Variability of Cosmogenic 35 S in Rain—Resulting Implications for the Use of Radiosulfur as Natural Groundwater Residence Time Tracer

: Information about groundwater residence times is essential for sustainable groundwater management. Naturally occurring radionuclides are suitable tools for related investigations. While the applicability of several long-lived radionuclides has been demonstrated for the investigation of long residence times (i.e., years, decades, centuries and more), studies that focus on sub-yearly residence times are only scarcely discussed in the literature. This shortage is mainly due to the rather small number of radionuclides that are generally suitable for the purpose and show at the same time adequately short half-lives. A promising innovative approach in this regard applies cosmogenic radiosulfur ( 35 S). 35 S is continuously produced in the stratosphere from where it is conveyed to the troposphere or lower atmosphere and ﬁnally transferred with the rain to the groundwater. As soon as the meteoric water enters the subsurface, its 35 S activity decreases with an 87.4 day half-life, making 35 S a suitable time tracer for investigating sub-yearly groundwater ages. However, since precipitation shows a varying 35 S activity during the year, setting up a reliable 35 S input function is required for sound data evaluation. That calls for (i) an investigation of the long-term variation of the 35 S activity in the rain, (ii) the identiﬁcation of the associated drivers and (iii) an approach for setting up a 35 S input function based on easily attainable proxies. The paper discusses 35 S activities in the rain recorded over a 12-month period, identiﬁes natural and anthropogenic inﬂuences, and suggests an approach for setting up a 35 S input function applying 7 Be as a proxy. management based on evaluation. information


Introduction
The knowledge of groundwater residence times is essential for the sustainable management of groundwater resources. This general fact can be underpinned with three specific aspects: (i) the vulnerability of an aquifer (regarding anthropogenic contamination) depends largely on the travel time needed by the meteoric water to cross the unsaturated soil zone covering the groundwater body. (ii) Sustainable groundwater abstraction management is mainly based on groundwater residence time evaluation. (iii) Groundwater residence time information is essential for evaluating groundwater migration patterns and the associated matter (and contaminant) transport.
Commonly, the investigation of groundwater residence times relies on field surveys that apply environmental tracers, i.e., tracers that occur naturally in the groundwater. Suitable in this regard are components that meet two general requirements: (i) their concentration in groundwater changes as a function of time and (ii) they enter the subsurface (dissolved in the meteoric water or any surface water) either with a time-constant concentration or with a specific time-variant input function.
Powerful residence time tracers include (besides stable isotopes such as 40 Ar, 3 He, δ 2 H, or δ 18 O) a range of ubiquitously occurring radionuclides. A key criterion for selecting site-specifically the most suitable radionuclide is its half-life, which should be in the same range as the groundwater residence time expected for the investigated aquifer domain. Rather long-lived radionuclides (such as 3 H, 14 C, 36 Cl, 39 Ar, 81 Kr, and 85 Kr) are generally suitable for investigating high groundwater ages ( Figure 1) and a great number of studies that aimed at investigating aquifer systems with water residence times between 10 and 10 6 years have been published. In contrast, radiotracer-based investigations that focus on sub-yearly groundwater residence times are scarcely discussed in the literature. This is because only a limited number of ubiquitously occurring radionuclides show a half-life short enough for covering the sub-yearly timeframe. Examples of suitable short-lived radionuclides are the frequently applied 222 Rn (e.g., [1][2][3]) or the radium species 224 Ra and 223 Ra (e.g., [4,5]).
Water 2020, 12, x FOR PEER REVIEW 2 of 15 function of time and (ii) they enter the subsurface (dissolved in the meteoric water or any surface water) either with a time-constant concentration or with a specific time-variant input function. Powerful residence time tracers include (besides stable isotopes such as 40 Ar, 3 He, δ 2 H, or δ 18 O) a range of ubiquitously occurring radionuclides. A key criterion for selecting site-specifically the most suitable radionuclide is its half-life, which should be in the same range as the groundwater residence time expected for the investigated aquifer domain. Rather long-lived radionuclides (such as 3 H, 14 C, 36 Cl, 39 Ar, 81 Kr, and 85 Kr) are generally suitable for investigating high groundwater ages ( Figure 1) and a great number of studies that aimed at investigating aquifer systems with water residence times between 10 and 10 6 years have been published. In contrast, radiotracer-based investigations that focus on sub-yearly groundwater residence times are scarcely discussed in the literature. This is because only a limited number of ubiquitously occurring radionuclides show a half-life short enough for covering the sub-yearly timeframe. Examples of suitable short-lived radionuclides are the frequently applied 222 Rn (e.g., [1][2][3]) or the radium species 224 Ra and 223 Ra (e.g., [4,5]).
The state-of-the-art situation is generally illustrated in Figure 1. The image reveals that the timeframe between about 1 and 12 months is not satisfactorily covered by the radioactive residence time tracers that are conventionally applied. A potential candidate to fill this gap is the naturally occurring radioisotope beryllium-7 ( 7 Be). It shows a half-life of 53.3 days and is hence (at least in this regard) suitable for investigating groundwater residence times of up to about eight months. 7 Be is produced through cosmic ray spallation of mainly oxygen and nitrogen within the upper troposphere and lower stratosphere (UTLS zone) [6][7][8]. There it attaches to aerosols, which are washed out by (dry deposition and) mainly precipitation [9,10]. Since no natural 7 Be sources exist in the subsurface, the 7 Be activity concentration of the meteoric water declines once it enters the ground. However, 7 Be has a high tendency to sorb not only to aerosols in the upper atmosphere but also to the soil matrix. Thus, a substantial and nonquantifiable share of the dissolved beryllium is retarded by the soil matrix [11,12], which (in contrast to its applicability as a tracer for atmospheric dynamics, [10,13]) impedes its applicability as a quantitative aqueous tracer in groundwater.
