Introduction

The atmosphere contains natural and anthropogenic radionuclides. The natural radionuclides are either produced in the upper levels of the troposphere and stratosphere by cosmic-ray interaction with molecules (cosmogenic) or are primordial and secondary transported from terrestrial areas due to Aeolian and diffusion processes. The anthropogenic radionuclides are produced in reactors and accelerators and are either accidentally released in the atmosphere or are remnants from nuclear weapon tests and accidents. In the atmosphere, most radionuclides have a strong tendency to be attached to aerosols forming the so-called radioactive aerosols. This work regards the use of the natural radionuclide of radon as a tracer in precipitation investigation.

Radon is the most abundant natural radionuclide in the atmosphere. The radon 222Rn has a half-life of t1/2 ≈ 3.8 d and the radon 220Rn (referred also as “thoron”) has a half-life of t1/2 ≈ 55 s. Radon is a radioactive inert gas, which emanates from terrestrial areas and is dispersed in the lower levels of the troposphere due to advection and diffusion processes. After its radioactive decay, radon progenies are produced (mainly radionuclides of Po, Pb, Bi and Tl). With the presence of aerosols, radon progenies get attached to them forming radioactive aerosols. In that form, they are involved in the formation of cloud droplets and raindrops. During precipitation, they are scavenged and removed from the atmosphere by wet deposition. So far, two scavenging processes have been referred to; in-cloud (rainout) and below-cloud (washout) scavenging. Inside the cloud (in-cloud processes), the radioactive aerosols are involved in activation/nucleation scavenging (vapour condensation around aerosols forming droplets) and impaction scavenging (aerosol collection by droplets). Below the cloud, only impaction scavenging (aerosol collection by raindrops) may take place. In the former case, radon progenies can be found both inside the volume of a raindrop and/or attached to its surface while in the latter only on its surface. Up to now, many works suggest that in-cloud scavenging is the dominant one (Fujinami 1997; Paatero and Hatakka 1999).

A simplified description of radon progenies’ behaviour in the atmosphere before and during precipitation starts with the in-cloud formation of droplets by condensing water vapours in the presence of atmospheric aerosols. In the case of radioactive aerosols, the forming droplets contain radon progenies and carry them throughout in-cloud movement and suspension. As droplets are tiny and numerous, offer an extensive effective surface for unattached radon progenies to continue the attachment, so the population of the attached radon progenies in droplets increases. The period while droplets are moving in the cloud is usually called droplets in-cloud residence time. During that time, radon in the cloud volume decays and produces progenies steadily supporting their population. Thus, during the in-cloud residence time, radon can be considered to be in radioactive equilibrium with its first progeny of 218Po (Horng and Jiang 2003; Minato 2007). As the number of droplets increased, more frequently they come into contact forming droplet clusters and successively bigger raindrops (coagulation process). However, the size of raindrops is much larger than that of droplets and also they are fewer in number so raindrops offer a less effective surface for radon progenies attachment. Consequently, the support of radon progenies by radon ceases and the state of radioactive equilibrium is ended. The period the raindrops spend in the cloud is called raindrop in-cloud residence time and during that period, radon progenies start to decay without support from radon thus, a radioactive disequilibrium state between them begins. Raindrops becoming heavier are moved downwards to the bottommost cloud region and finally, they leave the cloud volume, falling to the earth’s surface, spending some time (falling time) in the atmosphere below the cloud. Throughout the raindrops falling, radon progenies of the air may further be scavenged (washout) although this accumulation process is considered minor compared with the rainout process. The sum of the raindrop in-cloud residence time and falling time is usually called raindrop age. The disequilibrium between radon progenies contained in the raindrops can be converted into time and an estimation of the raindrop age can be realized based on radon progenies activity concentration measurements of rainwater (Cortès et al. 2001; Greenfield et al. 2003, 2008; Jun et al. 2008; Minato 2007; Moriizumi et al. 2015; Paatero and Hatakka 1999; Tsujimoto et al. 2006).

