurements for this Atmos. Meas. Tech., 13, 2241-2255, 2020 www.atmos-meas-tech.net/13/2241/2020/ specific experimental setup (e.g., tubing type and diameter, flow rate, aerosol size distribution). Finally, Schmithüsen et al. (2017) conducted an extensive European-wide 222 Rn/ 222 Rn progeny comparison study in order to evaluate the comparative performance of one-filter and two-filter measurement systems, determining potential systematic biases between them and estimating correction factors that could be applied to harmonize 222 Rn activity concentration estimates for their use as a tracer in various atmospheric applications. In this case, the authors employed a HRM monitor as the reference device. It was taken to nine European measurement stations to run for at least 1 month at each of them. This monitor was run in parallel to other onefilter and two-filter radon monitors operating at each station of interest.
Although several intercomparison campaigns have been carried out in the past, none of them have included simultaneous observations from one-filter, two-filter, and electrostatic deposition methods. Here, we present the results of a 3month intercomparison campaign carried out in the fall and winter of 2016-2017 in Gif-sur-Yvette (France) where, for the first time, co-located measurements from monitors based on the three measurement principles were included. Two two-filter 222 Rn monitors, two single-filter 222 Rn progeny monitors, and an electrodeposition monitor were run simultaneously under different meteorological and aerosol conditions sampling from heights of 2 and 100 m a.g.l.
The main objectives of the present study were to (i) compare the calibration and correction factors between all monitors required to derive harmonized atmospheric radon activity concentrations and (ii) analyze the influence that meteorological and environmental parameters, as well as sampling height, can have on the final determined 222 Rn activity concentration.
In the present paper the applied methodology is reported, including a short presentation of the 222 Rn/ 222 Rn progeny monitors participating in the campaigns, the sampling sites, and the statistical analysis carried out. Finally, the outcomes of the present study are discussed and compared with the ones from Schmithüsen et al. (2017).

Methods
In Sect. 2.1 a short description is given of the monitors compared in the experiment, mainly focusing on measurement techniques, instrument calibration, and maintenance. The main characteristics of these monitors are then summarized in Table 1. Section 2.2 presents the French atmospheric stations of Orme de Mérisiers (ODM) and Saclay (SAC) where the two phases of the intercomparison campaign were realized. Section 2.3 briefly describes the devices used to measure the environmental parameters and the atmospheric aerosol concentration at the above sites during the experi-ments. Finally, the statistical analysis applied is described in Sect. 2.4.

Dual-flow-loop two-filter detectors
The two 1500 L dual-flow-loop two-filter detectors included in this exercise were designed and built at the Australian Nuclear Science and Technology Organisation (ANSTO). This model of detector, which will henceforth be named ANSTO, is based on a previous design by Thomas and Leclare (1970), with some early iterations of the modified design being described by Whittlestone and Zahorowski (1998) and Brunke et al. (2002). The subsequent evolution of two-filter detectors in recent decades, and the current principle of operation, has been described in detail by Williams and Chambers (2016) and Griffiths et al. (2016).
During the measurement campaign ambient air was sampled continuously at a rate of about 83 L min −1 through a 50 mm i.d. HDPE inlet tube and a 400 L delay volume to allow decay of the short-lived 220 Rn (T 1/2 = 56 s). The air stream then passes through the first filter, which removes all ambient aerosols as well as 222 Rn and 220 Rn progeny. The filtered sample, now containing only aerosol-free air and 222 Rn gas, enters the main delay volume (1500 L) where 222 Rn decay produces new progeny. The newly formed 218 Po and 214 Po are then collected on a second filter and their subsequent α decays are counted with a ZnS photomultiplier system. Atmospheric 222 Rn activity concentrations are then calculated from the α count rate and the flow rate through the chamber.
The detection limit (L D ) of two-filter detectors is directly related to the volume of the main delay chamber. Here, L D is understood to represent the ambient radon concentration at which the estimated counting error of the instrument reaches 30 %. The L D of the 1500 L model used in this study was around 0.03 Bq m −3 . Under normal operation ANSTO monitors are automatically calibrated in situ every month by injecting radon into the sampling air stream from a wellcharacterized Pylon 226 Ra source (ca. 41 kBq radium at SAC station) for 5 h at a fixed flow rate of ∼ 100 cc min −1 . Automatic instrumental background checks, each lasting 24 h, are also performed every 3 months to keep track of long-lived 210 Pb accumulation on the detectors' second filter (which should be changed every 5 years). Based on a calibration source uncertainty of 4 %, coefficient of variability of valid monthly calibrations of 2 %-6 %, and counting uncertainty of around 2 % for radon concentrations ≥ 1 Bq m −3 , the total measurement uncertainty of 1500 L ANSTO radon detectors is typically between 8 % and 12 % (k = 2). The ANSTO monitors have low maintenance requirements, but due to their dimensions (2.5-3 m long) it can be challenging to in-  (2000) ∼ 8 -Sampling inlet height correction -Need of large pump stall them at stations with space restrictions. As an alternative to the 1500 L detectors, a 700 L model is also available, which is more portable and has a L D of 0.04-0.05 Bq m −3 . The combination of detector volume, operating flow rate, and radon decay chain results in ANSTO monitors having a response time of ∼ 45 min, which can be corrected for in postprocessing (Griffiths et al., 2016). Two ANSTO monitors were used during this study. As explained later in the text these monitors are permanently running at SAC and ODM stations. No calibration source was available when the ANSTO monitor was installed at the ODM site, so calibration and background information derived prior to transport have been used.

