Meteorological Approach in the Identification of Local and Remote Potential Sources of Radon: An Example in Northern Iberian Peninsula

This paper presents a meteorological approach to identify local and remote sources driving the variability of surface daily radon concentrations. To this purpose, hourly 222Rn concentration and surface meteorological measurements, and air mass trajectories at Bilbao station (northern Iberian Peninsula) during the period 2017–2018 have been taken as reference. To investigate the potential transport pathways and potential 222Rn sources, the backward trajectory cluster analysis, trajectory sector analysis (TSA), and potential source contribution function (PSCF) are applied. On average, the diurnal 222Rn cycle shows the expected behaviour, with larger concentrations during the night and minimum concentrations during the daylight hours, with differences in the seasonal amplitudes. According to daily differences between maximum and baseline values, 222Rn daily cycles were grouped into six groups to identify meteorological conditions associated with each amplitude, and potential source areas and transport routes of 222Rn over Bilbao. The trajectory cluster and the TSA method show that the main airflow pathways are from the south, with small displacement, and the northeast, while the analysis of surface wind speed and direction indicates that the highest amplitudes of 222Rn concentrations are registered under the development of sea-land breezes. The PSCF method identified south-western and north-eastern areas highly contributing to the 222Rn concentration. These areas are confirmed by comparing with the radon flux map and the European map of uranium concentration in soil. The results have demonstrated the need in combining the analysis of local and regional/synoptic factors in explaining the origin and variability of 222Rn concentrations.


Introduction
The radioactive noble gas radon (hereby, 222 Rn), a colourless, odourless, and chemically inert gas is part of the natural decay chain of 238 U and originates directly from the decay of 226 Ra, practically ubiquitous in rocks and soils. 222 Rn is released from soils and rocks and reaches the atmosphere through three main processes: (i) Emanation -release from the solid mineral grains to the filled pores; (ii) Transport of the emanated radon through the pores of the soil/rock into the ground surface; (iii) Exhalation the radon released to the atmosphere and the exhalation rate is defined as the amount of radon released per surface and time unit [1].
Due to the relatively long half-time (T 1/2 ≈ 3.8 d), it tends to accumulate in buildings reaching elevated concentrations in indoor air (up to thousands Bq/m 3 ), which, represents a health hazard for the inhabitants. Indeed, the inhalation of 222 Rn and its progenies is the largest source of public exposure to naturally occurring radioactivity [2]. It is responsible 2 of 17 for half of the natural dose received by the global population [2] and it is a major cause of lung cancer after smoking [3]. On the other hand, radiation exposure from inhalation of outdoor 222 Rn and its progeny is less compared to those indoors and it is known to have no major impact on health [3]. 222 Rn released in the atmosphere is quickly diluted reaching a concentration in the range between 1 Bq/m 3 and 100 Bq/m 3 [4]. It depends not only on the magnitude of the release rate from the soil but also on atmospheric mix phenomena. However, outdoor concentrations can reach hazardous levels in regions with high natural radiation or NORM-Naturally Occurring Radioactive Material sites (i.e., phosphogypsum piles [5]).
Although outdoor 222 Rn does not represent a significant health risk to the general population, it plays an important role in scientific research. For this reason, it is routinely measured in many parts of the world due to its use to analyse different research topics, such as (1) to study atmospheric transport and mixing processes within the planetary boundary layer [6], (2) to characterise air mass history and fetch at remote sites [7], (3) to indirectly estimate greenhouse gas (GHG) fluxes using the Radon Tracer Method (RTM) [8,9] and in radiation protection the estimation of radon priority areas (RPA) [10], (4) to analyse residues from certain NORM industries [11], and (5) to characterize radon wash-out peaks in the ambient dose rate data, which are exchanged in the EURDEP early warning system for radiological/nuclear accidents (https://remon.jrc.ec.europa.eu/About/Rad-Data-Exchange, accessed on 14 October 2022). The EMPIR project 19ENV01 TraceRadon (http://traceradonempir.eu/, accessed on 1 October 2022), which started in 2020, represents one example of research focused on atmospheric radon and radon flux. The project aims to improve traceable low-level radon activity concentration and radon flux measurements and to provide the necessary infrastructure for measuring both [12].
Considering the relevance of 222 Rn measurements in different scientific disciplines, the analysis and understanding of the day-to-day variability and the identification of potential sources are needed. In this sense, the characterization of the variability of outdoor 222 Rn concentrations in the boundary layer results from a combination of local and regional/synoptic meteorological factors [7,13]. An increase in atmospheric pressure, rainfall, and snowfall decrease the radon flux whereas an increase in wind speed or temperatures increases it, while, on the contrary, the contribution from remote sources is strongly influenced by the transport pattern affecting the regions [14]. The influence of each contribution can be evaluated by analysing both meteorological observation data and model results [11,15].
In this context, this study aims to investigate the 222 Rn activity concentrations, characterise the transport pathways, and identify potential sources of 222 Rn in Bilbao over a continuous time span ranging from 2017 to 2018 by means of meteorological and 222 Rn measurements, and meteorological methods. The local footprint has been analysed using the wind and precipitation data, while the regional footprint analysis, which offers information about the synoptic atmospheric circulation, has been performed by using the HYSPLIT model by calculating backward trajectories. In this sense, air mass back trajectory analysis is frequently used to solve inverse pollutant transport by estimating the path of an air parcel backward in time at a receptor site by means of different methodologies [16,17]. In the present study, the trajectory cluster analysis and trajectory sector analysis (TSA) are used to identify transport pathways, as well as the potential source contribution function (PSCF), which is used to investigate potential sources [18,19]. To finalize, those areas identified by applying these meteorological methods are then compared with the Radon Flux Maps [20] and with the European Map of Uranium in soil available on REMONwebportal (https://remon.jrc.ec.europa.eu/About/Atlas-of-Natural-Radiation/Digital-Atlas/ Uranium-in-soil/Uranium-concentration-in-soil-, accessed on 20 October 2022) This work is organized as follows. In Section 2, we first describe the 222 Rn and meteorological datasets, the scheme to calculate air mass trajectories, and the methodologies applied. Then, in Section 3, we present: (1) the temporal variability of the 222 Rn; (2) the set of different daily cycles of 222 Rn identified by comparing maximum and baselines concentrations; (3) the set of local and synoptic meteorological conditions associated with each group of daily cycles, and the location of the potential sources, and (4) the verification of these sources with those identified in the 222 Rn exhalation rate map and the European Map of Uranium. We draw the main conclusions in Section 4.

