1 Introduction

Precipitation is one of the best scavengers for removing particulate matter and gaseous pollutants from the atmosphere. The efficiency of this process reflects in their chemical compositions and pH values (Migliavacca et al. 2005; Prathibha et al. 2009). Chemical composition of precipitation is an important issue for many regions worldwide (e.g. Asia, eastern North America, Europe) as it can lead to eutrophication, affect ecosystems or contribute to global climate change (Calvo et al. 2010). Many studies focused on the chemical composition of rainwater from a spatio-temporal point of view (Avila 1996; Puxbaum et al. 1998; Galy-Lacaux and Modi 1998; Al-Momani et al. 2000; Topçu et al. 2002; Safai et al. 2004; Tu et al. 2005; Yoboue et al. 2005; Zhang et al. 2012) and identified several factors that can affect the chemical composition of rainwater: local pollutants, meteorological conditions, location, processes responsible for transport of pollutants (Rocha et al. 2003; Zhao et al. 2008). Chemical composition of precipitation can also provide information about both the main sources of pollutants and their migration processes (Ayres et al. 1995; Zunckel et al. 2003). A difference between ion constituents in rainwater on natural and anthropogenic sites exists. In coastal areas precipitation is mainly dominated by marine aerosols with high concentrations of Na+ and Cl− (Parungo et al. 1990; Ceron et al. 2002), whereas nitrogen (N) and sulphur (S) compounds dominate rural, urban and agricultural areas and reflect anthropogenic influence, such as fossil fuels combustion, fertilizers application on fields or specific sewage treatments (Das et al. 2005; Górka et al. 2014). The chemical composition of precipitation was also determined in Poland (Jóźwiak and Kozłowski 2005; Kacorzyk et al. 2012) including Wrocław and its vicinity (Kmieć et al. 1994; Czyżyk and Rajmund 2011). Chemical composition of precipitation and bulk precipitation in Wroclaw is monitored by the Voivodeship Inspectorate for Environmental Protection (VIfEP) that publishes results in yearly reports (e.g. VIfEP Report 2010).

Isotope approaches as complementary tools to chemical composition of precipitation have already proven their added value to identify the origin of pollutants (e.g. Caron and Tessier 1986; Lee et al. 2012; Xiao et al. 2012; Lim et al. 2014). Similar approaches have already been used for characterising precipitation in Wroclaw (SW Poland) (Górka et al. 2008; Szynkiewicz et al. 2008; Górka et al. 2011). Isotopes analysis of inorganic N, an important pollutant, have also been presented by other authors (e.g. Hoering 1957; Moore 1977; Freyer 1978a, b; Heaton 1987; Xiao et al. 2012; Felix et al. 2013) however not for Poland yet to our knowledge. Atmospheric NH3 is emitted from soils, oceans and is excreted by animals (Xiao and Liu 2002); anthropogenic sources include domestic animal excreta, fertilizers and biomass burning (Prospero et al. 1986; Hegg et al. 1988; Denterer and Crutzen 1994; Russell et al. 1998; Xiao and Liu 2002; Battye et al. 2003; Sapek 2013; Prenni et al. 2014). Nitrogen isotope composition of ammonium (δ15N(NH4 +)) shows wide variations from δ15N highly depleted in 15N (~ − 18.0 ‰) to highly enriched in 15N (~10.0 ‰), depending on the origin of NH4 + (Moore 1977; Freyer 1978a, b; Heaton 1987). We are studying here, for the first time in Poland, the δ15N(NH4 +) in order to trace the origin of nitrogen compounds in precipitation. Principal Component Analysis (PCA) can also provide additional information about processes and sources affecting the overall composition of precipitation (Plaisance et al. 1996; Cao et al. 2009; Calvo et al. 2010). The aim of this study is thus to characterise the processes involved in the genesis of major ions in precipitation in Wroclaw. We are also investigating seasonal variability based on relationships between chemical and isotope characteristics. This study uses a novel approach for characterising precipitation using an approach coupling chemical, isotope and statistical data from a 1-year (2010) monitoring in Wrocław.