A promising alternative approach for covering sub-yearly groundwater residence times is based on the application of naturally occurring radioactive sulfur as aqueous tracer ("radiosulfur"; 35 S, [14]). Comparable to 7 Be, the radionuclide is continuously produced by cosmic ray spallation (of 40 Ar) in the stratosphere where it rapidly (ca. one second) oxidizes to sulphate. Due to its short half-life of 87.4 days, some of the produced 35 S decays to 35 Cl during stratosphere/troposphere air exchange. Still, most 35 SO4 2− gets dissolved in meteoric water and is transferred as precipitation to the earth's surface and finally to the groundwater [15]. The state-of-the-art situation is generally illustrated in Figure 1. The image reveals that the timeframe between about 1 and 12 months is not satisfactorily covered by the radioactive residence time tracers that are conventionally applied.
A potential candidate to fill this gap is the naturally occurring radioisotope beryllium-7 ( 7 Be). It shows a half-life of 53.3 days and is hence (at least in this regard) suitable for investigating groundwater residence times of up to about eight months. 7 Be is produced through cosmic ray spallation of mainly oxygen and nitrogen within the upper troposphere and lower stratosphere (UTLS zone) [6][7][8]. There it attaches to aerosols, which are washed out by (dry deposition and) mainly precipitation [9,10]. Since no natural 7 Be sources exist in the subsurface, the 7 Be activity concentration of the meteoric water declines once it enters the ground. However, 7 Be has a high tendency to sorb not only to aerosols in the upper atmosphere but also to the soil matrix. Thus, a substantial and nonquantifiable share of the dissolved beryllium is retarded by the soil matrix [11,12], which (in contrast to its applicability as a tracer for atmospheric dynamics, [10,13]) impedes its applicability as a quantitative aqueous tracer in groundwater.
A promising alternative approach for covering sub-yearly groundwater residence times is based on the application of naturally occurring radioactive sulfur as aqueous tracer ("radiosulfur"; 35 S, [14]). Comparable to 7 Be, the radionuclide is continuously produced by cosmic ray spallation (of 40 Ar) in the stratosphere where it rapidly (ca. one second) oxidizes to sulphate. Due to its short half-life of 87.4 days, some of the produced 35 S decays to 35 Cl during stratosphere/troposphere air exchange. Still, most 35 SO 4 2− gets dissolved in meteoric water and is transferred as precipitation to the earth's surface and finally to the groundwater [15]. 35 SO 4 2− activities in rain range generally between about 10 and 150 mBq/L ( [16][17][18][19]; own measurements). As soon as the 35 S containing meteoric water (C 0 ) enters the subsurface, 35 S decay is not supported by 35 S production anymore, which makes the decline of the initial 35 S activity concentration in the groundwater (C t ) an indicator for its residence time (t) in the ground (determined by the 35 S decay constant λ) (Equation (1)).
In contrast to 7 Be, sulphate 35 SO 4 2− is not retarded by the soil or aquifer matrix. This makes radiosulfur a residence time tracer suitable for groundwater ages between about three and nine months (i.e., between one and three 35 S half-lives, [20][21][22]). Still, scanning the literature reveals that only a few related case studies have been published so far. These studies are generally limited to high geographical elevations, where snowmelt is the dominant hydrological recharge event. This boundary condition limits 35 S input into the groundwater to the peak snowmelt, i.e., it simplifies the 35 S input to a distinct value [20,21,23,24]. This neglects the fact that 35 S is likely to show considerable variability in the rain over the course of a year (e.g., [25,26]).
This variability is mainly due to large-scale atmospheric circulation dynamics, i.e., due to stratosphere/troposphere air mass exchange and convective tropospheric air circulation, which are, to a large degree, bound to seasonal patterns in atmospheric processes [10]. Besides such temporal variations, significant latitudinal variations of 35 S in the atmosphere and the rain are also likely. The varying deflection of cosmic rays causes this lateral variability due to the Earth's magnetic field's increased strength closer to the equator [10]. Another influential factor for 35 S variability in the rain is the rainfall dynamics, i.e., sulphate washout by precipitation [27,28]. The resulting regional and seasonal variability of 35 S in the rain suggest that the experiences reported for the use of 35 S as aqueous residence time tracer in alpine/subalpine watersheds are of only limited use in moderate climates. That implies that the variability of the 35 S activity in the rain (and the associated drivers) needs to be investigated and understood more comprehensively. Setting up region-specific long-term 35 S input functions is mandatory. Understanding the challenges related to this requirement calls for (i) an exemplary systematic long-term investigation of the 35 S activity in the rain in a certain study region, (ii) the identification of the drivers of the temporal long-term 35 S variability and (iii) the suggestion of an easily attainable 35 S proxy that can be applied for supporting the setup of a regional 35 S input function (the latter is desirable because long-term 35 S datasets are hardly ever available and 35 S in water analysis is rather laborious).