Electrostatic deposition monitor
The atmospheric radon monitor (ARMON) used in this experiment was designed and built at the Institut de Tècniques Energètiques (INTE) of the UPC. The ARMON is a portable instrument based on the electrostatic deposition method, consisting of alpha spectrometry of positive ions of 218 Po electrostatically collected on a detector (Hopke, 1989;Pereira and da Silva, 1989;Tositti et al., 2002). The ARMON is described in detail in Grossi et al. (2012).
Sampled air with a flow rate between 1 and 2 L min −1 is first filtered to remove ambient 222 Rn and 220 Rn progeny and then pumped through a ∼ 20 L spherical detection volume uniformly covered internally with silver. Within this volume the newly formed 222 Rn and 220 Rn progeny, i.e., positive 218 Po and 216 Po ions, respectively, are electrostatically collected on a passivated implanted planar silicon (PIPS) detector surface by an electrostatic field inside the spherical volume. An 8 kV potential is applied between the PIPS detector base and the sphere walls. As for the ANSTO detector, the sensitivity of this instrument type depends on the detector volume. The design of the monitor employed in this study has a L D of about 0.07 Bq m −3 in agreement with the definition given above. Grossi et al. (2012) reported a minimum detection limit for this instrument of around 0.2 Bq m −3 in agreement with the definition of Gilmore (2008). The measurement efficiency of the electrodeposition method is reduced due to neutralization of the positive 218 Po in recombination with OH − ions in the sampled air (Hopke, 1989). Consequently, it is necessary to dry the sampled air as much as possible before it enters the detection volume. To this end, a dew point of < −40 • C was maintained at both intercomparison sites using a cryocooler, consisting of a vessel tube where sampling air was passing through before reaching the radon monitor (Grossi et al., 2018).
Each ARMON is calibrated at the INTE-UPC 222 Rn chamber (Vargas et al., 2004) under different 222 Rn and relative humidity conditions (Grossi et al., 2012). The radon chamber of the INTE-UPC is a 20 m 3 installation, which allows control of the exhalation rate (0-256 Bq min −1 ) and the ventilation air flow rate (0-100 L min −1 ). The 222 Rn source is a dry powder material containing 2100 kBq 226 Ra activity enclosed in the source container (RN-1025 model manufactured by Pylon Electronics). The calibration factor F cal of Atmos. Meas. Tech., 13, 2241-2255, 2020 www.atmos-meas-tech.net/13/2241/2020/ the ARMON used in this study was 0.39 counts per minute (cpm) per becquerel per cubic meter with an uncertainty of 10 % (k = 2). The correction factor for the humidity influence inside the sphere was 6.5 × 10 −5 per part per million H 2 O (ppm) with a maximum uncertainty of 10 % (k = 2). The total uncertainty of the atmospheric radon activity concentration measured by the ARMON is about 20 % (k = 2) for atmospheric 222 Rn levels in the range of a few hundred becquerels per cubic meter but could increase up to 30 %-35 % (k = 2) when atmospheric 222 Rn levels decrease to a few becquerels per cubic meter due to the increase in the error of the alpha counts. The total uncertainty includes the calibration factor F cal , the background due to the presence of 212 Po from 220 Rn, the net 218 Po counts, and the humidity correction factor (Grossi et al., 2012;Vargas et al., 2015). Every 1-2 years the progeny filter at the ARMON inlet should be changed. The detection volume of the ARMON is safely isolated because it is located within an external wooden cube of 0.18 m 3 .