2 Materials and methods

2.1 Study area

Wrocław is the largest city (630,000 inhabitants) in the Lower Silesia region in southwestern Poland, at the feet of the Sudety Mountains. Surface area of the city is 293 km2, and consists of: agricultural lands (132 km2), forests and woodlands (16.5 km2), water (9.6 km2), wasteland (3.9 km2), roads (28.48 km2), habitations (85.08 km2), the irrigation fields of Osobowice (11 km2) and others (17.44 km2). On its western part Wrocław is surrounded by agricultural lands. 5 large rivers flow through the city: the Odra River and its 4 tributaries (Oława, Widawa, Bystrzyca, Ślęza). 3 trans-national roads run through Wrocław. There are also two large railway stations, two inland navigation ports and one international airport. Wrocław is influenced by a temperate climate characterized by both continental and marine contexts. Wind directions are mainly western (27.6 % days) and southern (23.1 %) with an average speed around 2.5 m · s−1. The mean annual temperature is 9.6 °C and the yearly rainfall average is 589 mm (Dubicki et al. 2002).

2.2 Samples collection

Rain samples were collected using a rain collector located in downtown Wrocław (Poland) (Fig. 1).

Fig. 1
figure 1

Location of sampling area (Wroclaw, SW Poland)

Full size image

Forty-one precipitation samples (wet deposition) were collected at a 2 m height from January 2010 to December 2010 via a funnel (diameter of 0.8 m) into 5 L plastic containers covered by plastic a grid to prevent contamination. The funnel was placed in an open space to avoid any canopy effect.

2.3 Chemical analysis

Each sample was filtered through a 1.2 μm GF/C filter to remove organic pollution. Electrical conductivity (EC) was measured using a Multi 340i WTW multiparameter gauge with a TetraCon 325/S electrode (precision of ±1 %). pH of rainfall was calculated from the equation: pH = 6.37 + log Alk (meq.L−1) (Gomółka 1998). For alkalinity samples were titrated with 0.1 M HCl (methyl orange used as a pH indicator) using automatic pipette (OP200 HTL; precision of 20 μL) for estimating the presence of carbonic acid. Cations (Na+, NH4 +, K+, Mg2+, Ca2+) and anions (Cl−, NO2 −, NO3 −, SO4 2−) were measured using Alliance Waters High-Performance Liquid Chromatograph (HPLC) 2695 with a 432 Conductivity Detector and a 2998 Photodiode Array Detector, an IC-Pak Anion HR column (glauconate/borate eluent) and IC-Pak Cation M/D column (EDTA/HNO3 eluent). Before measurements each sample was filtered through a 0.2 μm syringe filter. The analytical precision of measurement for these ions was 0.01 mg · L−1. Using a 100-μL injection, the estimated method’s detection limits (MDLs) were, in ppb: Cl− (50), NO2 − (50), NO3 − (75), SO4 2− (75), K+ (15), Na+ (5). Mg2+ (10), NH4 + (5), Ca2+ (15).

Ion were preconcentrated using ion exchange resins. Before gravity dripping through resins, samples were filtered through the GF/C filter and then placed in a 1 L glass separatory funnel. Bio-Rad AG 50W-X8 (hydrogen form) resin was used for the analysis due to its relatively high affinity for NH4 + (Silva et al. 2000). Prior to dripping ions from the resin, the reagents as well as quartz wool and quartz tubes were prepared according to following steps:

  1. 1.

    CuO and Cu were calcinated at 400 °C for about 4 h in a muffle furnace to clean the reagents.

  2. 2.

    The CaO fraction of 1–2 mm was prepared following the procedures described in Boyd et al. (1994) and Kendall and Grim (1990).

  3. 3.

    Quartz tube (20 cm length and 9 mm diameter) and quartz wool were calcinated at 900 °C for about 1 h.

2.4 Procedure for eluting NH4 + ions from the cation exchange resin column and analysing its δ15N

Following the procedure described in Lehmann et al. (2001), the cation exchange resin loaded with NH4 + was dried in a muffle furnace at 50 °C and homogenized using an agate mortar. It was then loaded in a cleaned quartz tube with combustion reagents in the following order: (i) ground resin containing the NH4 + ions, (ii) 100 mg of copper oxide (CuO), (iii) 200 mg of copper (Cu), (iv) 500 mg of calcium oxide (CaO), 100 mg of CuO, (v) 200 mg of Cu and (vi) 500 mg of CaO. The tube was pumped down to 10−3 mbar and sealed hermetically and burnt for 10 h in a muffle furnace following the modified procedure of Kendall and Grim (1990): (i) raising the temperature up to 900 °C within 1 h (ii) maintaining the temperature at 900 °C for 4 h, (iii) cooling down to 600 °C for 3 h and (iv) maintaining for 2 h at 600 °C and finally cooling down to room temperature. At 900 °C, organic pollutants are burnt and AgNO3 reacts with oxygen liberated by the CuO. At 600 °C the carbon dioxide generated by the combustion of organic pollutants reacts with CaO to form calcium carbonate. At room temperature, created water is adsorbed onto the CaO particles to form Ca(OH)2. During the cooling steps nitrogen oxides are also reduced to N2 by Cu, which does not require additional cryogenic purification and is ultimately used to measure the isotope ratio 15N/14N (Kendall and Grim 1990).

