Relationship between precipitation chemistry and ... - Science Direct

30 downloads 0 Views 884KB Size Report
Bulk precipitation samples were collected at Montseny (Catalonia, NE Spain) from ... trajectory analysis to identify the source areas of pollutants in precipitation.
Atmospheric Environment 33 (1999) 1663—1677

Relationship between precipitation chemistry and meteorological situations at a rural site in NE Spain Anna Avila *, Marta Alarco´n Centre de Recerca Ecolo% gica i Aplicacions Forestals, Universitat Auto% noma de Barcelona, Barcelona, Spain Departament de Fı& sica i Enginyeria Nuclear, Universitat Polite% cnica de Catalunya, Vilanova i La Geltru& , Barcelona, Spain Received 24 February 1998; received in revised form 2 August 1998; accepted 4 August 1998

Abstract Bulk precipitation samples were collected at Montseny (Catalonia, NE Spain) from 1983 to 1994 and analysed for major cations and anions. The samples were classified for provenance based on meteorological synoptic maps and back trajectory analysis to identify the source areas of pollutants in precipitation. The meteorological classification was compared to an independent grouping based on multivariate data analysis (Clustering and Principal Component Analysis). Alkaline rain (mean pH"7.2) was associated to African trajectories. Local events produced neutral rains (mean pH"5.5). Acid rain was associated to rains of Atlantic origin (mean pH"4.8) and to European rains (mean pH"4.4), which also presented the highest mean concentrations of NH> (57 leq l\), NO\(49 leq l\) and   SO\(103 leq l\). However, European events were only a small fraction of the total precipitation (10% of the cases).  Marine rains accounted for 52% of the events, and African and Local for 20 and 18%, respectively. During the 11 year period there was a decreasing trend for the frequency of European events.  1999 Elsevier Science Ltd. All rights reserved. Keywords: Acid rain; Trajectory analysis; Cluster analysis; South-western Europe; Seasonality

1. Introduction The chemistry of precipitation is highly variable as a result of the different processes involved in element scavenging by cloud water and rain water. Particularly, the composition of precipitation is influenced by (1) the strength of emission sources, (2) the chemical reactions in the atmosphere, and (3) the mechanisms of scavenging of the moving air masses. In the last two decades, much attention has been devoted to acid rain, claimed to be responsible for soil and freshwater acidification and forest decline (Drablos and Tollan, 1980; Hutchinson and * Corresponding author. Fax: 0034 9358 11312.

Havas, 1980) and damage to historical buildings (Cheng et al., 1987). A particular subject of study has been the analysis of the relationship between the meteorological factors determining air mass trajectories and the enrichment of pollutants of these air masses (Colin et al., 1989; Pio et al., 1991; Sanusi et al., 1996). Some of these studies, notably those undertaken in the Scandinavian peninsula, central Europe and North America, have identified the source areas of acid rain and its pollutant load, therefore providing a base for claiming reductions of emissions (Munn and Rodhe, 1971; Miller et al., 1978; Singh et al., 1987). In Spain, the relationship between meteorological transport patterns and rain acidity was first documented

1352-2310/99/$ — see front matter  1999 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 2 - 2 3 1 0 ( 9 8 ) 0 0 3 4 1 - 0

1664

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

for northern Spain (Ezcurra et al., 1988; Diaz-Caneja et al., 1989). These studies concluded that acid rain was mostly associated to industrial sources in southern France and northern Spain. A study in the Catalan Pyrenees reported that neutral and acidic rains were originated in regions north of the Pyrennes while alkaline events predominated in rains from the south—southeast (Camarero and Catalan, 1993). In the Valencia Autonomous Community (eastern Spain), the analysis of backtrajectories and synoptical patterns for rain events from a network of 27 stations indicated three main routes of transport to the region: (1) the Atlantic ocean, (2) Europe, and (3) North Africa (Carratala`, 1993). The most acidic rains corresponded to the transport from Europe. These studies presented evidence for the transboundary character of the acid rain transport to northern and eastern Spain, although local contribution was also present. In the Montseny mountains (45 km to the northeast of Barcelona), previous studies have reported strong fluctuations of the rain acidity, with pH values ranging from 3.98 to 8.01 (Avila, 1996). Rain events in the alkaline pH range (pH"6—8) were related to the transport of Saharan air masses containing calcareous dust and clay minerals (Avila et al., 1997, 1998). However, little is known of the transport patterns responsible for the more acidic events. In this paper, we use a statistical approach to analyse the chemical features of the precipitation considering its meteorological origin. The aim of the paper is to identify the source areas of pollutants responsible for the rain acidity at a site in northern Spain. The study comprises a high number of cases therefore providing statistical reliability, and a 11-year of record therefore allowing for interannual comparisons. In particular, the objectives were: (1) to characterise the mean chemistry for the main transport routes reaching the site, (2) to determine the main synoptic patterns associated with acid rains, and (3) to analyse the seasonal trends of the meteorological situations responsible for the transport patterns.