The presented paper discusses 35 S activities in the rain recorded over a 12-month period at a distinct location in Germany and suggests an approach regarding the setup of a regional 35 S input function. The approach applies 7 Be in atmospheric air as 35 S proxy. Furthermore, the results are backed and evaluated based on the time series of the aerosol optical depth (AOD) and the activity concentration of naturally occurring 212 Pb in atmospheric air.

Radiosulfur Sampling and Measurement
All 35 S data of the study was attained in a sampling campaign that aimed at collecting samples from as many major rain events as possible over the course of 12 months (August 2019-July 2020). A sampling station was installed, uncovered by trees, in a rural environment close to the city of Leipzig, Germany. The area is located at an elevation of 155 masl, characterized by an average annual rainfall of 550 mm, and an average annual temperature of 11 • C.
All rain samples were collected from a 31 m 2 inclined laminated rain collector plane via a downpipe into a 200-litre HDPE container. Hence, all rain events of up to 6.5 mm precipitation (88% of the 60 sampled rain events) were quantitatively collected. The container was isolated from any other environmental depositions besides rain. The rain collector surface, the downpipe and the container were cleaned regularly, during autumn (leaves) and springtime (pollen) and after storm events.
All related meteorological data was recorded at a professional meteorological station located only 3 km from the rain sampling location. Hence, all samples could be associated with specific rain events and their intensity.
Each rain sample was representatively transferred from the 200-litre sample container into a 20-litre plastic canister. Since no elevated sulphate concentrations were expected in the samples [29], all sample processing and measurement followed the suggestions for low-sulphate sample preparation for liquid scintillation counting (LSC) given by [30]. A Quantulus GCT 6220 liquid scintillation counter was used for all 35 S activity measurements. To allow reasonable statistical reliability of the counting results (counting error < 1%), each measurement lasted 24 h. The detection background was counted and subtracted from the sample counts by measuring a 35 S-dead (and 3 H-dead) background vial before each measurement. To optimize the signal-to-noise-ratio of the measurement results, the specific LSC detection options were set as given in Table 1.

Using Atmospheric 7 Be as 35 S Proxy
In Central Europe we usually observe increased activity concentrations of 7 Be in the lower atmosphere during the warm and dry summer season. The main reasons for this annual pattern are the following.
(i) The tropopause, i.e., the boundary layer between the stratosphere and troposphere rises during the warm season. This upward shift increases the import of 7 Be from the UTLS zone to the lower atmosphere [13,31]. (ii) The zone between the two latitudinal circulation features the Hadley Cell (in the northern hemisphere in the south) and the Ferrel Cell (in the north) is called the Hadley Ferrel Divergence Zone (HFDZ). Within the HFDZ air masses are continuously conveyed from the upper part to the lower part of the troposphere ( Figure 2) [32]. While (in the northern hemisphere) the HFDZ is generally active at a latitude of about 30 • north it migrates, with the start of the warm season, northward and approaches Central Europe annually during the summer.
As discussed in Section 1, both 7 Be and 35 S are primarily created in the lower stratosphere both triggered by cosmic ray spallation. It can hence be assumed that their production rates correlate. Furthermore, both nuclides are conveyed by the same abovementioned large-scale atmospheric processes to the troposphere and lower atmosphere as illustrated in Figure 2. Hence, it was postulated that exceptionally high (or low) 7 Be activities in the lower atmosphere indicate the simultaneous presence of exceptionally high (or low) 35 S activities in the rain.

Using the Aerosol Optical Depth (AOD) and 212 Pb in Air as Supporting Data
Besides the two large-scale atmospheric processes mentioned above in points (i) and (ii), a third reason can be given for the increase of the activity concentration of 35 S in the rain during the warm and dry season: (iii) a less intense 35 S washout from the atmosphere by rain. The less rain falls, the less intense is the washout, which results in an increase of the particle concentration in the rain when a rare precipitation event occurs. The triggering dry conditions are usually accompanied by an increased rate of resuspension of mineral dust from the earth's surface (windblown dust). The intensity of the latter effect can be evaluated based on two parameters: the aerosol optical depth (AOD) and the 212 Pb activity concentration in air.
The AOD indicates the attenuation of incoming solar radiation in the atmosphere by mineral dust, sea salts, volcanic ash, smoke from wildfires, pollution from factories and others. The potential sources of these attenuators indicate that the AOD is (regionally and temporarily) impacted by natural extreme events such as dust storms (with a global hotspot on the Arabian Peninsula) and/or anthropogenic activities such as land clearing fires (with a global hotspot in the Amazon Basin), or human-made air pollution (with global hotspots in northern India and eastern China). However, in Central Europe the impact of such effects on the AOD is less pronounced (although windblown atmospheric dust with a remote origin, such as northern Africa, Sahara, can be an issue). Besides, choosing the AOD wavelengths 1640 and 1020 nm for our investigation allowed focusing on the impact of mineral dust.