One-filter monitors
One-filter detectors measure the decay rates of aerosolbound 222 Rn progeny directly accumulated by air filtration . The 222 Rn activity concentration is then calculated assuming a constant disequilibrium factor (F eq ) for a given site and sampling height between 222 Rn and the measured progeny in the sampled air.
In the present study two monitors based on this method were used. One, named HRM here, was developed at the Institute of Environmental Physics of Heidelberg University, Germany, and is described in detail by Levin et al. (2002). Rosenfeld (2010) describe the most recent version of this monitor for which the electronics, data acquisition, and evaluation hardware and software were modernized. The HRM measurement is based on α spectrometry of 222 Rn daughters attached to atmospheric aerosols collected on a static quartz fiber filter (QMA ∅ 47 mm) using a surface barrier detector (Canberra CAM 900 mm 2 active surface). The L D of the HRM is about 0.07 Bq m −3 at a flow rate of about 20 L min −1 with an uncertainty smaller than 15 % (k = 2) for atmospheric 222 Rn levels above 2 Bq m −3 . This includes the uncertainty of the line loss correction (see below). Since onefilter detectors have no need for any delay chambers but use only a compact filter holder with an integrated detector and pre-amplifier, the HRM is a small instrument with high portability. Regarding maintenance requirements, the quartz fiber filter should be changed monthly.
During the measurement campaign carried out at the Saclay station, where air samples were collected via a 100 m Decabon tubing (see below), the line loss correction of Levin et al. (2017) was applied to all data of the HRM. No loss of aerosol was assumed in the short tubing used at Orme de Mérisiers station. Here we report 214 Po activity concentrations for both sites. However, for the 100 m intake height at Saclay we would not expect any disequilibrium, meaning that, based on the results from Schmithüsen et al. (2017), the reported 214 Po activity concentrations directly correspond to 222 Rn activity concentrations. By contrast, for the 2 m intake height at ODM we expect a 214 Po/ 222 Rn disequilibrium of about 0.85 to 0.9.
The second type of one-filter monitor participating in this study was built at the Laboratoire des Sciences du Climat et de l'Environnement, LSCE, France (Polian, 1986;Biraud, 2000;Schmithüsen et al., 2017). Within this paper this monitor will be called the LSCE monitor. This monitor uses a moving filter band system, which allows the determination of atmospheric 222 Rn activity concentration based on measurements of its progeny 218 Po and 214 Po. Attached 222 Rn progeny are collected on a cellulose filter (Pöllman-Schneider) over a 1 h period at a flow rate of 160 L min −1 and after this aerosol sampling period, the loaded filter is moved to the α spectrometry for a 1 h measurement period by a scintillator from Harshaw Company Co. and photomultiplier from EMI Electronics Ltd. (Biraud, 2000). The L D is about 0.01 Bq m −3 with an uncertainty of about 20 %.
Regarding maintenance on a regular basis, the LSCE monitor's filter roll has to be changed every 3 weeks. Automatic detector background is performed every 3 weeks, and counting efficiency is manually tested with an americium source. The instrument is designed to measure radioactive aerosols a few meters above the ground level. An inlet filter is installed to block black carbon or dirt deposition when the instrument is installed in urban areas as the flow rate drops below 9 m 3 h −1 . The instrument is about 25 cm high, 40 cm long, and 25 cm deep, and it can be easily deployed at a station.