Study Area
The study area is Bilbao city, in the Basque province (northern Spain, Figure 1). Bilbao has located about 16 km from the Cantabric sea, following the valley of the Nervion river, which acts as a preferential channel for surface winds reaching Bilbao. The relief surrounding Bilbao is dominated by NW-SE oriented folds, with an altitude between 80 m and 700 m ( Figure 1). Bilbao has a humid oceanic climate with a predominance of westerly winds. During the studied period (2017-2018), temperature averages between 11 • C in winter and 21 • C in summer, while annual precipitation averages over 830 mm, with a yearly average of 142 days with at least 1 mm. Previous analysis [21] has demonstrated that this area is typically affected by air masses from the northwest, west, north, northeast, and south, which are combined in spring and summer with the development of mesoscale circulations in surface levels, i.e., sea-land breezes.
teorological datasets, the scheme to calculate air mass trajectories, and the methodologies applied. Then, in Section 3, we present: (1) the temporal variability of the 222 Rn; (2) the se of different daily cycles of 222 Rn identified by comparing maximum and baselines con centrations; (3) the set of local and synoptic meteorological conditions associated with each group of daily cycles, and the location of the potential sources, and (4) the verifica tion of these sources with those identified in the 222 Rn exhalation rate map and the Eu ropean Map of Uranium. We draw the main conclusions in Section 4.

Study Area
The study area is Bilbao city, in the Basque province (northern Spain, Figure 1) Bilbao has located about 16 km from the Cantabric sea, following the valley of the Ner vion river, which acts as a preferential channel for surface winds reaching Bilbao. The relief surrounding Bilbao is dominated by NW-SE oriented folds, with an altitude be tween 80 m and 700 m ( Figure 1). Bilbao has a humid oceanic climate with a predomi nance of westerly winds. During the studied period (2017-2018), temperature averages between 11 °C in winter and 21 °C in summer, while annual precipitation averages ove 830 mm, with a yearly average of 142 days with at least 1 mm. Previous analysis [21] has demonstrated that this area is typically affected by air masses from the northwest, west north, northeast, and south, which are combined in spring and summer with the devel opment of mesoscale circulations in surface levels, i.e., sea-land breezes. This area could be classified as low natural radioactivity background based on the value of U contents in soil, gamma dose rate, and radon fluxes. According to the European Map of Uranium in soil (https://remon.jrc.ec.europa.eu/About/Atlas-of-Natural-Radiation/Digital-Atlas/Uranium-in-soil/Uranium-concentration-in-soil-, accessed on 10 October 2022), Bilbao area shows values of uranium between 1.6 and 2.4 mg/kg. The map of natural gamma radiation in Spain-MARNA shows values below 10 µR/h (https://www.csn.es/en/mapa-de-radiacion-gamma-natural-marna-mapa, accessed on 10 October 2022). Looking to the map obtained by [14], which provides detailed information on the 222 Rn flux term in Spain, with a spatial resolution of 0.02 deg by downscaling the Spanish terrestrial gamma map (MARNA), this area presents an average 222 Rn flux of less than 40-50 Bq/(m 2 h), while in nearby areas and to the east there is a higher average of 222 Rn flux, at around 80-90 Bq/(m 2 h). Therefore, these low natural radioactivity levels make this region could be considered as a background station, and then, useful to identify the regional and remote sources of 222 Rn.