To investigate the behaviour of radon progenies in the atmosphere before and during precipitation, several methods have been developed. Commonly, liquid samples collected from rainwater or snow are analysed using laboratory detectors based on (a) gamma-ray spectroscopy (Fujinami 1997; Greenfield et al. 2002; Minato 1980; Moriizumi et al. 2015; Paatero 2000; Paatero and Hatakka 1999; Tsujimoto et al. 2006), (b) alpha and beta counting methods (Gómez Escobar et al. 1996) and (c) combined methods (Seftelis et al. 2008, 2007; Stoulos and Ioannidou 2020). Also, automated sampling methods (Casanovas et al. 2013; Cortès et al. 2001) and in situ methods in the air (Livesay et al. 2014; Mercier et al. 2009) and in the marine environment have been implemented (Patiris et al. 2021; Pensieri et al. 2018; Tsabaris et al. 2008; Tsabaris and Ballas 2005).

The main objective of this work concerns the assessment of the use of a medium-resolution underwater detector for in situ continuous gamma-ray spectrometry in real atmospheric conditions before, after and during precipitation events. This method is a rather straightforward approach needing no maintenance procedures, sampling or measurement preparation. It aims to be utilized for experiments in remote areas, without surveillance or personnel involvement. The main questions of the work regard to a. assess the possibility to detect and examine radon progenies variations before, during and after precipitation events, b. investigate the main factors that have an impact on the radon progenies variations, c. reveal any association between radon progenies variations with precipitation parameters, d. find the optimum conditions and limitations of the proposed in situ method for prolonged continuous monitoring of the atmosphere before and during precipitation.

Study area and experimental set-up

The in situ continuous monitoring of the radon progenies of lead 214Pb, bismuth 214Bi and thallium 208Tl in the atmosphere was realized utilizing the gamma-ray spectrometer GeoMAREA (Tsabaris et al. 2019). The system is based on a (2″ × 2″) CeBr3 scintillation crystal which provides medium-resolution gamma-ray spectra (≈ 6% at 661.6 keV) in the energy range of 150–2600 keV. Also, the intrinsic radiation of the crystal results in the internal background at the energy range 1400–1800 keV which is much lower than that of LaBr3 detection crystals. The detection system was installed on an auxiliary platform (which provided energy by a waterproof battery pack and real-time communication with a PC) with the detection crystal facing upwards toward the atmosphere. The installation was realized on the roof of the HCMR’s facilities in the region of Anavyssos, Greece (Fig. 1), approximately 15 m above the ground to minimize any potential variations due to radon emanation variation from the soil. Continental areas are extended toward the north and west while the Aegean Sea is toward the east and south. The detector acquired one spectrum per hour and meteorological data were obtained every 10 min by the meteorological station of the HCMR’s Poseidon network located approximately 5 m away from the detector. The duration of the experiment covered the season from 1 November 2017 until 1 April 2018. During that period, 180 distinct rainfall events occurred. GeoMAREA system was programmed to hourly acquire one gamma-ray spectrum and then the auxiliary communication system of the platform (Alexakis and Tsabaris 2021) sent each spectrum in (near) real-time mode to the operation centre for further analysis.

Fig. 1
figure 1

The study area is located near Anavyssos city, Southern Attica, Greece (a). The underwater GeoMAREA gamma-ray spectrometer deployment with the detection crystal (CeBr3) straightforwardly facing up to the air (b)

Full size image

Methods

The determination of the volumetric activity concentration of radon progenies in absolute units (Bq m−3) employing the applied in situ method, demands the estimation of the full energy photo-peak efficiency (FEPE) for each corresponding photo-peak. However, several methodological issues do not allow the accurate estimation of FEPE. The major issue concerns the lack of a voluminous calibration source with dimensions and total volume as the effective volume in the atmosphere of the detector. The effective volume is the volume around the detector within it, emitted gamma-rays have a (nonzero) possibility to be detected. The atmospheric air has a very low density (approximately 1.2 kg m−3) so the possibility of a gamma-ray interacting with an atom to be absorbed in the air before reaching the detector is also very low. In other words, the detector can detect gamma-rays emitted from very far locations. Thus, the effective volume is huge and within it, many time-varied factors may influence a theoretical estimation by Monte Carlo simulation codes of FEPE (e.g. surrounding structures, moving objects, atmospheric changes).