2. Study site The sampling site was at 700 m a.s.l. in a clearing of the dense holm oak (Quercus ilex L.) forest at La Castanya Biological Station in the Montseny mountains (41°46N, 2°21E). The site is 45 km to the NE from Barcelona and 27 km to the W from the Mediterranean sea. The Montseny is a natural park: it is extensively forested and the main human activities are agricultural and recreational. Major vegetation types in the Montseny mountains are holm oak forests, and agricultural and pastoral land. The lithology of the site is formed by metamorphic schists, with chlorite, quartz, muscovite and albite as major mineral phases. The mean annual precipitation was 876 mm

Fig. 1. Map showing the areas for meteorological separation. E"European trajectories, M"Mediterranean trajectories, A"African trajectories and O"Atlantic Ocean trajectories.

(S.D."195 mm) for the period August 1983—July 1994. About 3% of this precipitation was snowfall.

3. Materials and methods 3.1. Precipitation sampling Precipitation for analysis was collected weekly, but during some wet periods in autumn and spring it was collected daily. Rain water was sampled by four continuously open collectors placed 1.5 m above the ground in a clearing of the forest. As the collectors were continuously open to the atmosphere, we sampled bulk precipitation (wet precipitation plus the fraction of dry deposition which sediments gravitationally; Eriksson, 1953). The inclusion of this fraction of dry deposition is known to alter the original wet deposition chemistry to some degree depending on the relative importance of the dry flux (Gascoyne and Patrick, 1981). The importance of particle sedimentation increases with dry climate and arid environments (Popp et al., 1984). On sites with a relatively dust-free atmosphere, as is usually the case of our sampling site at La Castanya, this dry deposition flux would be much less important. To assess the role of dry deposition in our samples we have used two approaches: (1) we have calculated the correlation between the number of dry days previous to the rain and the ion concentrations of bulk precipitation, and (2) we have analysed 20 blanks of distilled water used to wash the collector surface after a week of exposure without rain. The correlation between the element concentrations and the previous dry period was non-significant (p'0.5 for all elements). Concentrations in blanks

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

were very low (range 0.2—6 leq l\1) and represented a small fraction of the rain mean ion concentrations (0.2—12%). Therefore, we can conclude that dry deposition of sedimented particles did not modify to a great extent the chemical signal of the wet precipitation at our site. Consequently, we consider that the bulk precipitation data here presented retain the principal features of wet precipitation, although strictly speaking they do not correspond to wet precipitation. Furthermore, to minimise the role of dry deposition, we have eliminated from the data set all samples with precipitation lower than 2 mm. This included 36 events amounting to 52 mm (0.5% of total precipitation). Upon sampling, the rain water was taken to the laboratory where conductivity, pH and alkalinity were measured within 24 h in unfiltered samples. Samples were subsequently filtered through 0.45 lm pore size Millipore filters and the filtrate was deep-frozen for later analysis of the main anions (SO\, NO\ and Cl\) and cations   (Na>, K>, Ca> and Mg> and NH>). Alkalinity was  measured with a conductometric method (Golterman et al., 1978) until 1990, and by Gran titration thereafter with an Orion 960 Autochemistry system. Differences between both the methods for a common period of comparison (5 months) were lower than 15%. pH was measured with an Orion pH meter with an electrode for low ionic-strength solutions. Concentrations of Na>, K>, Ca> and Mg> in the filtered rain water were determined by atomic absorption spectrometry, SO\, NO\ and Cl\ were   determined by ion chromatography, and NH> by flow  injection analysis and gas diffusion (FIAstar Application Note 50-01/84, Tecator). All ion analyses included synthetic samples of known concentrations within the runs to check for precision and accuracy of the results. Data were screened for analytical quality by the quotients sum of cations/sum of anions (in terms of leq l\) and calculated conductivity/measured conductivity. The averages for these quotients were 1.09 and 1.02, respectively, for the 331 samples. Precipitation was measured by a tipping-bucket recording gauge and, each sampling day, in a standard non-recording gauge of 16 cm diameter.