Radon Measurements
The 222 Rn measurements have been carried out by the Faculty of Engineering of Bilbao (43.26 • N, −2.9 • W). The radiological station located in Bilbao is a combined Particulate and Iodine Monitoring system Type 9850-6 (Moving Filter Monitor BAI 9100 D, IOD-131 monitor BAI9103-2 and Gamma Dose Rate Detector LB6360, Berthold trademark positioned on the roof of a building at 10 m above ground level [22]. At this height, the thoron contribution can be assumed negligible, so we do not expect fluctuations in radon/thoron decay product concentration [10]. This equipment has been checked empirically in Berthold every five years to determine any effect on the calibration factor in order to keep consistent high-quality measurements. Figure 1 shows the moving Filter Monitor BAI 9100 D with air caption and gamma dose rate detector LB6360. This Monitor BAI 9100 D consists of the following blocks: (a) a dust collection unit BAI9100D and (b) an Alpha-/Beta-Detector BAI9300AB with ZnS coated Plastic Scintillator and Preamplifier unit LB2030 for simultaneous, separate Alpha and Beta activity measurement and Radon activity concentration obtained by the Alpha Beta Pseudo-coincidence Difference (ABPD) method. The ABPD method uses the specific measurement of the Bismuth-214 decay into Polonium-214 and the Bismuth-212 decay into Polonium-212 to compensate for natural activity. A good pseudo-coincidence stage should contain a second circuit to detect radon coincidences. These should be used to compensate for the pseudo-coincidence stage for random coincidences at higher count rates to avoid overcompensation i.e., suppression of potential artificial radioactivity events. The ABPD module block diagram has two stages implemented the pseudo-coincidence stage ( 214 Bi/ 214 Po and 212 Bi/ 212 Po decays) and the random (A/B) coincidence stage. Gamma Dose Rate Detector LB6360 is a proportional counter with a calibration factor of 0.105 µGy h per cps. The sampling time is 10 min, and the calibration is completed annually with 241 Am and 36 Cl [22]. The methodology can introduce uncertainties in radon measurements in the order of ±10%.
The estimation of radon concentration from radon progenies requires the application of an equilibrium factor, which is known to have great variability, from 0.13 to 0.91 [10]. Measurements of the equilibrium factor were not available. So, it was decided to carry out our analysis without correcting for the equilibrium factor considering the output of the instruments, being aware that referring to 222 Rn concentration we are referring to 222 Rn estimated through its progeny ( 214 Bi and 214 Po).
Working with data, it is almost impossible to acquire them without gaps due to power failure or other technical problems. Hourly measurements from 2017 to 2018 are selected based on the availability of 222 Rn measurements (82% in 2017 and 88% in 2018), thus guaranteeing the largest statistical sample. In this case, there is one main gap in data, in August 2017 (Figure 2a).

Meteorological Parameters
The following meteorological variables were continuously measured in situ using a Campbell automatic weather station together with the CR1000 datalogger: temperature, relative humidity, precipitation, and wind speed and direction. These measurements were obtained at the same location as the 222 Rn equipment, and with a time resolution of one hour. From this set of hourly measures, the quality criterion of 75% was applied to calculate daily values, i.e., at least 75% of the hourly and daily records. tion is 18.7 Bq/m at 8:00 UTC, which is associated with the accumulation of Rn in the lower layers of the atmosphere due to the enhancement of atmospheric stability. This process favours the accumulation of freshly exhaled radon in lower layers [36]. After thi increase at night, 222 Rn concentrations decrease during the daytime, reaching a minimum value of 7.7 Bq/m 3 on average at 17:00-18:00 UTC. This decrease is due to solar radiation which breaks the night-time stability and increases the vertical radon dilution.