Thus, this work is based on examining the variation of specific parameters before, during and after precipitation events. The parameters regard a) the total counting rate (TCR in s−1) of each spectrum which is the rate of all gamma-rays detected regardless of their energy and b) the counting rate (CR in s−1) of net areas of photo-peaks of interest which is calculated by the subtraction of the Compton background under each photo-peak. More specifically in the analysis of each spectrum, the following steps were realized:

  1. (a)

    GeoMAREA is a medium-resolution scintillation detector and is prone to slight spectrum shifts (± 10 channels) due to temperature variations. So, each spectrum was automatically re-calibrated according to the energy by algorithmic routine using the 609 keV photo-peak of 214Bi and the 1460 keV photo-peak of 40 K which were the most intense photo-peaks.

  2. (b)

    Calculation of the TCR by dividing the number of all detected counts with the acquisition time of 3600 s.

  3. (c)

    Calculation by an algorithmic routine of the CR of selected energy regions of interest (RoIs) corresponding to 222Rn radon progenies of 214Pb: 295, 352 keV and 214Bi: 609 keV photo-peaks and to 220Rn progeny 208Tl: 2614 keV photo-peak.

  4. (d)

    Calculation of the net CR for selected spectra. As the total number of the collected spectra was too high (3371) for manual analysis, selected spectra were analysed by the gamma-ray spectrum analysis software SPECTRW (Kalfas et al. 2016) to obtain the net counting rate by Compton background subtraction in the aforementioned photo-peaks. More than 550 spectra were manually analysed which were selected by the whole range of TCR.

  5. (e)

    The CR values of the RoIs and net areas of each photo-peak as calculated in steps c) and d) were found to be very strongly linearly related. Thus, from the linear regression model, the net areas of all acquired spectra were indirectly calculated and verified by manual calculations.

In the case of the spectra acquired during precipitations, another step took place initially. From the spectrum acquired during a precipitation event (Fig. 2a blue line), the number of counts of each channel was subtracted by the number of counts of the corresponding channel of the spectrum (Fig. 2a red line) obtained just before the event initialization. The subtraction channel by channel removes the contribution of radon progenies that are not related to the precipitation event itself (e.g. radon progenies of the dry air, the floor of the roof and all the surrounding structures). In such a way, a spectrum (Fig. 2b black line) representing the concentration of radon progenies solely in rainwater was derived. The rainwater spectrum amounts to the gamma-rays emitted from (a) raindrops in the atmosphere around the detector, (b) fall-out rainwater on the floor of the roof and the structures that surround the detector and (c) rainwater on the detector’s surface.

Fig. 2
figure 2

Two acquired spectra during a rainfall (blue line) and before the rainfall (red line) (a) and the derived rainwater spectrum (b). The vertical lines depict the photo-peaks corresponding to 222Rn progenies of 214Pb: 295, 352 keV and 214Bi: 609 keV (orange line) and to 220Rn progeny 208Tl: 2614 keV photo-peak (green line)

Full size image

Results and discussion

Qualitative analysis of spectra before and during rainfalls

In Fig. 3, the spectra acquired during (blue line) and before rainfall (red line) as well as the derived spectrum of rainwater (black line) are presented magnified for three energy ranges of interest (RoI).