Sites
The present intercomparison study was carried out at two stations located 30 km southwest of Paris in the fall and winter of 2016-2017 (Fig. 1). Both stations, 3.5 km apart, belong to the LSCE and are located in a region with a radon flux of ca. 5-10 mBq m −2 s −1 in winter, according to output of the Karsten et al. (2015) model.
Phase I of the measurements started at Orme des Mérisiers (ODM, latitude 48.698, longitude 2.146, 167 m above sea level) and ran between 25 November 2016 and 23 January 2017. Here, LSCE and ANSTO (for convenience named ANSTO_ODM here) monitors are routinely running. During Phase I of the intercomparison exercise these two monitors were operated in parallel with a HRM and an ARMON. The sampling height for all radon detectors at ODM was 2 m a.g.l.
Phase II of the exercise was realized at Saclay (SAC, latitude 48.730, longitude 2.180; Fig. 1) between 25 January and 13 February 2017. At this location the sampling inlet height was at 100 m a.g.l. At SAC station another ANSTO monitor (from now on labeled ANSTO_SAC) was already running. In addition, during Phase II this detector was running in parallel with the portable ARMON and HRM detectors. The LSCE monitor did not participate in Phase II of the experiment.
Meteorological parameters were also available at both stations during the intercomparison periods at heights corresponding to the radon measurements (2 and 100 m a.g.l.). In the case of the ODM site, atmospheric aerosol concentrations were also measured for this period.

Environmental parameters and atmospheric aerosol concentration
Meteorological data used within this study were available from continuous measurements carried out at the SAC and ODM stations at 100 m and at 10 m a.g.l., respectively. The measurements were performed with a Vaisala weather transmitter WXT520 (Campbell Scientific) for (1) wind speed and direction (accuracies of ±3 % and ±3 • C, respectively) and (2) humidity and temperature (accuracies of ±3 % and ±0.3 • C, respectively). In addition, the atmospheric aerosol concentration was measured at the ODM site using a fine dust measurement device Fidas ® 200 S (Palas) at 10 m a.g.l. The measurement range is between 0 and 20 000 particles cm −3 . All the accuracies refer to the manufacturer's specifications.

Correlation factors between monitors
To study the correlation between responses of the different detectors, linear regression models were calculated using hourly atmospheric radon activity concentrations from each monitor. The linear regression fits were calculated following Krystek and Anton (2007), relative to the two portable detec-tors, ARMON and HRM, because they both were measuring at SAC and at ODM.

Analysis of the influence of the environmental and meteorological parameters on detector response
The present study intended to build upon the findings of Xia et al. (2010) and Schmithüsen et al. (2017) regarding the possible influence of meteorological conditions on the response of radon and radon progeny monitors. With this in mind, the ratio between hourly atmospheric 222 Rn activity concentrations measured and/or obtained by the HRM, LSCE, and ANSTO monitors and that measured by the ARMON were calculated, and their variability was analyzed in relation to hourly atmospheric temperature, relative humidity, and atmospheric aerosol concentration measured at ODM and at SAC. Not enough rain data were available to be used in this study. For this part of the study, the ARMON was used as a reference since it was the only direct radon monitor running at both sites.

Results
Hourly time series of atmospheric 222 Rn, in the case of the ARMON and ANSTO monitors, and 222 Rn progeny ( 214 Po activity concentration) for the HRM and LSCE monitors, measured at ODM and SAC during Phase I and Phase II of the intercomparison experiment, are presented in Figs. 2 and 3, respectively. In each of the previous figures, a zoom plot has also been reported as an example of the response of each monitor to the sub-diurnal atmospheric radon variability. As shown, all monitors running at both sites follow this variability, with 222 Rn and 222 Rn progeny data measured or estimated by the three different measurement tech-Atmos. Meas. Tech., 13, 2241-2255, 2020 www.atmos-meas-tech.net/13/2241/2020/ niques showing the same general patterns. Table 2 summarizes the mean, minimum, and maximum hourly atmospheric radon or radon progeny activity concentrations measured by each monitor for both campaigns. For further information, Figs. S1 and S2 in the Supplement show the time series of the differences (absolute) and of the ratios (relative) between the hourly 214 Po or 222 Rn activity concentrations measured by HRM, LSCE, and ANSTO monitors and those measured by the ARMON.