The AOD is continuously monitored by NASA at several AERONET sites worldwide. For our study, we used a dataset (covering the complete 12-month period of the 35 S sampling campaign) recorded as daily averages at an altitude of 115 masl at the Institute for Tropospheric Research, which is located about 30 km north-east of our rain sampling station. 212 Pb is a short-lived (half-life 12.64 h) progeny of the 232 Th decay chain and is easily detectable by gamma spectrometry due to its strong gamma line at 238.6 keV. Thorium and its progeny occur as a minor constituent of most soils on earth. Therefore, the 212 Pb activity concentration in the atmosphere can also be used as an indicator for the presence of terrestrial mineral dust (and hence for its washout by precipitation). Comparable to 7 Be, 212 Pb is also continuously monitored by the CTBTO as routine parameter. We used a 12-month 212 Pb dataset recorded (simultaneously to the 7 Be dataset) at the CTBTO station RN33 in the Black Forest Mountains for our study. Long-term 7 Be datasets are generally available since the parameter is recorded as a routine parameter by the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO). A worldwide network of 80 CTBTO stations has been set up to monitor compliance with the Comprehensive Nuclear-Test-Ban Treaty. The resulting 7 Be datasets are freely available [10]. This makes 7 Be an easily attainable 35 S proxy.
Our study evaluated the recorded 35 S dataset versus a 7 Be time series recorded at the CTBTO station RN33, located (as sole CTBTO station in Germany) at Mount Schauinsland in the Black Forest Mountains. Due to the large spatial scale of 7 Be and 35 S production and migration processes in the atmosphere (as discussed in Section 1), the distance between the 7 Be and the 35 S recording stations (Black Forest Mountains and close to the City of Leipzig, respectively) is of only minor relevance.

Using the Aerosol Optical Depth (AOD) and 212 Pb in Air as Supporting Data
Besides the two large-scale atmospheric processes mentioned above in points (i) and (ii), a third reason can be given for the increase of the activity concentration of 35 S in the rain during the warm and dry season: (iii) a less intense 35 S washout from the atmosphere by rain. The less rain falls, the less intense is the washout, which results in an increase of the particle concentration in the rain when a rare precipitation event occurs. The triggering dry conditions are usually accompanied by an increased rate of resuspension of mineral dust from the earth's surface (windblown dust). The intensity of the latter effect can be evaluated based on two parameters: the aerosol optical depth (AOD) and the 212 Pb activity concentration in air.
The AOD indicates the attenuation of incoming solar radiation in the atmosphere by mineral dust, sea salts, volcanic ash, smoke from wildfires, pollution from factories and others. The potential sources of these attenuators indicate that the AOD is (regionally and temporarily) impacted by natural extreme events such as dust storms (with a global hotspot on the Arabian Peninsula) and/or anthropogenic activities such as land clearing fires (with a global hotspot in the Amazon Basin), or human-made air pollution (with global hotspots in northern India and eastern China). However, in Central Europe the impact of such effects on the AOD is less pronounced (although windblown atmospheric dust with a remote origin, such as northern Africa, Sahara, can be an issue). Besides, choosing the AOD wavelengths 1640 and 1020 nm for our investigation allowed focusing on the impact of mineral dust.
The AOD is continuously monitored by NASA at several AERONET sites worldwide. For our study, we used a dataset (covering the complete 12-month period of the 35 S sampling campaign) recorded as daily averages at an altitude of 115 masl at the Institute for Tropospheric Research, which is located about 30 km north-east of our rain sampling station. 212 Pb is a short-lived (half-life 12.64 h) progeny of the 232 Th decay chain and is easily detectable by gamma spectrometry due to its strong gamma line at 238.6 keV. Thorium and its progeny occur as a minor constituent of most soils on earth. Therefore, the 212 Pb activity concentration in the atmosphere can also be used as an indicator for the presence of terrestrial mineral dust (and hence for its washout by precipitation). Comparable to 7 Be, 212 Pb is also continuously monitored by the CTBTO as routine parameter. We used a 12-month 212 Pb dataset recorded (simultaneously to the 7 Be dataset) at the CTBTO station RN33 in the Black Forest Mountains for our study.

Anthropogenic 35 S Sources
Using 35 S as an environmental aqueous tracer with a distinct input function requires the consideration of anthropogenic 35 S sources and their potential impact. In nuclear facilities (i.e., power plants, reprocessing facilities, and research reactors) minor amounts of 35 S are continuously produced via the nuclear reactions 35 Cl(n,p) 35 S [33] and, to a lesser degree, 34 S(n,γ) 35 S [34]. The reactions are triggered by the activation of impurities in coolant liquids or solid reactor materials [35]. Furthermore, 35 S is in some nuclear facilities (in low activities) applied as process tracer. Both the unintentional production and the process tracer application of 35 S result in releases of the radionuclide from nuclear facilities, which occur with the emission of liquid effluents and/or coolant gas.
An evaluation of the impact of anthropogenic 35 S emissions on human health (carried out 43 years ago, [36]) concluded that not more than 4.8 MBq 35 S reach the overall group of worldwide adults by food ingestion per year. Due to this minor human exposure, 35 S does not belong to the group of radionuclides originating from anthropogenic activities systematically monitored and reported by the CTBTO or any national radiation protection organization. Hence, data on gaseous 35 S emissions from nuclear facilities are scarce.
A few historical data exist, though. A report by [35] summarized gaseous 35 S emissions from ten (west) German nuclear power plants for the 30-month period between January 1983 and June 1985. The facility-specific annual values ranged between 0.8 and 6.8 GBq/a, summing up to an annual gaseous 35 S emission averaged for (west) Germany of 24.7 ± 6.8 GBq/a.