Fig. 3
figure 3

The spectra acquired during (blue line) and before a rainfall (red line) (a–c) as well as the derived spectrum of rainwater (black line) (d–f) are magnified for three energy ranges of interest 200–700 keV (a, d), 1300–1600 keV (b, e) and 2500–2700 keV (c, f) related to 214Pb and 214Bi, 40 K and 208Tl, respectively

Full size image

The first RoI regards the energy range (200–700) keV where three photo-peaks of radon progenies (214Pb: 295 keV, 214Pb: 352 keV and 214Bi: 609 keV) are significantly increased during the rainfall (Fig. 3a blue line). These photo-peaks are observed in the rainwater spectrum (Fig. 3d black line), revealing the presence of 214Pb and 214Bi in the rainwater. They are well-formed and separated allowing the calculation of the net area counting rate by the subtraction of the Compton background. Similar observations can be found in other energy ranges where photo-peaks of 214Bi are emitted (e.g. 1120, 1764, 2220 keV).

The second RoI regards the energy range (1300–1600) keV where the photo-peak of 40 K (1461 keV) is expected. During the rainfall, the height of the photo-peak slightly increases while its width is significantly broadened (Fig. 3b blue line) in comparison with the 40 K photo-peak before the rainfall (Fig. 3b red line). This observation is not attributed to an increase of 40 K but to the eight, low intensity, overlapped photo-peaks of 214Bi in that energy range (namely 1377 keV with 4%, 1385 keV with 0.77%, 1401 keV with 1.38%, 1408 keV with 2.48%, 1509 keV with 2.2%, 1538 keV with 0.4%, 1543 keV with 0.35% and 1583 keV with 0.72%). This justification is also strengthened by the absence of a well-formed photo-peak in the rainwater spectrum (Fig. 3e black line) centred on 1461 keV. Thus, although the sum of the counts in the vicinity of 1461 keV increases during the rainfall, no observable increment of 40 K activity concentration can be considered.

The third RoI regards the energy range (2500–2700) keV where the photo-peak of 208Tl (progeny of radon 220Rn) is expected centred on 2614 keV. In the spectrum during the rainfall (Fig. 3c blue line), no increment of the counts can be observed in comparison with the counts of 208Tl photo-peak before the rainfall (Fig. 3c red line). Also, in the spectrum of rainwater (Fig. 3f black line), no photo-peak can be observed centred on 2614 keV. Although 220Rn gas, like 222Rn, emanates from terrestrial areas and is transported inside the atmosphere, it has a very short half-life (55.6 s). This short period seems not to be enough for 220Rn to reach the heights in the atmosphere (several kilometres) where cloud formation processes occur and to be sufficiently engaged in droplets formation and scavenging processes, before its total radioactive decay.

Time series

The analysis of the derived rainwater’s spectra (Fig. 2b black line) was used to create time series of the results with hourly averaged values of the quantities of rain height (in mm) and rain rate (in mm h−1). A representative plot is depicted in Fig. 4 regarding the period of 16–19 November 2017. The total counting rate of the rainwater’s spectra is depicted in Fig. 4a and b along with rain height and rain rate, respectively. The counting rate of the net area of 351 keV and 609 keV photo-peaks corresponded to radon progenies of 214Pb and 214Bi are depicted in Fig. 4c and d along with rain height and rain rate, respectively.

Fig. 4
figure 4

Representative time series regarding the period 16–19 November 2017. In a, b, the TCR (in s−1) of the derived rainwater’s spectra are depicted (black points) along with the rain height (in mm) and rain rate (in mm h−1) (blue bars), respectively. In c, d the CR (in s−1) of the net area of the 214Pb: 351 keV (red points) and 214Bi: 609 keV (green points) photo-peaks are depicted along with the rain height and rain rate, respectively

Full size image

In Fig. 4a and b, the TCR of rainwater’s spectra increment during the rainfall events can be observed and similarly the increment of the net area counting rates of 214Pb and 214Bi in Fig. 4c and d are depicted. However, their increase seems not to be proportional or related to the rain height or the rain rate. During all the rainfall events, the increase starts in the first 60 min of the rainfall and the values minimize and return to values before the event after 3 h from the end of each rainfall. This period is comparable with 5–6 times the half-lives of 214Pb (26.8 min) and 214Bi (19.9 min). During that period, their population decreases by at least 97% due to radioactive decay.