Phase I: ODM site
During Phase I the LSCE, HRM, ARMON, and ANSTO_ODM monitors were operating in parallel, sampling air from the same height (2 m a.g.l.). The mean temperature over Phase I of the campaign was 2.9 • C with an interquartile range of 0.10 to 5.8 • C. The mean relative humidity was 80 % with an interquartile range of 73 % to 89 %. An average accumulated rain per day of 13 mm was recorded. The main wind patterns during Phase I were from the northeast and southwest, with speeds typically between 1 and 7 m s −1 . The mean atmospheric aerosol concentration observed at ODM during Phase I was 505 particles cm −3 with an interquartile range of 233 to 660 cm −3 . The means of the atmospheric 222 Rn activity concentration measured by the ARMON and the ANSTO_ODM are in the same order ( Table 2). The means of the atmospheric 214 Po activity concentrations measured by the LSCE monitor were ca. 50 % lower and ca. 30 % lower by the HRM than the atmospheric 222 Rn activity concentration. Table 2 shows the slopes (b) and intercepts (a) of the linear regression fits calculated between the hourly atmospheric 222 Rn and 214 Po activity concentrations measured by the ARMON and/or the HRM and the other 222 Rn and 222 Rn progeny monitors deployed in Phase I. The calculated slopes were in the range of 0.62 to 1.17, and the R 2 values varied between 0.90 and 0.96. The slope closest to unity was calculated between the ARMON and ANSTO_ODM monitors and was 0.96±0.01, while the lowest slope was observed between the ARMON and LSCE monitors and was 0.62±0.01. The highest correlation (R 2 = 0.96) was found between the HRM and LSCE monitors. The plots of the linear regression fits of Phase I are shown in the left panels of Figs. S3, S4, and S5. Notably, the offset (a value) of the regression between the ANSTO and ARMON detectors at ODM is considerably greater than that at SAC. The regression slopes are also slightly different. These differences are likely related to the limited calibration and background information available for the ANSTO_ODM detector for this intercomparison project. In particular, a substantial component of the instrumental background signal is site specific. This is likely responsible for much of the change in offset value.

Phase II: SAC station
Phase II lasted 18 d. The mean temperature during this period was 5 • C with an interquartile range of 2 to 8 • C. The mean relative humidity was 86 % with an interquartile range of 80 % to 94 %. An average accumulated rain per day of 3 mm was recorded. The main wind patterns during this phase at 100 m a.g.l. were from the south and southwest with speeds typically between 3 and 10 m s −1 . Figure 3 shows the hourly atmospheric 222 Rn and 214 Po activity concentrations observed at SAC during Phase II by the ARMON, HRM, and ANSTO_SAC instruments. Table 2 reports the means, minima, and maxima of the atmospheric data measured during Phase II by all participating monitors. In this case, the mean atmospheric 222 Rn and 214 Po activity concentrations measured by all monitors agreed within the instrumental errors. At 100 m a.g.l. the slopes of the hourly fits of the monitor's response in this case were all close to unity. The calculated offsets also decreased at 100 m a.g.l. relative to 2 m a.g.l. The plots of the linear regression fits of Phase II are shown in the right panel of Figs. S5 and S6. During the period of 30 January to 1 February 2019, the HRM shows significantly lower values than ANSTO and ARMON. This period coincides with saturated air humidity conditions. Figure S7 presents two plots to summarize the results of the slopes and offsets calculated at both ODM and SAC stations relative to the ARMON. Figures 2 and 3 show a larger hourly variability of the HRM and ARMON signals compared with the ANSTO ones. This difference in variability is likely due to a larger uncertainty of the HRM and ARMON detectors for atmospheric 222 Rn levels of around 1 Bq m −3 . In addition, it has to be taken into account that only an approximated form of the Griffiths et al. (2016) response time correction could be applied to the output of the ANSTO detectors. Further investigations should be carried out to clarify these differences and to exactly quantify the detectors' uncertainties for the low 222 Rn concentrations typical for outdoor environmental monitoring at or above 100 m a.g.l.