We tried to compare this averaged anthropogenic release rate to the overall 35 S inventory of the lower atmosphere. Related atmospheric data is scarce, too. Average values of several-day periods published thirty years ago [25] revealed 35 S activities in atmospheric air (both attached to aerosols and as SO 2 gas) between about 0.107 and 0.225 MBq/km 3 . Another available data collection includes atmospheric 35 S activities recorded in Germany and Norway between 1970 and 1979 [37]. These datasets revealed activities ranging between about 0.01 MBq/km 3 and 0.09 MBq/km 3 with an overall mean of 0.034 MBq/km 3 . However, we consider this data less representative for the troposphere, i.e., our realm of interest, since all air samples were taken at ground level.
Based on the little available information given above, we assume for further estimations an average 35 S activity in the precipitation-influenced part of the troposphere over Central Europe of 0.1 MBq/km 3 . For approximately assessing the impact of anthropogenic gaseous radiosulfur release on the overall 35 S inventory of this part of the troposphere, we roughly estimated the atmospheric 35 S inventory over Germany for the lower 10 km of the atmosphere (i.e., the layer that is subject to both 35 S removals by meteoric water and 35 S potential input by anthropogenic activities). If an averaged atmospheric activity of 0.1 MBq/km 3 is assumed, the atmospheric 35 S inventory over West Germany (with an area of about 249 × 10 3 km 2 ) sums up to about 249 GBq. Balancing the decay of this activity requires an overall 35 S import of 720 GBq/a. This approximate 35 S import estimation reveals that the anthropogenic share of it (24.7 ± 6.8 GBq/a as shown above) equals only about 3.4% of the overall 35 S input. Even though this estimation is very approximate, we can conclude that anthropogenic 35 S release is generally insignificant for the application of 35 S as a naturally occurring aqueous tracer. However, as we demonstrate in our study and discuss in more detail below, the possibility of unusual anthropogenic peak releases has nevertheless to be considered.

Cross-Evaluation of the Recorded Time Series
The following five data time series (as introduced above) were recorded simultaneously over a twelve-month period and cross-evaluated: (i) precipitation intensity, (ii) 35 S in the rain, (iii) 7 Be in air, (iv) 212 Pb in air, and (v) AOD at wavelengths 1640 and 1020 nm. The time series (i) and (ii) were recorded and evaluated as discrete values reflecting individual rain events; the time series (iii), (iv) and (v) were obtained as daily values and evaluated as 15-day moving averages. All datasets were collected in Germany; time series (i), (ii) and (v) in the vicinity of the city of Leipzig; time series (iii) and (iv) at CTBTO station RN33, located in the Black Forest Mountains. Due to the immense spatial scale of the investigated atmospheric processes (cf. Figure 2), the distance between the two recording stations (ca. 500 km) is of only minor relevance. Figure 3 displays the recorded time series (i) rain intensity and (ii) 35 S activity concentration in the rain. The plot illustrates that samples for 35 S analyses were taken for all major rain events throughout the twelve-month sampling campaign. Furthermore, the figure shows that most rain events (88%) resulted in less than 6.5 mm of precipitation and could be collected quantitatively within the 200-litre HDPE container.
Water 2020, 12, x FOR PEER REVIEW 7 of 15 and (iv) at CTBTO station RN33, located in the Black Forest Mountains. Due to the immense spatial scale of the investigated atmospheric processes (cf. Figure 2), the distance between the two recording stations (ca. 500 km) is of only minor relevance. Figure 3 displays the recorded time series (i) rain intensity and (ii) 35 S activity concentration in the rain. The plot illustrates that samples for 35 S analyses were taken for all major rain events throughout the twelve-month sampling campaign. Furthermore, the figure shows that most rain events (88%) resulted in less than 6.5 mm of precipitation and could be collected quantitatively within the 200-litre HDPE container. The most intense rain event by far (37 mm) occurred on 9 September 2019 and lasted for the whole day. The related sample represents the rain that fell towards the end of the event. The detected unusually low 35 S activity (if compared to the two much less intense rain events that occurred before and after) illustrates the effect of dust and aerosol (and hence 35 SO4) washout by the rain during the initial hours of the event, i.e., previous to the event-specific sampling. A similar observation has been made during the major rain event on February 3rd (17.2 mm) that lasted for several hours, too. A sample taken at the beginning of the event showed a 35 S activity of 48 mBq/L; a sample taken towards the end of the event contained only 17 mBq/L (the mean value of 33 mBq/L was used related to this event for further evaluations).
Naturally, the recording of the 35 S data was bound to rain events. Due to rather dry conditions, particularly in January 2019 and between mid-March and the end of June 2020, not all periods of the twelve-month monitoring campaign could be documented satisfactorily with 35 S samples. Still, the 35 S dataset and the continuously recorded time series of 7 Be, 212 Pb and AOD (Figures 4 and 5; cf. Figure  9) reveal that all four parameters showed maxima in the second half of August 2019 and in April/May 2020. A third less distinct maximum was observed for 35 S and 7 Be in the first half of January 2020.