Air Mass Trajectories and Cluster, TSA and PSFC Methods
Kinematic 3-D back trajectories were computed for a period of 2 years (2017-2018) eight times per day (at 00:00, 03:00, 06:00, 09:00; 12:00; 15:00; 18:00; 21:00 UTC). This set of backward trajectories was calculated by the HYSPLIT model, which was run with meteorological data from the Global Data Assimilation System reanalysis archive (http://ready.arl.noaa.gov/, accessed on 10 September 2022) with 1-degree global latitudelongitude projection and a temporal resolution of 6 h. These files are widely used by the scientific community to study the regional atmospheric circulation [23] and to perform a qualitative analysis of the different source regions influencing the atmospheric 222 Rn concentrations [24][25][26]. Trajectories were computed at 100 m above ground level (agl) and with a period of 96 h. This time period agrees with the half-life of radon (3.8 days), providing unambiguous evidence of terrestrial influence over the sampling point [27,28]. This time is also used in previous studies over the Iberian Peninsula dealing with long-range transport [28][29][30][31].
Based on the set of back trajectories calculated, trajectory cluster analysis, trajectory source apportionment (TSA), and the potential source contribution function (PSCF) were applied to investigate the source-receptor relationship for the 222 Rn concentrations in Bilbao. Details regarding the HYSPLIT k-means clustering algorithm are provided at the ARL NOAA site (ready.arl.noaa.gov/HYSPLIT.php, accessed on 1 October 2022). The TSA is a statistical approach that computes average concentrations from various directions, and hence, is useful to evaluate the effect of air masses from various directions on atmospheric concentrations of different substances [32]. In the present study, we use the calculated back trajectories and 3-h 222 Rn concentrations to identify those regions that highly impact the receptor at Bilbao. Equations used in [33] were used to calculate the relative contribution (%C) from each of the sector 12 sectors of 30 • each, numbered clockwise from 0 • north.
On the other hand, the PSCF method estimates pollutant sources in upwind areas by analysing airflow trajectories and a given concentration threshold [34]. From the base that air mass alone cannot be used to identify specific source regions of pollutants [35], this method links residence time with high concentrations through a conditional probability field and provides a value to each grid cell, indicating high PSCF values the location of potential emission sources [36] in the present case, of 222 Rn over Bilbao: where N ij is the total number of airflow trajectories' endpoints that fall in the ij th grid and n ij is the total number of airflow trajectories' endpoints for which the measured 222 Rn concentration exceeds a given threshold in the same grid. To calculate the PSCF, the whole geographic region covered by the backward trajectories was divided into a gridded array, with the centre of Bilbao as the midpoint and containing grid cells of 0.5 • × 0.5 • . It is important to note that a grid with no endpoints cannot be identified as a source area in the analysis even though there are known emission sources in the grid cell, as well as studies, have demonstrated that great uncertainty exists in the calculation result when N ij is extremely small. To better reflect this uncertainty the weight function (W ij ) defined in [33] was applied: Then the value of PSCF is interpreted as the probability where the concentration of 222 Rn higher than the creation level was related to the passage of air parcels through the ith cell. These cells are indicative of areas of high potential contributions for 222 Rn in Bilbao.  Figure 2 also shows the composite diurnal cycles for 222 Rn concentrations during the whole sampling period. On average, the night-time maximum of the hourly concentration is 18.7 Bq/m 3 at 8:00 UTC, which is associated with the accumulation of 222 Rn in the lower layers of the atmosphere due to the enhancement of atmospheric stability. This process favours the accumulation of freshly exhaled radon in lower layers [36]. After this increase at night, 222 Rn concentrations decrease during the daytime, reaching a minimum value of 7.7 Bq/m 3 on average at 17:00-18:00 UTC. This decrease is due to solar radiation, which breaks the night-time stability and increases the vertical radon dilution.