Comparison with past studies
The results obtained in the present study of the slopes (b) and of the offsets (a) of the regression lines calculated between ANSTO or LSCE monitors against the HRM are here compared with the ones presented by Schmithüsen et al. (2017). Table 3 shows a summary of this comparison. All slopes (correction factors) are defined as (routine station monitor) divided by HRM because this last was used as the reference instrument by Schmithüsen et al. (2017). Data in Table 3 need to be analyzed taking into account that a unique traceability chain is not yet available for atmospheric radon measurements and the different monitors routinely running at the different stations could have differ- ent calibration chains (e.g., radon source, primary standard). Generally speaking, for both studies, it can be observed that the correction factor between the atmospheric 214 Po activity concentration measured by HRM and the atmospheric 222 Rn activity concentration measured by ANSTO at each station approaches unity with the increase in the height of the sampling input. By contrast, the offsets of the regression fits decrease with the increase in the input height.
The only case where the compared instruments were exactly the same and at the same height is for the Orme des Mérisiers station. Here the slope between the atmospheric 214 Po activity concentration measured by LSCE and HRM is equal to 0.76 ± 0.01. This number is slightly larger but within uncertainties comparable to the number reported by Schmithüsen et al. (2017) of 0.68 ± 0.03 (see Table 3). Figure 4 shows the variability of the ratio between hourly atmospheric 214 Po and/or 222 Rn activity concentration measured by each monitor relative to those measured by the AR-MON versus the hourly means of ambient temperature and relative humidity. Analysis was carried out at ODM (Fig. 4a, b) and at SAC (Fig. 4c, d) versus ambient temperature (Fig. 4, left panels) and relative humidity (Fig. 4b, d) measured at the corresponding stations. Figure 5 shows the same variability plotted in relation to the ANSTO_ODM at ODM (Fig. 5a, b) and to the ANSTO_SAC at SAC (Fig. 5c, d) versus the hourly means of ambient temperature (Fig. 5a, c) and relative humidity (Fig. 5b, d).

Influence of the weather conditions on the ratio between 214 Po and 222 Rn measurements
Data do not show any evident patterns at 100 m a.g.l. (SAC station), which could indicate that there is not any impact Atmos. Meas. Tech., 13, 2241-2255, 2020 www.atmos-meas-tech.net/13/2241/2020/    on 222 Rn or 222 Rn progeny measurements due to change in ambient temperature and relative humidity, at least not until saturated conditions are achieved. By contrast, a small decrease, of about 10 −2 • C −1 , is observed in the ratio between the 214 Po activity concentration (measured by the HRM and LSCE monitors) and the 222 Rn activity concentration (measured by the ANSTO_ODM and ARMON monitors) with the increase in the ambient temperature (Fig. S8) at 2 m a.g.l. (ODM station). This temperature dependency may be rather due to the effect of atmospheric activity concentrations, increasing during nighttime, on the disequilibrium between radon and its progeny. However, this influence on measured 214 Po/ 222 Rn ratios is really small compared with other observed effects (e.g., loss of progeny within the sample tube  or atmospheric aerosol concentration; see below). Looking at Fig. 5, there appears to be less scatter in the point clouds (particularly at SAC) when the ANSTO_SAC monitor is used as the reference, likely attributable to the lower measurement uncertainty of the ANSTO monitor used at this station. In Fig. 6 the ratio of the hourly atmospheric 222 Rn or 222 Rn progeny activity concentration measured by the HRM ( 214 Po in Fig. 6a), the LSCE ( 214 Po in Fig. 6b), and the ANSTO_ODM ( 222 Rn in Fig. 6c) Table 4. The slope of the ratio between the ANSTO_ODM and AR-MON monitors in relation to the variability of the logarithm of the hourly atmospheric aerosol concentration is close to zero, and the intercept is close to 1. The ratio between the hourly atmospheric aerosol-bound radon progeny data measured by the two one-filter radon progeny monitors and the one measured by the ARMON seems to decrease with decreasing aerosol concentration ( Fig. 6a and b). However, this effect only becomes evident when atmospheric aerosol concentration is lower than 300 particles cm −3 .