Excluded from this summarizing evaluation is a distinct peak in 35 S that was recorded in the second half of June 2020. While 35 S showed in the previous months an average value of about 25 mBq/L and reached maximum values around 140 mBq/L, it peaked on 19th June 2020 suddenly at 363 mBq/L declining subsequently to a still considerably high 148 mBq/L (24th June 2020). Figure 4 highlight this event in red. However, in order to illustrate the previous 35 S data more clearly, the 35 S axis of the three figures is scaled to a maximum value of 200 mBq/L, i.e., the 363 mBq/L peak is not fully shown (it is fully shown in Figure 5, though). Simultaneously to the sudden 35 S increase, the The most intense rain event by far (37 mm) occurred on 9 September 2019 and lasted for the whole day. The related sample represents the rain that fell towards the end of the event. The detected unusually low 35 S activity (if compared to the two much less intense rain events that occurred before and after) illustrates the effect of dust and aerosol (and hence 35 SO 4 ) washout by the rain during the initial hours of the event, i.e., previous to the event-specific sampling. A similar observation has been made during the major rain event on February 3rd (17.2 mm) that lasted for several hours, too. A sample taken at the beginning of the event showed a 35 S activity of 48 mBq/L; a sample taken towards the end of the event contained only 17 mBq/L (the mean value of 33 mBq/L was used related to this event for further evaluations).
Naturally, the recording of the 35 S data was bound to rain events. Due to rather dry conditions, particularly in January 2019 and between mid-March and the end of June 2020, not all periods of the twelve-month monitoring campaign could be documented satisfactorily with 35 S samples. Still, the 35 S dataset and the continuously recorded time series of 7 Be, 212 Pb and AOD (Figures 4 and 5; cf. Figure 9) reveal that all four parameters showed maxima in the second half of August 2019 and in April/May 2020. A third less distinct maximum was observed for 35 S and 7 Be in the first half of January 2020. be located in Leipzig, Germany (latitude: 51.200000; longitude: 12.200000), the trajectory paths fan out (back in time) in the main to the northwest but with their specific 60 h backward direction starting points scattering over an area ranging in between Brittany in the far southwest (northern France) and the Gdańsk Bay in the far northeast (northern Poland). Hence, their informative value is (at least in the given case) limited. This limitation is, e.g., due to potential vertical mixing of air masses along the way, an uncertainty that is substantially increasing with the backtracking time.   Figure 5). The unusually high AOD values recorded around this peak (max. 0.161) are not displayed either (cf. Figure 5).  Figure 5). The unusually high AOD values recorded around this peak (max. 0.161) are not displayed either (cf. Figure 5). As discussed in Section 2.2, 7 Be can be considered a direct proxy for 35 S. In general, 35 S in rain and 7 Be in air correlated reasonably well during the twelve-month period of the study (Figure 4). Still, the resulting correlation coefficient of the recorded twelve-month time series revealed the uncertainty of the information resulting from an evaluation of a 7 Be time series (even if supported by 212 Pb and AOD datasets) if used as proxy for 35 S (Figure 6). As it is shown in Figure 7, both parameters, measured and predicted, followed the same (seasonal) trend during the first 10 months of the campaign. However, from 13-16 June the comparative analysis of both parameters revealed an uncharacteristic behavior of 35 S.
While the calculated RMSE of the two time series between 15th August 2019 and 13th June 2020 was 0.4, the RMSE for the complete time series was 0.7. This significantly higher prediction error for the complete period of the measurement campaign was due to the anomalous peak that was observed between June and July 2020. Excluded from this summarizing evaluation is a distinct peak in 35 S that was recorded in the second half of June 2020. While 35 S showed in the previous months an average value of about 25 mBq/L and reached maximum values around 140 mBq/L, it peaked on 19th June 2020 suddenly at 363 mBq/L declining subsequently to a still considerably high 148 mBq/L (24th June 2020). Figure 4 highlight this event in red. However, in order to illustrate the previous 35 S data more clearly, the 35 S axis of the three figures is scaled to a maximum value of 200 mBq/L, i.e., the 363 mBq/L peak is not fully shown (it is fully shown in Figure 5, though). Simultaneously to the sudden 35 S increase, the AOD rose rather abruptly at the beginning of June. It reached a maximum value of 0.161 (8 June 2020), while the previous top values ranged around 0.125 (this AOD peak is also not shown in Figure 4c but fully displayed in Figure 5. A more detailed look at the AOD dataset reveals that the AOD was dominated by "fine mode" aerosol (90%), which is generally of anthropogenic origin (while "coarse mode" aerosol would rather indicate mineral dust). In contrast to 35 S and AOD, 7 Be and 212 Pb did not show significantly elevated values in the second half of June 2020. Figure 5 displays the recorded time series 35 S and AOD, with the 35 S axis scaled to a maximum value of 400 mBq/L; i.e., including the full 35 S peak recorded in late June (in red) and also fully displaying the simultaneously recorded ("fine mode" aerosol, i.e., anthropogenic) AOD peak. Back trajectory analysis that allows determining the origin of the related air masses (HYSPLIT Model provided by the Air Resources Laboratory (ARL) of the US National Oceanic and Atmospheric Administration (NOAA)) suggests for the time atmospheric particle transport to the sampling site from the northwest (UK, the Netherlands). However, this general assumption is based on a set of 27 specific back trajectory ensembles (calculated by means of NOAA Hysplit Web Interface), that differ quite substantially in their lateral tracks. In fact, while the end point of all ensembles was defined to be located in Leipzig, Germany (latitude: 51.200000; longitude: 12.200000), the trajectory paths fan out (back in time) in the main to the northwest but with their specific 60 h backward direction starting points scattering over an area ranging in between Brittany in the far southwest (northern France) and the Gdańsk Bay in the far northeast (northern Poland). Hence, their informative value is (at least in the given case) limited. This limitation is, e.g., due to potential vertical mixing of air masses along the way, an uncertainty that is substantially increasing with the backtracking time.