Analysis of the N isotope composition was carried out on a dual inlet Delta V Advantage mass spectrometer in the Laboratory of Isotope Geology and Geoecology at the Institute of Geological Sciences, University of Wrocław. International standards IAEA N-1 and N-2 were used to calibrate the data. δ15N are reported relative to atmospheric nitrogen (N2-Air). The error of the δ15N determination is below 0.5 ‰.

2.5 Statistical analysis

Principal Component Analysis (PCA) is a very proven statistical method to characterise the spatio-temporal variability of large data population (e.g. Zhang et al. 1992; Manly 1998, Eder et al. 2014). PCA allows to abstract factors describing the variability of analysed population by using correlation matrixes and were used here with the intent to: (i) analyse the variability of several precipitation events simultaneously, (ii) discriminate sources of ions in precipitation, (iii) determine the respective contributions of those sources. PCA was carried out for 11 parameters: electrical conductivity (EC), HCO3 −, Cl−, NO2 −, NO3 −, SO4 2−, Na+, NH4 +, K+, Mg2+ and Ca2+. PCA followed several steps: (i) elimination, (ii) testing of the distribution normality, (iii) choosing parameters for analysis, (iv) standardisation, (v) parameters transformation using the PCA method, (vi) “varimax” rotation, (vii) factor’s loading and factor’s scores extraction and interpretation of its variability (Johnson 1978; Manly 1998).

3 Results

3.1 Weather conditions

Air temperature varied from −11 °C (December 4, 2010) to 26 °C (July 13, 2010) with an average value of 10.6 °C (median: 10 °C) (Table 1; Supplementary Materials). The average annual temperature (8.4 °C) for 2010 was about 1.2 °C lower than the 9-year average measured between 2000 and 2009 (Dubicki et al. 2002). We divided the calendar year into three distinct periods according to the meteorological data: vegetative season and 2 heating seasons. Vegetative season is defined as the period where the daily average temperature is higher than 5 °C. For 2010 we thus adopted the following periods: heating season I (January 14 to April 04), vegetative season (April 9 to October 16) and heating season II (October 13 to December 25).

The sum of yearly precipitation (751 mm) (Meteorological Bulletin 2010) was significantly higher than average value observed from 2000 to 2009 (589 mm) (Dubicki et al. 2002).

The calculated pH in precipitation ranged from 4.5 (December 25, 2010) to 5.8 (May 3, 2010), with a mean value of 5.2 (median: 5.2) (Table 1; Supplementary Materials) and correspond to the ranges determined by VIEP in Wroclaw (VIfEP Report 2010). EC in precipitation varied from 12 (December 25, 2010) to 162 μS · cm−1 (March 13, 2010), with a mean value of 71 μS.cm−1 (median: 61 μS.cm−1) (Table 1; Supplementary Materials).

3.2 Chemical composition of precipitation

NO2 − concentrations varied from <0.01 mg · L−1 (May 21, 31, November 14, and December 12, 25, 2010) to 2.9 mg · L−1 (April 26, 2010) with an average value 0.9 mg · L−1 (median: 0.7 mg · L−1), NO3 − ranged from <0.01 mg · L−1 (May 21 and November 14, 2010) to 16.5 mg · L−1 (April 26, 2010) with an average value of 4.6 mg · L−1 (median: 3.71 mg · L−1), whereas NH4 + concentrations ranged from <0.01 mg · L−1 (May 19, 21 and December 4, 2010) to 7.9 mg · L−1 (April 26, 2010) with an average value of 2.8 mg · L−1 (median: 2.1 mg · L−1) (Table 1).