Characterization of 222 Rn Concentrations
This general daily cycle presents seasonal differences (Figure 2b), in line with the corresponding seasonal box-and-whisker plots displayed in Figure 3. Meteorological seasons were defined as winter, January to March, spring April to June, summer July to September, and autumn October to December. There is a large increase in the maximum daily values from Winter (14.2 Bq/m 3 ) to Summer (26.2 Bq/m 3 ), being similar to the minimum ones, reaching the highest in Autumn (10.2 Bq/m 3 ). The period of high 222 Rn concentrations coincides with the largest influence of stable synoptic conditions (e.g., Azores anticyclone) over the Iberian Peninsula, which favours the development of mesoscale processes, such as sea-land breezes, and hence, increasing recirculation and stagnation processes of surface air. In Figure 3, the highest hourly concentrations, close to 75 Bq/m 3 , are mainly reached from August to early October, which causes an annual pattern with low radon levels in spring months and an increase in autumn. In this sense, it is necessary to mention the considerable difference between maximum 222 Rn concentrations and 95th percentile values in each season, ranging from 30.1 Bq/m 3 (Winter) to 40.2 Bq/m 3 (Summer). The maximum value tends roughly to be twofold or even threefold compared to the 95th percentile. This suggests the existence of relative high-radon events with less than 5% occurrence throughout the year, either in summer or winter, which reach similar maximum hourly concentrations of 74.7 Bq/m 3 and 71.6 Bq/m 3 respectively. These high concentrations observed in Winter can be associated with the small vertical development of the nocturnal atmospheric boundary layer.
seasons were defined as winter, January to March, spring April to June, summer July to September, and autumn October to December. There is a large increase in the maximum daily values from Winter (14.2 Bq/m 3 ) to Summer (26.2 Bq/m 3 ), being similar to the minimum ones, reaching the highest in Autumn (10.2 Bq/m 3 ). The period of high 222 Rn concentrations coincides with the largest influence of stable synoptic conditions (e.g., Azores anticyclone) over the Iberian Peninsula, which favours the development of mesoscale processes, such as sea-land breezes, and hence, increasing recirculation and stagnation processes of surface air. In Figure 3, the highest hourly concentrations, close to 75 Bq/m 3 , are mainly reached from August to early October, which causes an annual pattern with low radon levels in spring months and an increase in autumn. In this sense, it is necessary to mention the considerable difference between maximum 222 Rn concentrations and 95th percentile values in each season, ranging from 30.1 Bq/m 3 (Winter) to 40.2 Bq/m 3 (Summer). The maximum value tends roughly to be twofold or even threefold compared to the 95th percentile. This suggests the existence of relative high-radon events with less than 5% occurrence throughout the year, either in summer or winter, which reach similar maximum hourly concentrations of 74.7 Bq/m 3 and 71.6 Bq/m 3 respectively. These high concentrations observed in Winter can be associated with the small vertical development of the nocturnal atmospheric boundary layer. Figure 3. Seasonal box plots of 222 Rn hourly concentrations at Bilbao. The rectangle represents 50% of data (interquartile range from 25th to 75th percentile), the small cross identifies the mean, the continuous horizontal line inside the rectangle identifies the median (50th percentile), the squares and circles identify the 90th and 10th percentiles respectively, and the whiskers extend between the 95th and 5th values.

Baseline and Maximum Daily 222 Rn Concentrations: Daily Differences
One way to investigate and identify the impact of local and regional contributions to 222 Rn activity concentrations is to check the variability of radon daily cycles and to analyse the meteorological conditions associated with different amplitudes. The amplitude is defined as the difference between the maximum and minimum daily 222 Rn concentra- Figure 3. Seasonal box plots of 222 Rn hourly concentrations at Bilbao. The rectangle represents 50% of data (interquartile range from 25th to 75th percentile), the small cross identifies the mean, the continuous horizontal line inside the rectangle identifies the median (50th percentile), the squares and circles identify the 90th and 10th percentiles respectively, and the whiskers extend between the 95th and 5th values.