As discussed in Section 2.2, 7 Be can be considered a direct proxy for 35 S. In general, 35 S in rain and 7 Be in air correlated reasonably well during the twelve-month period of the study (Figure 4). Still, the resulting correlation coefficient of the recorded twelve-month time series revealed the uncertainty of the information resulting from an evaluation of a 7 Be time series (even if supported by 212 Pb and AOD datasets) if used as proxy for 35 S ( Figure 6).
As discussed in Section 2.2, 7 Be can be considered a direct proxy for 35 S. In general, 35 S in rain and 7 Be in air correlated reasonably well during the twelve-month period of the study (Figure 4). Still, the resulting correlation coefficient of the recorded twelve-month time series revealed the uncertainty of the information resulting from an evaluation of a 7 Be time series (even if supported by 212 Pb and AOD datasets) if used as proxy for 35 S (Figure 6). As it is shown in Figure 7, both parameters, measured and predicted, followed the same (seasonal) trend during the first 10 months of the campaign. However, from 13-16 June the comparative analysis of both parameters revealed an uncharacteristic behavior of 35 S.
While the calculated RMSE of the two time series between 15th August 2019 and 13th June 2020 was 0.4, the RMSE for the complete time series was 0.7. This significantly higher prediction error for the complete period of the measurement campaign was due to the anomalous peak that was observed between June and July 2020. As it is shown in Figure 7, both parameters, measured and predicted, followed the same (seasonal) trend during the first 10 months of the campaign. However, from 13-16 June the comparative analysis of both parameters revealed an uncharacteristic behavior of 35 S.

Natural Drivers of the Recorded Variations in 35 S
The simultaneous maxima of 35 S and 7 Be recorded in August 2019 and April/May 2020, respectively (Figure 4; cf. Figure 9), indicated a significantly intensified air mass exchange between higher and lower atmosphere during these times. This increased interaction is thought to be bound to two annually recurring processes that trigger an increased aerosol import from the lower stratosphere to the upper troposphere, namely (i) the cyclical altitudinal shift of the tropopause and (ii) the cyclical lateral or latitudinal shift of the HFDZ. In 2019, the HFDZ reached central Germany around late August/September. In 2020 the approach had already occurred in April [38]. This earlier arrival of the HFDZ in 2020 was due to an early northward extension of the Hadley Cell and a corresponding stalling of the jet streams, an effect that is thought to be due to decadal climate variability or/and global warming [13].
Besides these two large-scale atmospheric processes, the less intense washout of 35 S and 7 Be from the lower atmosphere during the dry season can be named the third reason for the increase of 35 S in the rain and 7 Be in the air in August 2019 and April/May 2020. The 35 S and 7 Be datasets alone cannot evaluate the relevance of this effect. However, its substantial impact was revealed by the simultaneous significant increase of the AOD and the even more obvious increase of the 212 Pb activity in the air (Figure 4b,c).
The minor increase in 212 Pb observed in January 2020 suggests reduced particle washout as the While the calculated RMSE of the two time series between 15th August 2019 and 13th June 2020 was 0.4, the RMSE for the complete time series was 0.7. This significantly higher prediction error for the complete period of the measurement campaign was due to the anomalous peak that was observed between June and July 2020.

Natural Drivers of the Recorded Variations in 35 S
The simultaneous maxima of 35 S and 7 Be recorded in August 2019 and April/May 2020, respectively (Figure 4; cf. Figure 9), indicated a significantly intensified air mass exchange between higher and lower atmosphere during these times. This increased interaction is thought to be bound to two annually recurring processes that trigger an increased aerosol import from the lower stratosphere to the upper troposphere, namely (i) the cyclical altitudinal shift of the tropopause and (ii) the cyclical lateral or latitudinal shift of the HFDZ. In 2019, the HFDZ reached central Germany around late August/September. In 2020 the approach had already occurred in April [38]. This earlier arrival of the HFDZ in 2020 was due to an early northward extension of the Hadley Cell and a corresponding stalling of the jet streams, an effect that is thought to be due to decadal climate variability or/and global warming [13].
Besides these two large-scale atmospheric processes, the less intense washout of 35 S and 7 Be from the lower atmosphere during the dry season can be named the third reason for the increase of 35 S in the rain and 7 Be in the air in August 2019 and April/May 2020. The 35 S and 7 Be datasets alone cannot evaluate the relevance of this effect. However, its substantial impact was revealed by the simultaneous significant increase of the AOD and the even more obvious increase of the 212 Pb activity in the air (Figure 4b,c).
The minor increase in 212 Pb observed in January 2020 suggests reduced particle washout as the primary reason for the less distinct maxima of 35 S and 7 Be that were observed simultaneously (AOD data is, unfortunately, missing for these days). During a five-week period starting on 26 December 2019, only minimum amounts of rainfall occurred in the wider area of the rain sampling station (overall 8 mm), thereby building up an increased aerosol concentration within the lower atmosphere.

Potential Anthropogenic Causes for Recorded Variations in 35 S
While the three maxima discussed in Section 3.2 can be associated to natural drivers, the significantly elevated 35 S values recorded in June 2020 (displayed in Figure 5) can hardly be explained with natural processes. The unusually high 35 S activities, which peaked on June 19th with a value of 363 mBq/L, did not reflect any increase in 7 Be, which excludes the stratosphere as the origin of the radiosulfur.