Table 1 Descriptive statistics for the main parameters
Full size table

Ca2+ varied from 0.1 mg · L−1 (December 25, 2010) to 6.9 mg · L−1 (July 13, 2010) with a mean value of 2.2 mg · L−1 (median: 1.5 mg · L−1), Mg2+ from 0.05 mg · L−1 (December 25, 2010) to 1.0 mg · L−1 (December 5, 2010) with a mean value of 0.3 mg · L−1 (median: 0.2 mg · L−1), K+ from <0.01 mg · L−1 (December 5, 25, 2010) to 23.0 mg · L−1 (November 14, 2010) with a mean value 2.9 mg · L−1 (median: 1.6 mg · L−1), Na+ from <0.01 mg · L−1 (April 5, August 31, October 13, 2010) to 12.9 mg · L−1 (December 5, 2010) with a mean value of 2.4 mg · L−1 (median: 1.7 mg · L−1), SO4 2− from <0.01 mg · L−1 (April 11, November 22, December 25, 2010) to 24.5 mg · L−1 (November 10, 2010) with a mean value of 8.1 mg · L−1 (median: 5.6 mg · L−1), Cl− from 0.7 mg · L−1 (May 19, 2010) to 18.6 mg · L−1 (November 10, 2010) with a mean value of 4.1 mg · L−1 (median: 2.8 mg · L−1) and HCO3 − from 0.7 mg · L−1 (December 25, 2010) to 17.1 mg · L−1 (March 3, 2010) with a mean value of 5.7 mg · L−1 (median: 4.2 mg · L−1) (Table 1).

3.3 Nitrogen isotope composition (δ15N) of NH4 +

δ15N of NH4 + showed large fluctuations during the study period from −11.5 (January 28) to +18.5 ‰ (August 31) with an average value of 2.1 ‰ (median: 1.3 ‰) (Table 1).

4 Discussion

4.1 Sources and process controlling the chemical composition of rainwater

The chemical compositions of precipitation showed during 2010 a significant variability with time (Figs. 2 and 3) that may indicate that either one or different processes are involved or/and that this/these process(es) vary(ies) with time. PCA can help alleviate this difficulty by distinguishing processes that control the chemical composition of rainwater (Zhang et al. 1992) as well as help estimate the intensity of these processes by the study of factor scores. Extreme negative values (<−1) indicate that the specific processes that each factor represents do not affect samples. In contrast, positive values (> + 1) indicate they are strongly impacted by these processes. Values between −1 and +1 define the “mixing zone” of individual processes (Manly 1998; Jolliffe 2002).

Fig. 2
figure 2

Temporal variability of concentration of anions in precipitation water sampled in Wrocław in 2010

Full size image
Fig. 3
figure 3

Temporal variability of concentration of cations in precipitation water sampled in Wrocław in 2010

Full size image

PCA, in our case, explains 82 % of the observed chemical variations (Table 2). The remaining 18 % were considered as “random noise” that can not be interpreted using this technique (Drever 1997; Manly 1998). Results suggest that chemical and isotope compositions of our precipitation are formed/controlled by at least three distinct factors (Table 2) interpreted as natural and anthropogenic processes or a mixing of both.

Table 2 Results of the Principal Component Analysis (PCA) for precipitation water collected in Wrocław in 2010
Full size table

4.1.1 Factor 1

Factor 1 accounted for nearly 40 % of the total variance, had high positive factors for EC, HCO3 −, NO3 −, NO2 −, NH4 + and SO4 2− (Table 2) probably hinting that they may have a common origin. Strong correlation between mineral forms of nitrogen and sulphate ions is observed in many rainwater studies worldwide (e.g. Plaisance et al. 1997; Cao et al. 2009; Calvo et al. 2010) and usually is interpreted by the authors as a mixing of emissions from different anthropogenic sources.