Furthermore, the acquisition time of 60 min is almost 3 times the half-lives of 214Pb and 214Bi. As 214Pb is a parent of 214Bi, it would be reasonable for the built-up of 214Pb at the beginning of a rainfall event to precede the built-up of 214Bi. Equivalently, 214Bi decay would be reasonable to last further after the end of a rainfall event. However, none of such observations was revealed as the variations of 214Pb and 214Bi depicted in Fig. 4c and d appear so similar. Thus, the selected acquisition time of 60 min seems to be too long to distinctly investigate differences in radon progenies’ built-up and decay processes during and after rainfalls.

Also regarding TCR values minimization, it is observed that occurs at the same time as the decay of 214Pb and 214Bi after the end of each rainfall. This is strong evidence that the increment of TCR of the rainwater’s spectra resulted mainly from the fall-out of radon 222Rn progenies and TCR decreases because of their decay. This assertion is confirmed by the very strong positive correlation of TCR with both 214Pb and 214Bi net area CRs as depicted in Fig. 5. From an operational point of view, this provides a rapidly produced parameter (as spectrum analysis is not needed) for rainfall investigation. Taking also into account the very low uncertainty of TCR (less than 0.3% for acquisition of 1 h), this parameter proved to be very effective for rainfall identification and investigation.

Fig. 5
figure 5

The net area of the 214Pb: 351 keV (red points) and 214Bi: 609 keV (green points) photo-peaks over the TCR of the derived rainwater spectra. The results of the linear regression model are presented in the two insert legends

Full size image

Rainwater’s radon progenies association with meteorological parameters

Scatter plots of radon progenies net area CRs with the measurements of temperature, pressure, humidity and dew point did not reveal any linear relation. However, further statistical analysis was realized to unveil any other possible association. The data of TCR and radon progenies net areas CRs (214Pb: 352 keV, 214Bi: 609 keV) were divided into two groups. The first group contains measurements without rainfall (No rainfall, N = 3075) and the second group those during rainfalls (Rainfall, N = 296). For both groups, according to the Shapiro–Wilk normality test, the distribution of the results cannot be approached by a normal distribution so the nonparametric Spearman rho and Kendall tau rank association methods were selected. In Tables 1 and 2, the statistical association of TCR and radon progenies’ net area CRs (214Pb: 352 keV, 214Bi: 609 keV) with temperature, humidity, dew point and atmospheric pressure are presented according to the Spearman rho and Kendall tau rank association methods. Table 1 contains the results for the case of the “No rainfall” group and Table 2 contains the results for the case of the “Rainfall” group. Considering the association coefficients rho and tau, there is no strong or even moderate association between radon progenies and the meteorological parameters either during rainfalls or dry meteorological conditions.

Table 1 Statistical association of TCR and radon progenies’ net area CRs (214Pb: 352 keV, 214Bi: 609 keV) with meteorological parameters during dry periods (no rainfalls) according to Spearman rho and Kendall tau methods. Statistically significant results (confidence level 0.05) are depicted with bold fonts
Full size table
Table 2 Statistical association of rainwater’s TCR and radon progenies’ net area CRs (214Pb: 352 keV, 214Bi: 609 keV) with meteorological parameters during periods of rainfalls according to Spearman rho and Kendall tau methods. Statistically significant results (confidence level 0.05) are depicted with bold fonts
Full size table

Rainwater’s radon progenies dependency on wind direction

Another possible association was investigated regarding the atmospheric mass movement on the concentration of radon progenies in the rainwater. The wind direction data recorded by the meteorological station installed 15 m above the ground was used for the analysis. However, the wind direction can be drastically different in altitudes of cloud formation so the use of this parameter is indicative. The wind direction measurements were hourly averaged and divided into groups according to 16 distinct orientations with a step of 22.5o, namely north (N): 0, north-northeast (NNE): 22.5°, northeast (NE): 45°, east-northeast (ENE): 67.2°, east (E): 90°, etc. Also, the 17th group regards the case of rainfall under calm, no windy conditions. Then, the values of TCR and radon progenies’ net areas CRs (214Pb: 352 keV, 214Bi: 609 keV) of the derived rainwater’s spectra were attributed to the corresponding wind direction group. In Fig. 6, the results are presented in form of boxplots. The bold black line of each box is the median of the values’ distribution for each wind direction. With the red line, the median of the values’ distribution under calm, no windy conditions, is also depicted.