A potential explanation might be derived from related news that was issued in June 2020. Several public media reported at the time that elevated activities of anthropogenic radionuclides ( 134 Cs, 137 Cs and 103 Ru) had been observed in the lower atmosphere over Northern Europe in mid-June 2020, i.e., simultaneously with the 35 S peak in our study (e.g., reported by the CTBTO Executive Secretary Lassina Zerbo, [39]). The unusual activity concentrations had been detected at the CTBTO station RN63 (Stockholm, Sweden) ( Figure 8).  [39]. The black star shows the rain sampling location. Shaded in blue is the area that resulted from the modelled 60 h back trajectory ensembles discussed in Section 3.1.
Additionally to the three anthropogenic radionuclides reported by the CTBTO, our own inquiries revealed also an increase in 7 Be at the Swedish station RN63 (not at the German station RN33, though) simultaneously to the 35 S spike observed in our study (Figure 9). Hence it could be speculated that the CTBTO observation of a minor activity release in Northern Europe points towards the same anthropogenic source responsible for the significantly elevated 35 S activities detected in June close to Leipzig. However, evidence for this speculation can neither be provided based on the detected nuclide pattern nor on the modelled 60h back trajectory ensembles discussed in Section 3.1,  [39]. The black star shows the rain sampling location. Shaded in blue is the area that resulted from the modelled 60 h back trajectory ensembles discussed in Section 3.1.
Additionally to the three anthropogenic radionuclides reported by the CTBTO, our own inquiries revealed also an increase in 7 Be at the Swedish station RN63 (not at the German station RN33, though) simultaneously to the 35 S spike observed in our study (Figure 9). Hence it could be speculated that the CTBTO observation of a minor activity release in Northern Europe points towards the same anthropogenic source responsible for the significantly elevated 35 S activities detected in June close to Leipzig. However, evidence for this speculation can neither be provided based on the detected nuclide pattern nor on the modelled 60 h back trajectory ensembles discussed in Section 3.1, which do not overlap with the possible source region as detected by the CTBTO (even though the overall northern direction is the same). Furthermore, we cannot explain, why this observed 35 S spike correlates with distinctly elevated AOD values recorded at the Leipzig AERONET site (cf. Figure 5).  [39]. The black star shows the rain sampling location. Shaded in blue is the area that resulted from the modelled 60 h back trajectory ensembles discussed in Section 3.1.
Additionally to the three anthropogenic radionuclides reported by the CTBTO, our own inquiries revealed also an increase in 7 Be at the Swedish station RN63 (not at the German station RN33, though) simultaneously to the 35 S spike observed in our study (Figure 9). Hence it could be speculated that the CTBTO observation of a minor activity release in Northern Europe points towards the same anthropogenic source responsible for the significantly elevated 35 S activities detected in June close to Leipzig. However, evidence for this speculation can neither be provided based on the detected nuclide pattern nor on the modelled 60h back trajectory ensembles discussed in Section 3.1, which do not overlap with the possible source region as detected by the CTBTO (even though the overall northern direction is the same). Furthermore, we cannot explain, why this observed 35 S spike correlates with distinctly elevated AOD values recorded at the Leipzig AERONET site (cf. Figure 5).

Conclusions
The aims of our study were (i) to identify the drivers of the long-term variation of the 35 S activity concentration in rain and (ii) to evaluate the applicability of the routinely recorded 7 Be in air as proxy

Conclusions
The aims of our study were (i) to identify the drivers of the long-term variation of the 35 S activity concentration in rain and (ii) to evaluate the applicability of the routinely recorded 7 Be in air as proxy for setting up a 35 S input function that allows the use of 35 S as a groundwater residence time tracer. Our twelve-month sampling campaign results reveal that the time series of both 35 S and 7 Be show generally comparable yearly patterns with elevated values during the warm season. This pattern can be explained with two large-scale seasonal atmospheric processes: (i) the cyclical altitudinal shift of the tropopause and (ii) the cyclical latitudinal shift of the HFDZ. Additionally, the evaluated time series of 212 Pb in air and AOD revealed a noteworthy impact of (iii) reduced aerosol washout from the lower atmosphere by rain during the dry season.
Hence, if a groundwater study is to be executed to use 35 S as aqueous residence time tracer, we recommend the analysis of at least a few recent rain samples for their 35 S activity concentration and set up a 35 S input function based on these specific values, considering the general long-term pattern of higher 35 S activities in summer and lower 35 S activities in winter as indicated by regionally recorded datasets of 7 Be, 212 Pb and AOD. As a rough estimate we found the 35 S activity concentrations during the summer about three times as high as during the winter.
Furthermore, we recommend paying attention to the available time series of anthropogenic radionuclides recorded at CTBTO stations located in the wider area of the study site to become aware of any potential anthropogenic 35

S releases.
As the uncertainty in the input function will propagate into the calculation of the water age and age distribution patterns, it needs to be investigated further. Two key issues should be in focus in this regard: (i) dry deposition of 35 S and its impact on the input function and (ii) impact of rain duration and intensity on the 35 S activity (as shown for 22 Na by [40]).
Finally, we want to point out that each rain sample taken to set up a 35 S input function should represent the complete rain event. Long-lasting rains can result in a successive decrease in the 35 S activity concentration due to a washout of 35