NO2 − and NO3 − yielded maximum concentrations in spring at the beginning of vegetative season (Figs. 2 and 3) with values of 2.9 mg · L−1 and 16.5 mg · L−1, respectively (Table 1; Supplementary Materials; Fig. 2). These concentrations are unexpectedly higher than those reported for the Lower Silesia Region and rural areas in other countries (Cao et al. 2009; Wang and Han 2011). For 2010, the Voivodeship Inspectorate for Environmental Protection for whole voivodeship (VIfEP Report 2010) reports combined (NO2 − + NO3 −) concentrations from 0.02 mg · L−1 to 3.6 mg · L−1, still lower than our data obtained in Wrocław. Annual averages of NO3 − concentrations in the Sudety Mountains (SW Poland) ranged from 1.9 to 2.7 mg · L−1 between 1988 and 1992 (Kmieć et al. 1994) and from 0.6 to 1.5 mg · L−1 in the vicinity of Wroclaw between 2002 and 2010 (Czyżyk and Rajmund 2011). Our NH4 + concentrations in precipitation ranged from <0.01 mg · L−1 to 7.9 mg · L−1 in 2010. Kmieć et al. (1994) reported for 1994 NH4 + values from 2.3 to 3.9 mg · L−1 similar to the highest values we observed during the vegetative season (Fig. 4). Those spring maximum concentrations of N mineral forms may be related to the plants biological activity caused by the application of fertilizer onto agricultural fields during spring (Freyer 1978a, b). The concomitant NH4 + peak may be related to an increase in the rate of NH4 + volatilization from agricultural fields (fertilized with manure and/or urea). NH4 + volatilization results from the temperature-controlled bacterial enzymatic activity, where the urease enzyme catalyses the release of NH4 + from urea (Torello and Wehner 1983). Therefore an increase in the NH4 + volatilisation rate is expected during spring especially when considering other local conditions: predominance of sandy soils with expected low cations exchange capacities, high soil moisture, periodical flooding, crop residues (Ernst and Massey 1960; Torello and Wehner 1983; Tomassi-Morawiec et al. 1998).

Fig. 4
figure 4

Temporal variability of factor scores calculated via PCA for precipitation water sampled in Wrocław in 2010

Full size image

SO4 2− concentrations, from <0.01 mg · L−1 to 24.5 mg · L−1, were higher during heating periods (related to fossil fuels combustion) than during the vegetative season. An elevated spring concentration pick was also observed (Fig. 2). Similarly to NO2 − and NO3 2− SO4 2− concentrations were higher in Wrocław than in the whole voivodeship (0.3 to 15.1 mg · L−1; VIfEP Report 2010). In the near Sudety Mountains SO4 2− concentrations in precipitation range from 9.1 to 15.4 mg · L−1 (Kmieć et al. 1994) and from 3.1 to 9.4 mg · L−1 in the vicinity of Wroclaw (Czyżyk and Rajmund 2011). Additionally, during the heating season II two higher concentrations of NO3 − and NO2 − (coupled to high SO4 2− episodes) were observed (Fig. 1). Analysis of the variability of NO3 −, NO2 −, NH4 +, SO4 2− ions and PCA correlation shows that in Wroclaw they likely result from a mixing of two or three main anthropogenic sources: 1) application of fertilizers, 2) combustion of fossil fuels and 3) emissions from local Osobowice sewage irrigation fields. Though processes have not been fully characterised, SO4 2−, NO3 − and NH4 + can also be produced as secondary inorganic aerosols (SIA) trough photochemical reactions between gaseous pollutants and primary aerosols (Baker and Scheff 2007; Behera and Sharma 2011). Still, multiple atmospheric reactions may explain the strong correlation existing between these ions and ultimately play a role in the formation of the chemical composition of precipitation in Wrocław. Analysis of factor 1 scores shows that highest values (>1) are observed during the vegetative season (spring maximum; Fig. 4), indicating that these processes are dominant during this period. During the rest of the year they are contributing less (“mixing zone” in Fig. 4). Processes represented by factor 1 largely determine the rainwater mineralization as evidenced by its strong correlation with both EC and HCO3 − concentrations.

4.1.2 Factor 2

Factor 2 accounted for 26 % of the total variance and had high positive factor loadings for K+, Ca2+ and Mg2+. The fact that the PCA identifies these components as a factor suggests that their main sources in precipitation are probably distinct than those for NO3 −, NO2 −, NH4 + and SO4 2−. Ca2+ concentrations, obtained in this study, ranged from 0.1 mg · L−1 to 6.9 mg · L−1 consistent with concentrations reported by VIfEP (0.1-7.1 mg · L−1), Kmieć et al. (1994; 2.0 to 2.7 mg · L−1) and Czyżyk and Rajmund (2011; 0.5 to 2.5 mg · L−1. VIfEP reports slightly higher Mg2+ concentrations (0.1–1.6 mg · L−1) than those measured in this study (0.1–1.0 mg · L−1). Previous studies (1988–1992 and 2002–2010) yielded lower concentration ranges for Mg2+, from 0.3 to 0.5 mg · L−1 and from 0.1 to 0.4 mg · L−1, respectively (Kmieć et al. 1994 and Czyżyk and Rajmund 2011). K+ concentrations as high as 22.9 mg · L−1 significantly higher than those measured by VIfEP (0.1–2.9 mg · L−1) as well as in previous studies: 1.1 to 3.8 mg · L−1 (Kmieć et al. 1994) and 0.5 to 2.7 mg · L−1 (Czyżyk and Rajmund 2011).