Fig. 6
figure 6

Box plot of the distribution of the values regarding the TCR (a) and the net area of 214Pb: 352 keV (b) and 214Bi: 609 keV (c) photo-peaks according to the wind direction. The first group (---) regards the values during no wind conditions and the red line depicts the corresponding median

Full size image

With exception of the wind directions WSW and NNW where the sample size was too low (n = 2 and n = 0, respectively), in all other cases the median values of the distribution of the TCR (Fig. 6a) and radon progenies’ net areas CRs 214Pb: 352 keV (Fig. 6b), 214Bi: 609 keV (Fig. 6c) are higher than the median under calm, no windy conditions. Also, there are directions where the median is significantly higher (N, E) and others where is significantly lower (SSE, SSW). This finding reveals a considerable dependency between radon progenies concentration in rainwater with the wind direction. This has been also reported in other studies realized in different locations (Jasaitis et al. 2013; Jun et al. 2008; Paatero and Hatakka 1999).

Furthermore, the dependency of the radon progenies in rainwater is not similar to the dependency of the rain height and rain rate on the wind direction as can be observed in Fig. 7. More specifically, in Fig. 7, boxplots present the results of the analysis of rain height (Fig. 7a) and rain rate (Fig. 7b) with the wind direction. The higher median values of rain height and rate regard the wind directions W, SW, and ENE. The wind directions where radon progenies are higher (N, E) are mostly related to low (N) and low-medium (NE) medians of rain height and rate.

Fig. 7
figure 7

Box plot of the distribution of the values regarding the rain height (a) and rain rate (b) according to the wind direction. The first group (---) regards the values during no wind conditions and the red line depicts the corresponding median

Full size image

The interpretation of the dependency was performed by the simulation of the atmospheric masses’ movement by models for atmospheric trajectory and dispersion calculations. In this work, the HYSPLIT (Rolph et al. 2017; Stein et al. 2015) model was used to provide backward trajectories of atmospheric masses from specified geographical location, altitude and time according to the following set-up:

  1. (a)

    The coordinates of the detector’s position set the location of the model. As altitude, an indicative height of 5000 m above the ground level (AGL) was selected, representing a typical altitude of cloud formation.

  2. (b)

    Four backward trajectories were set to go back to the past for a period of 48 h. This period is comparable with radon half-life (4.8 d) and is typical for cloud formation, movement and precipitation initialization. The first (red line) and the second (blue line) backward trajectories start 2 h and 1 h before the time of the rainfall initialization. The third (green line) and fourth (turquoise line) trajectories start 1 h and 2 h after the rainfall initialization. In this way, the movement of the atmospheric masses is simulated 2 h before, during and 2 h after the first record of rainfall from the meteorological station.

  3. (c)

    Among the recorded rainfalls, indicative events were selected with the criterion of the rainwater’s TCR to be very low or very high.

In Figs. 8 and 9, the backward trajectories of four representative rainfall events are depicted. In Fig 8a and b, rainwater’s TCR values were very low indicating two rainfall events of rainwater lacking radon progenies. In Fig 9a and b, TCR values were very high, indicating two rainfall events of rainwater rich in radon progenies. For each rainfall event, four back trajectories of atmospheric masses lasting 48 h were selected covering a period of 2 h before and 2 h after the rainfall initialization. The time between two consecutive points of the trajectories is equal to 6h so, following the points of the trajectories in the reverse direction, the altitude and the location of atmospheric masses over time can be estimated on the map.