Baseline and Maximum Daily 222 Rn Concentrations: Daily Differences
One way to investigate and identify the impact of local and regional contributions to 222 Rn activity concentrations is to check the variability of radon daily cycles and to analyse the meteorological conditions associated with different amplitudes. The amplitude is defined as the difference between the maximum and minimum daily 222 Rn concentrations. Therefore, to perform this analysis, the magnitude of the "background" and maximum daily values are needed to be set. The background value represents the starting point upon which the local sources and sinks act to produce the final observed concentration. Due to the location of Bilbao, in a coastal area but distant 16 km to the Cantabrian Sea, 222 Rn activity concentrations associated with the maritime wind sector are relatively high, and hence, the methodology based on selected hourly atmospheric radon concentration measurements in the maritime wind sector [27] is not applicable in this case. For this reason, and in the present study, we have used the methodology used in [5,24] which is based on taking as reference the average of the hourly 222 Rn measurements in the afternoon 8 of 17 minimum period (16:00-20:00 UTC) ( Figure 2) as our daily background reference value at each of the sites. Figure 4 (orange line) displays the daily 222 Rn background evolution in Bilbao during the whole sampling period, which has an average of 6.7 ± 0.2 Bq/m 3 . On average, the background value presents low and similar values from February to August (at about 5 Bq/m 3 ), while it progressively increases in September-October (9-10 Bq/m 3 ) and it reaches the maximum on December 13 Bq/m 3 . This evolution, for instance, is not similar to the one obtained in [5] in Huelva (southern Iberian Peninsula), in which 222 Rn background increases from May to September, with a decrease from October to November. This difference can be justified according to differences in the sampling sites and the set of meteorological conditions at different scales influencing 222 Rn activity concentrations. relatively high, and hence, the methodology based on selected hourly atmospheric radon concentration measurements in the maritime wind sector [27] is not applicable in this case. For this reason, and in the present study, we have used the methodology used in [5,24] which is based on taking as reference the average of the hourly 222 Rn measurements in the afternoon minimum period (16:00-20:00 UTC) (Figure 2) as our daily background reference value at each of the sites. Figure 4 (orange line) displays the daily 222 Rn background evolution in Bilbao dur ing the whole sampling period, which has an average of 6.7 ± 0.2 Bq/m 3 . On average, the background value presents low and similar values from February to August (at about 5 Bq/m 3 ), while it progressively increases in September-October (9-10 Bq/m 3 ) and it reaches the maximum on December 13 Bq/m 3 . This evolution, for instance, is not similar to the one obtained in [5] in Huelva (southern Iberian Peninsula), in which 222 Rn background increases from May to September, with a decrease from October to November. This dif ference can be justified according to differences in the sampling sites and the set of me teorological conditions at different scales influencing 222 Rn activity concentrations.   1  16  31  46  61  76  91  106  121  136  151  166  181  196  211  226  241  256  271  286  301  316  331  346  361  376  391  406  421  436  451  466  481  496  511  526  541  556  571  586  601  616  631  646  661  676  691   Once defined and calculated, daily differences between background and maximum daily values are obtained. On average during the whole sampling period, this difference is 11.3 Bq/m 3 and it presents a monthly variability over the whole sampling period. Figure 4 also displays the monthly average of this difference. It is clear the rising tendency from March to August, and how during the warm period the highest values are reached. This evolution can be justified by the progressive increase from May to September in the frequency of mesoscale circulations favoured by the dominance of synoptic stable conditions causing low gradient pressure forcing in this region [21], while, on the other hand, the decrease is likely due to the more unstable meteorological conditions registered from October onwards, with more influence of synoptic conditions.
The variable under study is the diurnal cycle of hourly 222 Rn concentrations associated with different daily backgrounds and maximum differences. We have grouped each full day of data depending on this difference. 222 Rn daily cycles were classified into six different ranges using the values of the percentiles of daily differences, i.e., P5 (the lowest values), P25, P50 (the intermediate values), P75, and P95 (the highest values). The ranges are: less than −0.7 Bq/m 3 , from −0.7 to 2.9 Bq/m 3 , from 2.9 to 8.5 Bq/m 3 , from 8.5 to 17.4 Bq/m 3 , from 17.4 to 31.9 Bq/m 3 and greater than 31.9 Bq/m 3 . The use of percentiles as a reference has been previously used in the characterization of the local, mesoscale and synoptic conditions associated with different ranges of substances in the atmosphere [37,38]. Once defined and calculated, daily differences between background and maximum daily values are obtained. On average during the whole sampling period, this difference is 11.3 Bq/m 3 and it presents a monthly variability over the whole sampling period. Figure 4 also displays the monthly average of this difference. It is clear the rising tendency from March to August, and how during the warm period the highest values are reached. This evolution can be justified by the progressive increase from May to September in the frequency of mesoscale circulations favoured by the dominance of synoptic stable conditions causing low gradient pressure forcing in this region [21], while, on the other hand, the decrease is likely due to the more unstable meteorological conditions registered from October onwards, with more influence of synoptic conditions.
The variable under study is the diurnal cycle of hourly 222 Rn concentrations associated with different daily backgrounds and maximum differences. We have grouped each full day of data depending on this difference. 222 Rn daily cycles were classified into six different ranges using the values of the percentiles of daily differences, i.e., P5 (the lowest values), P25, P50 (the intermediate values), P75, and P95 (the highest values). The ranges are: less than −0.7 Bq/m 3 , from −0.7 to 2.9 Bq/m 3 , from 2.9 to 8.5 Bq/m 3 , from 8.5 to 17.4 Bq/m 3 , from 17.4 to 31.9 Bq/m 3 and greater than 31.9 Bq/m 3 . The use of percentiles as a reference has been previously used in the characterization of the local, mesoscale and synoptic conditions associated with different ranges of substances in the atmosphere [37,38]. Figure 5 displays the daily evolution of 222 Rn concentrations according to each range. The maximum 222 Rn concentrations increase from 6.1 Bq/m 3 (Range 2) to 49.5 Bq/m 3 (Range 6), while the minimum ranges from 4.7 Bq/m 3 (Range 2) to 11.4 Bq/m 3 (Range 1). It is interesting to point out how in Range 1 the 222 Rn concentrations during the daytime are higher than at night, while in the rest of the daily cycles, the concentration of 222 Rn increases during the night and decreases during the daytime.

Meteorological Link with Daily Cycles: From Local to Synoptic
The temporal variability of local meteorological factors influences radon activity concentration in the air [39]. Correlation coefficients for daily differences and meteorological parameters were then computed as a measure of the strength of the relationship between daily differences and local meteorological factors [40]. This analysis is based on the set of non-zero pairs between daily radon differences and daily meteorological values, so pairs in which either radon or meteorological values were empty are not included.