For K+, Ca2+ and Mg2+ (Fig. 3) and factor 2 scores (Fig. 4) the highest values were observed 2–3 times during the vegetative season and once during the heating season II. The rest of the year, values plotted in the “mixing zone” (Fig. 4). These ions can have either a crustal (Ca2+, Mg2+, K+ are present in weathering products minerals) or anthropogenic origin (e.g. vehicle roadside dust, industrial aerosol) (Plaisance et al. 1997; Ali et al. 2004; Cao et al. 2009; Wang and Han 2011). Once emitted in the atmosphere they can be flushed following interaction with rainwater. But we observed no significant correlation between the volume of precipitation and these ions concentrations except once during the heating season II. Low precipitation can also lead to an increase in their corresponding concentrations. K+, Mg2+ and Ca2+ are ubiquitous in fertilizers, including ammonium nitrate (NH4NO3) and ammonium sulphate ((NH4)2SO 4 ), which probably represent one of these ions sources in factor 1. In particular the fact that potassium fertilizers, such as KCl and K2SO4, are often applied or re-applied in autumn and winter season in Europe (Crous et al. 2008; Hejcman et al. 2012; Shield et al. 2014) may explain the K+ peak observed in autumn (Fig. 2). Potassium ions also tend to accumulate in top layers in fertilized soils, which can facilitate their transfer to the atmosphere during soil dust suspension (Shield et al. 2014). This suggests that factor 2 may represent a mixing of two processes: 1) natural and anthropogenic dusts emissions and 2) application of fertilizers (containing K+, Ca2+ and Mg2+) to tillage soils. The absence of correlation between components from PCA factors 1 and 2 also favours volatilisation during application of fertilizers, irrigation fields activity as the main source of NH4 + as further field tilling is expected to generate dust, the main component of factor 2.

4.1.3 Factor 3

Factor 3 accounted for 16 % of the total variance and yielded high positive factor values for Na+ and Cl−, that probably represent marine aerosols and dry deposition (Plaisance et al. 1997; Błaś et al. 2010; Calvo et al. 2010). Na+ concentration ranged from <0.01 to 12.9 mg · L−1 and Cl− from 0.7 mg · L−1 to 18.6 mg · L−1, higher than the values measured by VIfEP in Wrocław and the whole voivodeship (Na+ from 0.01 to 2.6 mg · L−1 and Cl− from 0.05 to 6.5 mg · L−1) and from previous studies: Na+ from 1.5 to 1.9 mg · L−1, Cl− from 3.8 to 5.0 mg · L−1 (Kmieć et al. 1994) and Na+ from 0.9 to 3.2 mg · L−1, Cl− from 3.1 to 5.7 mg · L−1 (Czyżyk and Rajmund 2011). The median value for the Cl/Na ratio in precipitation was above 1.17, the typical for seawater, but along the year extremely high (367) and low (0.15) Cl/Na ratios were also observed. Analysis of the variability of Na+, Cl− and factor 3 values suggest that the processes represented by factor 3 affected the composition of precipitation throughout the year but dominated during the winter periods.

4.2 Origin of NH4 + concentrations and δ15N(NH4 +) isotope compositions

The NH4 + concentrations ranged from <0.01 to 7.9 mg · L−1 with an average value of 2.5 ± 1.8 mg · L−1 and δ15N(NH4 +) values ranged from −11.5 to 18.5 ‰ with an average value of 2.1 ± 6.8 ‰, which both represent wider ranges than those obtained by other authors (Fig. 6) (Hoering 1957; Moore 1974; Freyer 1978a, b; Heaton 1987; Paerl and Fogel 1994; Gao 2002). This large fluctuation of δ15N(NH4 +) may indicate that several sources of nitrogen compounds are involved or that secondary processes are significantly fractionating initial isotope compositions.