Conclusions
In order to confirm and build upon the results obtained by Xia et al. (2010), Grossi et al. (2016), and , a 3-month intercomparison campaign was carried out in the south of Paris, France, in the fall-winter period of 2016-2017. For the first time, three fundamentally distinct radon and radon progeny measurement approaches were The results of this study show that 222 Rn and 222 Rn progeny measurements follow the same general patterns of diurnal variability, both close to and further up from the surface. The slopes of the linear regression fits between the radon and the radon progeny measurements, which represent the calibration factors, are not significantly different from one at 100 m height above ground (SAC), but they differ at the 2 m level (ODM). The latter is attributable to the disequilibrium known to exist between 222 Rn freshly emitted from the ground and its short-lived progeny in the lowest tens of meters of the atmosphere, the magnitude of which is known to decrease with distance from the surface, as shown in earlier work, and to be close to 1 at a height of 100 m and above (e.g., Jacobi and André, 1963;Schmithüsen et al., 2017).
For the 2 m level, we found a significant correlation of radon progeny activity concentrations between LSCE and HRM measurements (see Fig. S3). The slope, however, is only equal to 0.76±0.01. This result is comparable, considering its uncertainties, with the result reported by Schmithüsen  Table 3) based on the comparison of the same two monitors (HRM and LSCE) and at the same station (ODM) in March and April 2014.
Observations of the total atmospheric aerosol concentration available at the ODM station during the first 2 months of the experiment were used to investigate the influence of changing atmospheric aerosol concentrations on the response of the radon and radon progeny measurements. Under very low atmospheric aerosol loading (< 300 particles cm −3 ), the 222 Rn progeny monitors seem to underestimate the atmospheric 214 Po activity concentrations by up to 50 %. This effect may be attributable to loss of unattached 218 Po and 214 Po. Particle number concentrations below 300 particles cm −3 at continental stations are, however, very rare, and even during winter at Alpine stations like Schneefernerhaus such low particle concentrations are only occasionally observed (Birmili et al., 2009).
The comparison of results obtained in the present study with those reported in Schmithüsen et al. (2017) demonstrate that in order to harmonize atmospheric 222 Rn activity concentrations measured at different atmospheric networks it will be important to (i) have a well-established metrological chain; (ii) have traceable methods for measuring low-level atmospheric radon activity concentrations; (iii) harmonize the calculation of total uncertainty in atmospheric 222 Rn concentrations measured by all monitors when ambient radon is only a few becquerels per cubic meter or less and (iv) use a direct radon monitor as a mobile reference instrument, the response of which is not influenced by meteorological conditions or inlet tube dimensions and length.
Finally, the new portable ARMON seems to have a great potential for being used at atmospheric radon stations with space restrictions. It could also be useful as a mobile reference instrument to calibrate 222 Rn progeny instruments or fixed radon monitors. However, the total expanded uncertainty of the ARMON could increase for really low radon activity concentrations (< 1 Bq m −3 ) and when atmospheric 220 Rn is also present. This should be better investigated in the near future. The uncertainties related to the ANSTO detector response time correction when characteristics of the entire intake system have not been directly measured should also be investigated.
Author contributions. CG, OL, FRV, VK, MD, RC, JAM, SDC, and AC worked on the installation of the radon and the radon progeny monitors. In addition, they were in charge of the maintenance of Atmos. Meas. Tech., 13, 2241-2255, 2020 www.atmos-meas-tech.net/13/2241/2020/ the in situ and remote radon and radon progeny measurements dur-