Meteorological Link with Daily Cycles: From Local to Synoptic
The temporal variability of local meteorological factors influences radon activity concentration in the air [39]. Correlation coefficients for daily differences and meteorological parameters were then computed as a measure of the strength of the relationship between daily differences and local meteorological factors [40]. This analysis is based on the set of non-zero pairs between daily radon differences and daily meteorological values, so pairs in which either radon or meteorological values were empty are not included. A total of 645 days (88% of days within the period 2017-2018) were used. A positive correlation with temperature (r = 0.2, at 0.05 significance level (p = 4.2 × 10 −7 )) is obtained, which is mainly associated with the occurrence of high temperatures (hence higher radon flux from the soil) in those months with the maximum differences. On the other hand, negative correlations of radon with wind speed (r = −0.3, at 0.05 significance level (p = 2.8 × 10 −8 )) and rainfall (r = −0.2, at 0.05 significance level (p = 5.6 × 10 −18 )) agree with the dispersive nature of the high wind speed and the decrease in the exhalation due to rainfall. In detail, Figure 6 shows the influence of rainfall episodes with precipitation above 0.1 mm on daily differences of 222 Rn concentrations. This figure reports that similar daily values are observed under intense rainfall events, which is expected due to rainfall patterns affecting water-table depth and, consequently, soil-moisture content and therefore radon emanations [41]. This figure displays the progressive increase in daily differences between the number of rainy days and the accumulated rainfall.
concentrations. In this area, one pattern of sea-land breeze has been identified [42]. Briefly, the breeze presents nocturnal flows from the southwest from midnight until midday, followed by the arrival of flows from the northwest overnight. Wind intensity is also following this change in direction, showing a well-marked daily cycle in range 6, while on the contrary, it presents similar daily values in Range 1. On average, wind intensity decreases from 1.4 m/s (Range 1) to 0.8 m/s (Range 6). These conditions would represent a change from near stagnation in the case of Range 1, to a more synoptic influence in the case of Range 6 in this area. In other words, when the wind speed is low, local contributions may be of importance to understanding 222 Rn activity concentrations, while under strong winds regional transport and distant sources should be considered. This influence of sea-land breezes agrees with previous studies in which this mesoscale circulation plays an important role in the transport and distribution of pollutants [43,44].  We now investigate the wind dynamics at different scales associated with each type of daily cycle. Figure 7 shows the corresponding daily average wind evolution for each range during the sampling period. This figure displays an evolution from a constant arrival of north-westerly flows associated with Ranges 1-3 to a progressive combination of south-western winds during the morning and north-western flows overnight from range 4 to range 6, i.e., there is a change in surface winds following the increase in 222 Rn activity concentrations. In this area, one pattern of sea-land breeze has been identified [42]. Briefly, the breeze presents nocturnal flows from the southwest from midnight until midday, followed by the arrival of flows from the northwest overnight. Wind intensity is also following this change in direction, showing a well-marked daily cycle in range 6, while on the contrary, it presents similar daily values in Range 1. On average, wind intensity decreases from 1.4 m/s (Range 1) to 0.8 m/s (Range 6). These conditions would represent a change from near stagnation in the case of Range 1, to a more synoptic influence in the case of Range 6 in this area. In other words, when the wind speed is low, local contributions may be of importance to understanding 222 Rn activity concentrations, while under strong winds regional transport and distant sources should be considered. This influence of sea-land breezes agrees with previous studies in which this mesoscale circulation plays an important role in the transport and distribution of pollutants [43,44].
To complete the analysis, and to identify the synoptic scenarios associated with each daily range, Figure 8 displays the results of the cluster analysis for the 96-h backward trajectory of air mass over Bilbao for each range of 222 Rn daily differences previously defined. This figure shows the pathway in the horizontal and vertical directions followed by an air parcel upwind from Bilbao. Considering the results in the horizontal direction, in general, the results follow the fact that the higher the daily difference is, the shorter the trajectories are, and the larger displacements over land are present. The air mass analysis for Range 1 indicates the prevalence of westerly flows (70%) with origins and pathways over the Atlantic Ocean, and the arrival of northern circulations, with smaller displacement (30%). The influence of both circulations (westerly and northerly) remains in the backward trajectory cluster associated with Range 2, while in Range 3, air mass clusters representing the arrival of southern air masses and hence with more displacement over land start to be identified. In this range, the arrival of western circulations decreases. Ranges 4-6 point out the continuity in the arrival of westerly and northerly circulations but with less displacement than in previous ranges, i.e., air mass trajectories present less wind velocity. This trend is observed from Range 4 to 6, in which the most influence of nearby circulations (41%) in combination with the arrival of northerly flows (50%) is shown. This combination of airflows, showing the arrival of continental air masses from the northeast and those with small displacement, which agree with the development of mesoscale circulations in this area, and hence, in line with the sea-land breeze pattern identified in Figure 7, are those more influencing in reaching the highest differences in the daily cycles. Therefore, the meteorological scenario mostly influencing reaching the largest daily differences is associated is mostly associated with dry periods, relatively high temperatures and low wind speeds, and two different wind dynamics (1) the arrival of nearby and southerly air masses with the development of sea-land breezes, and (2) arrival of northeast airflow patterns. To complete the analysis, and to identify the synoptic scenarios associated with each daily range, Figure 8 displays the results of the cluster analysis for the 96-h backward trajectory of air mass over Bilbao for each range of 222 Rn daily differences previously defined. This figure shows the pathway in the horizontal and vertical directions followed by an air parcel upwind from Bilbao. Considering the results in the horizontal direction, in general, the results follow the fact that the higher the daily difference is, the shorter the trajectories are, and the larger displacements over land are present. The air mass analysis for Range 1 indicates the prevalence of westerly flows (70%) with origins and pathways over the Atlantic Ocean, and the arrival of northern circulations, with smaller displacement (30%). The influence of both circulations (westerly and northerly) remains in the