Temporal variations of NH4 + concentrations in precipitation clearly identify two distinct periods corresponding to two peaks in NH4 + concentrations (Fig. 5): (i) early spring, and (ii) early autumn. In both cases the maximum NH4 + concentration is probably due to (i) increased application of fertilizers in local agricultural fields 1) at the beginning of the vegetation period and then in autumn for autumn-sown cereal or 2) biological activity at the local Osobowice sewage irrigation fields. The variations in NH4 + concentrations do not seem to correlate with the corresponding δ15N(NH4 +) as we are observing a rough 15N enrichment along the whole season (Fig. 5).

Fig. 5
figure 5

δ15NN2-Air(NH4 +), NH4 + concentration in precipitation and air temperature during sampling period. Air temperature values from www.wunderground.com

Full size image

The lack of simple correlation of NH4 + concentrations (1/NH4 + ratio) with the corresponding δ15N(NH4 +) were confirmed by statistical tests (R = −0.07 and R = −0.08, respectively, p < 0.05). This is probably caused by a complex atmospheric open system controlled by numerous intricate nitrogen sources, conditions (e.g. temperature, pressure) and secondary processes that may fractionate nitrogen isotopes. However, our data show positive relations between the daily average temperature and δ15N(NH4 +), and daily average temperature and NH4 + concentrations (Fig. 5), confirmed by significant statistical correlations (R = 0.38 and R = 0.70 respectively; p < 0.05). Figure 6 reports variations of NH4 + concentrations and δ15N(NH4 +) for each identified 3 seasons (heating I, vegetation and heating II). Results indicate that all data can be explained by at least the three following end-members (two at low NH4 + concentrations and one dominant at NH4 + concentrations higher than 8 mg · L−1): 1) Low NH4 + and high δ15N (>20 ‰), 2) low NH4 + and low δ15N (<15 ‰) and 3) high NH4 + (>8 mg · L−1) and a δ15N around 0 ‰.

Fig. 6
figure 6

Relationship between ammonium concentration and δ15NN2-Air(NH4 +) in precipitation. Values of ammonium concentration and δ15NN2-Air(NH4 +) values are from: Hoering 1957; Moore 1974; Freyer 1978a, b; Heaton 1987; Paerl and Fogel 1994 and Gao 2002

Full size image

Three main processes may control the final δ15N(NH4 +) in this study (Fig. 7): 1) PROCESS 1: equilibrium fractionation between NH3 and NH4 +; 2) PROCESS 2: kinetic exchange reactions between NH3 and NH4 +; and 3) PROCESS 3: exchange reactions between atmospheric NH4 + in salt aerosols and NH4 + in precipitation. Variations of air temperature with δ15N(NH4 +) in precipitation (Fig. 7a) discriminate samples into two distinct groups depending on the season: (I) vegetative and (II) heating seasons.

Fig. 7
figure 7

Possible explanations of precipitation ammonium origin basing of δ15N(NH4 +). a Fractionation equation from Scalan 1958. b Vehicle exhaust, cow waste, volatilized fertilizer levels from Felix et al. 2013 and brown and hard coal level from Freyer 1978a, b. c Possible kinetic exchange reactions from Freyer 1978a, b and Heaton 1987; δ15N(NH4 +) levels for tropospheric aerosols from Moore 1977, fertilizer from Shearer et al. 1974 and for soil NH4 + from Freyer 1978a, b. Process 1 - Measured δ15N(NH4 +) values were converted to theoretical δ15N(NH3) assuming equilibrium exchange; Process 2 - Measured δ15N(NH4 +) values were converted to theoretical δ15N(NH3) assuming kinetic exchange; Process 3 - Measured δ15N(NH4 +) values were converted to theoretical δ15N(NH4 +) assuming atmospheric solid NH4 + exchange