Fig. 8
figure 8

Four back trajectories of atmospheric masses lasting 48 h before two representative rainfall events of very low rainwater TCR. Regarding the time of each event, red trajectory initiated 2 h before, blue 1 h before, green 1 h after and turquoise 2 h after the rainfall started. The altitude of the atmospheric masses is depicted in the subplot

Full size image
Fig. 9
figure 9

Four back trajectories of atmospheric masses lasting 48 h before two representative rainfall events of very high rainwater TCR. Regarding the time of each event, red trajectory initiated 2 h before, blue 1 h before, green 1 h after and turquoise 2 h after the rainfall started. The altitude of the atmospheric masses is depicted in the subplot

Full size image

For each rainfall event of very low rainwater TCR, the four trajectories came from totally different areas, North West Africa (Fig. 8a) and Central Europe (Fig. 8b). In the first case, the atmospheric masses move in high altitudes (> 4000 m AGL) for 48 h before the rainfall initialization. In the second case, the atmospheric masses came from altitudes higher than 5000 m AGL and remain above 4500 m AGL until reach the location of the study.

In the first case of a rainfall event of very high rainwater TCR, the four trajectories originate at high altitudes (> 4500 m AGL) above North Atlantic and pass Northwestern Europe (Fig. 9a). However, the passing of the masses from the terrestrial European areas occurred from much lower altitudes (> 1500 m AGL) before gaining altitude again above the Adriatic Sea. In the second case, two trajectories originate at high altitudes (> 6000 m AGL) above the Baltic and the North Sea and pass terrestrial areas of Western Europe and the West Mediterranean Sea at altitudes higher than 4000 m AGL. However, the other two trajectories originate at very low altitudes (< 1000 m AGL) above North African terrestrial areas and rapidly gain altitude passing the Libyan and the South Aegean Seas. Thus, the atmospheric masses of the second event are a mixture of masses originating from very distant areas and very different altitudes between very high to very low.

Radon concentration is much higher at low altitudes and decreases with altitude increase. In recent airborne radon experimental studies (Baldoncini et al. 2017; Williams et al. 2011), a critical altitude above which radon concentration significantly decreases is 1500 m AGL. From 1500–3000 m AGL radon concentration steadily decreases and above 3000 m AGL is one order of magnitude lower than that of 1500 m AGL (Williams et al. 2011). According to the aforementioned description of the atmospheric masses trajectories, masses which pass from terrestrial areas at altitudes lower than 1500 m AGL are significantly enriched in radon resulting in high rainwater TCR. On contrary, masses that move at altitudes higher than 3000 m AGL have a very low possibility to be enriched by radon resulting in low rainwater TCR.

Similar observations were found during HYSPLIT back trajectories runs regarding other recorded rainfall events as well as, in the case of changing the selected altitude of the scenario to 4000 m or 3000 m AGL. Thus, although more data and scenario analysis are necessary for a more comprehensive discussion, the dependency between radon progenies concentration in rainwater with the wind direction is attributed to the complicated trajectories the atmospheric masses follow before a precipitation event. Those passing over terrestrial areas at low altitudes (less than 1500 m AGL) have an increased possibility to be highly enriched in radon.

Rainwater’s radon progenies association with rain height and rate

Although in the plots of time series (representative period in Fig. 4) rainwater’s TCR and radon progenies’ net areas CRs 214Pb: 352 keV, 214Bi: 609 keV seem not to be associated with rain height and rain rate, a more comprehensive statistical analysis was followed. The analysis aimed to unveil any association of radon progenies related to high or low intensity and rainwater mass. The values of rain height and rain rate were processed by the “k-means” clustering method. The method partitions the whole population of data into k clusters/groups by minimizing the “with-in cluster” variance. The optimum number of clusters can be found either by visual observation or by statistical methods. In this work, the optimum number of clusters was determined by the Elbow method trying also to produce clusters of sufficient population number. Thus, the optimum number of clusters was found to be 2 for the rain height and 3 for the rain rate values.