Identification of Potential Sources
Once the meteorological scenarios associated with high radon peaks have been defined, this section presents the results obtained in the identification of potential sources of radon over Bilbao. In this sense, an important feature of Bilbao is its proximity to the coast, i.e., coastal sites present lower radon average concentrations than inland sites [14], and the low natural radioactivity levels in this region (Section 2.1). The TSA method is applied by using back trajectories and 3-h 222 Rn concentrations. This TSA analysis (Figure 9) firstly confirms the main influence of continental winds on the radon concentrations in Bilbao, and secondly, that air masses from the northeast (sectors 2-3) and from the southwest (sectors 8,9,10) are mainly associated with the highest radon concentrations, due to their large relative contribution to 222 Rn measured in Bilbao. These TSA results, then, reflect the highly polluted nature of the combination of southern and north-eastern air mass trajectories over Bilbao, remarking the results of the backward trajectory analysis (Figure 8). There is also a not negligible contribution of pure maritime sectors which could be associated with nearby circulations identified in Figure 8, and hence, associated with mesoscale circulations of sea-land breezes developed in the area. Those areas more influencing the highest 222 Rn concentrations registered at Bilbao, i.e., potential emission sources, are estimated by means of the PSCF model ( Figure 9). In the present study, this map reports the areas with the longest residence time of air masses associated with the highest 222 Rn activity concentrations. Considering previous studies [45,46] in which maximum ABL heights between 950 m in summer and 700 m (13 LT) in winter, and minimum heights of about 100 m were reported, we have only considered for the present analysis those trajectory endpoints at a height lower than 900 m. Figure 9 shows how the potential sources of 222 Rn in Bilbao are mainly placed in the south, as well as on the French east coast. These areas correspond with the areas with the highest radon flux as well as the area with the highest uranium contents in soil, as it is shown in Figure 10

Conclusions
In this study, local meteorological parameters and air mass backward trajectory cluster analysis, TSA, and PSCF methods were used to investigate local and synoptic meteorological influences on the temporal variability, transport pathways, and potential sources of 222  Therefore, combining the TSA and PSCF results would present two scenarios in which the highest radon concentrations would be measured in Bilbao. The first one is in which air masses from the northeast could transport radon concentrations from sources placed in the west of France to Bilbao, and the second one is in which south air masses from the south sweep sources placed in the Iberian Peninsula. In both scenarios, the development of sea-land breezes helps in reaching these peak concentrations due to the combination of oceanic air masses which bring markedly smaller concentrations, and continental air masses, which leads to the return of higher 222 Rn concentrations.

Conclusions
In this study, local meteorological parameters and air mass backward trajectory cluster analysis, TSA, and PSCF methods were used to investigate local and synoptic meteorological influences on the temporal variability, transport pathways, and potential sources of 222 Rn in the north of the Iberian Peninsula during the period 2017-2018. 222 Rn levels were characterized by a diurnal cycle with an early morning maximum and a minimum in the afternoon, although the combination of local and regional meteorological patterns causes differences in seasonal daily patterns, as well as in daily cycle amplitudes. For different daily amplitudes, backward trajectories were analysed. The largest daily amplitude of daily 222 Rn concentrations is found under the development of mesoscale circulations in this area, which favour the progressive accumulation of 222 Rn due to the scarce air mass circulation, which agrees with the influence of south and northeast sources. These results are taken as a reference to analyse the great significance of mesoscale circulations in explaining the variability of 222 Rn, although further investigation is needed to confirm the influence of regional sources on 222 Rn concentrations and to determine the long-term influence of individual meteorological parameters on radon levels in the air.