Full size image

If we assume an equilibrium isotope fractionation between NH3 and NH4 + in precipitation we can try to estimate the δ15N of NH3 from which NH4 + was formed. Volatilization of urea (e.g. manure) in soils follows a Rayleigh distillation that is highly favouring 15N, resulting in Δ15N in volatilized NH3 around 35 ‰ (Frank et al. 2004) and even 30–60 ‰ (Högberg 1997; Robinson 2001). Similarly, the process of volatilizing fertilizers (with δ15N around 0.2 ‰; Bateman and Kelly 2007) produces NH3 with δ15N between −48.0 and −36.3 ‰, which represents a similar nitrogen isotope fractionation (Δ) as for urea. However, isotopically describing the opposite process of dissolving NH3 in atmospheric water into NH4 + is probably more complicated. Still, we can estimate the δ15N of potential sources atmospheric NH3 sources from the values of δ15N(NH4 +) values and daily air temperature using the NH4 +-NH3 equation described in Scalan et al. (1958) and in Busigny and Bebout (2013). Results give a range of δ15N(NH3) from −35.0 to −13.2 ‰ sample group I during vegetative season (Fig. 7b) that implies involving two main sources of NH3 : 1) animal waste with δ15N(NH3) values from −28.5 to −22.8 ‰, (Felix et al. 2013) and fertilizers with δ15N(NH3) values from −48.0 to −36.3 ‰, (Felix et. al. 2013). While our isotope data do not contradict fertilisers or sewage irrigation fields as the main sources of atmospheric NH3 in Wroclaw during the vegetative season, it does not unambiguously comfort this hypothesis as NH3 volatilisation from fertilizer, manure, urea produces overlapping δ15N(NH3) ranges. If PROCESS 1 is actually controlling the atmospheric budget of NH3 in the atmosphere of Wroclaw it would imply that during the vegetative season fertilisers and manure are the dominant sources of NH3, compared to vehicle and industrial emissions. For samples in group II that correspond to heating seasons I & II (Fig. 7a) we cannot use the Scalan equation as it is not suited for temperatures below 0 °C. At these temperatures a nitrogen exchange between NH3-NH4 + in a solid matrix (snow) is rather impossible. Kinetic exchange reactions between NH3 and NH4 + in rainwater (PROCESS 2) is another possibility for explaining δ15N(NH4 +) of samples from group II (Fig. 7a). The induced nitrogen isotope effect is still subject to debate as some studies report that NH4 + ions in rainwater are slightly depleted in 15N comparing to NH3 (Freyer 1978a, b) and others that they are slightly enriched in 15N comparing to NH3 (Heaton 1987). It follows that whatever hypothesis is considered the δ15N(NH4 +) we measured is only slightly different form the initial δ15N(NO3) during heating seasons (sample group II) in Wrocław (Fig. 7). If we then compare our data with the δ15N(NH3) reported by Freyer (1978a, b) and Heaton (1987) fossil fuels combustion appears as the main vector of NH4 + in rainwater, a conclusion that is supported by results from earlier studies (Górka et al. 2011 and 2012).

The last process (PROCESS 3) to be tested corresponds to exchange reactions between atmospheric NH4 + in salts aerosols and NH4 + in precipitation (Fig. 7c). Atmospheric NH4 + in salts aerosols is mainly related to particles generated from suspension of soils (or soils fertilised with NH4 +) or secondary inorganic aerosols (SIA) generated in the atmosphere. When under the form of ammonium nitrate (NH4NO3) or ammonium sulfate ((NH4)2SO4), NH3 is susceptible to be transported over long distances (Heeb et al. 2008). Górka et al. (2012) measured δ15N in PM10 (particulate matter with an aerodynamical diameter <10 μm) samples from the atmosphere of Wrocław from 5.0 to 13.7 ‰. It follows that dissolution of aerosols in precipitation may take place, accompanied with a nil to negligible nitrogen isotope fractionation (Δ15N; Fig. 7a and c).

5 Conclusions

Combined chemical, isotopic and statistical analyses provided critical information about the genesis of ions in precipitation in the atmosphere of Wroclaw. A PCA analysis discriminated them into three independent factors that lead to the following conclusions:

  • HCO3 −, NO3 −, NO2 −, NH4 + and SO4 2− result of emissions from direct fertilizers usage, sewage irrigation fields activity and fossil fuel combustion.

  • The study of K+, Ca2+ and Mg2+ concentrations and PCA analysis suggest that fertilised soil, fossil fuel combustion (road traffic and industrial) and natural dust are the main sources.

  • Marine aerosols, as expected, bring Na+ and Cl− ions to the city.

The implication and relative contributions of the processes controlling the overall budget of these ions vary following three distinct seasons: heating seasons I and II and vegetative season. The study of δ15N(NH4 +) shows that the dissolving of primary and secondary salts aerosols within the atmoshere of Wroclaw is prevailing in the NH4 + budget. During the heating seasons (I & II) our data suggest that the main source of NH3 in precipitation is fossil fuels combustion while during the vegetative season local fertilisers and the Osobowice sewage irrigation fields are dominating.