More specifically, the values of rain height were separated into two clusters. The “Low Rain Height” cluster contains a number of N = 152 values in the range (0.2–0.8) mm while the “High Rain Height” cluster contains a number of N = 16 values in the range (0.8–2) mm. As regards the values of rain rate, they were separated into three clusters. The “Low Rain Rate” cluster contains a number of N = 91 values in the range (0.2–4.2) mm h−1, the “Moderate Rain Rate” cluster contains a number of N = 22 values in the range (4.2–10) mm h−1 and the “High Rain Rate” cluster contains a number of N = 6 values in the range (10–85) mm h−1. For each cluster, the nonparametric Spearman rho and Kendall tau rank association methods were selected to reveal any potential association of rain height and rate with the rainwater’s TCR and radon progenies’ net area CRs 214Pb: 352 keV, 214Bi: 609 keV. In Table 3, the results of the analysis are presented. In the cases when the association is statistically significant, the corresponding coefficient is presented in bold.

Table 3 Statistical association of rainwater’s TCR and radon progenies’ net area CRs (214Pb: 352 keV, 214Bi: 609 keV) with the data clusters of “Low Rain Height”, “High Rain Height”, “Low Rain Rate”, “Moderate Rain Rate” and “High Rain Rate” according to Spearman rho and Kendall tau method. Statistically significant results (confidence level 0.05) are depicted with bold fonts
Full size table

Rainwater’s TCR and radon progenies’ net area CRs 214Pb: 352 keV, 214Bi: 609 keV revealed a weak positive association (0.2–0.4) with the “Low Rain Height” cluster according to both association methods. Also, according to the Spearman rho method, they reveal a weak positive association (−0.2 to −0.4) with the “Low Rain Rate” cluster. Moreover, 214Bi: 609 keV net area CRs revealed a moderate negative association (−0.4 to −0.6) with the “High Rain Height” and “High Rain Rate” clusters according to both association methods. Although more data are needed for validation, that negative association indicates that intense rainfalls and/or rainfalls of higher rainwater quantity may be related to lower radon progenies falling out in comparison with rainfalls of lower intensity and rainwater quantity.

Conclusions

This work proposes an in situ gamma-ray spectrometry method using the medium-resolution scintillation (CeBr3) system “GeoMAREA” for continuous direct detection of gamma-ray spectra in the atmosphere 15 m AGL before, during and after rainfall events. The capabilities and limitations of the proposed method are summarized:

  1. a.

    Due to the huge volume and the complexity of the detection geometry, the full energy photo-peak efficiency cannot be estimated by experimental or simulation methods. From the derived rainwater spectra, the TCR and net area CRs in regions of interest were obtained and used for statistical analysis instead of volumetric activity concentration measurements.

  2. b.

    The analysis of the rainwater spectra reveals that the major contributors to the observed photo-peaks concern the progenies of 222Rn (214Pb, 214Bi) so the proposed method is capable to use 222Rn progenies variation to investigate precipitation. The increase in TCR and the net areas CRs of 214Pb: 352 keV and 214Bi: 609 keV photo-peaks was statistically significant to all of the 180 precipitation events, so they are effective parameters for rainfall identification.

  3. c.

    Although more data are needed for validation, no association was found between radon progenies and temperature, pressure, humidity and dew point, either during rainfalls or dry meteorological conditions. Also, a moderate negative association was found between radon progenies and high values of rain height and rate indicating that intense rainfalls and/or rainfalls of higher rainwater quantity may be related to lower radon progenies falling out in comparison with rainfalls of lower intensity and rainwater quantity.

  4. d.

    Preferable wind directions for rainwater rich in radon progenies revealed the impact of the atmospheric masses trajectories before a precipitation event. According to HYSPLIT modelling of selected rainfall events, air masses that pass over terrestrial areas at low altitudes (< 1500 m AGL) 48 h before the event result in rainwater enriched in radon progenies. On contrary, air masses that pass before an event over terrestrial areas at higher altitudes (> 3000 m AGL) result in rainwater of low radon progenies concentration.

Overall, the method is considered promising for continuous in situ measurements of radon progenies in the atmosphere and prolonged time series covering full seasons may extend the use of radon as a tracer for studies related to climate investigation.