3.2. Meteorological synoptic situations The synoptic meteorological situation associated with each precipitation sample was identified by means of the daily charts of the isobaric surfaces at 850 hPa (1200 UTC) of the Deutsche Wetterbericht and the Spanish Boletı´ n Nacional de Meteorologı´ a. Due to the extension of the data set, back trajectories could only be computed for 37 randomly selected cases. For the main body of observations, classification of air mass provenance was based on the synoptic meteorological situation. Such an approach, although lacking the accuracy of the back-trajectory calculation, has been frequently

1665

applied in similar studies (Ezcurra et al., 1988; Carratala`, 1993; Samara et al., 1992). Air mass movement from synoptic meteorological charts was compared to the 37 computed back-trajectories and a good correspondence of the classification of air mass origin with the two approaches was obtained. Because most of the precipitation events included weekly samples, we proceeded first to identify the rainy days for the weekly period. Next, we obtained the synoptic charts for each rain event and we compared the meteorological charts for the various rainy days in the period. If the weekly sample was obtained from many days with rain produced by a stable meteorological situation, we included the sample for statistical analysis; however, if the weekly sample was obtained from a set of different rainy days produced under different meteorological situations, then the sample was discarded. Fourteen observations were discarded due to mixing of synoptic situations within the sampling week. The retained data set consisted of 331 samples totalling a precipitation amount of 9402 mm, or 97% of the total precipitation above 2 mm in the 11-yr period. In 164 cases (50% of total), there was only one day with rain during the sampling period. It must be stressed that it is not infrequent for northeastern Spain to be affected by meteorological situations of the same characteristics for quite long periods, for example, by western air mass movement linked to Atlantic fronts crossing the Iberian peninsula (Clavero and Raso, 1979). Based on the meteorological synoptic situations and the back-trajectory analysis, we classified the precipitation events in four classes: Marine (Fig. 2), Local, European (Fig. 3) and African (Fig. 4). The Marine class included both Mediterranean and Atlantic air movements, Mediterranean situations, however, were less frequent (8% of total Marine). The Atlantic fluxes were produced by cold fronts crossing the Iberian peninsula from west to east (Fig. 2), a frequent situation during autumn and winter (Clavero and Raso, 1979). The Local class comprised events associated to meteorological situations with a very weak baric gradient. This situation is frequent in the Western Mediterranean, particularly in summer. At this time of the year, the Azores anticyclone, centred in front of Portugal, acts as a barrier against northern air fluxes. In this situation, local winds of thermal origin, such as breezes, are predominant and convective storms are often produced. Events without a well-defined air flux have also been assigned to the Local class. This occurred, for example, with the presence of a depression over Catalonia. This situation can be produced in winter and spring, and is often accompanied by heavy precipitation. The European class is associated with an advection from central and eastern Europe, produced by a depression crossing over Europe to the Italian peninsula, and an anticyclone over the northern Atlantic ocean (Fig. 3). This is frequent in winter and spring, and

1666

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

Fig. 2. Example of a Marine situation for the event on 7 November 1984. (A) Synoptic meteorological chart from the Deutsche Wetterbericht, (B) back-trajectory. 300 K indicates the altitude of the trajectory (1000—3000 m).

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

1667

Fig. 3. Example of an European situation for the event on 21 October 1991. (A) Synoptic meteorological chart from the Deutsche Wetterbericht, (B) back-trajectory. 300 K indicates the altitude of the trajectory (1000—3000 m).

1668

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

Fig. 4. Example of an African situation for the event on 11 October 1990. (A) Synoptic meteorological chart from the Spanish Boletin Nacional de Meteorologı´ a, (B) back-trajectory. 315 K indicates the altitude of the trajectory (4000—6000 m).

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

produces air fluxes over NE Spain which however give very little precipitation (Clavero and Raso, 1979). The African class corresponded to air fluxes from North Africa, usually associated to the presence of a depression over southern Portugal or north Africa at low levels, and an anticyclone over central or north Europe (Fig. 4). This type of situation has been reported to occur frequently over the Western Mediterranean during spring and summer (from April to August; Bergametti et al., 1989).

1669

Table 1 Squared Euclidean distances between clusters. They indicate the degree of differentiation between clusters in a four-dimensional space

Cluster Cluster Cluster Cluster

1 2 3 4

Cluster 1

Cluster 2

Cluster 3

Cluster 4

0.000

1.128 0.000

5.922 10.306 0.000

1.514 3.983 8.444 0.000

3.3. Back-trajectory analysis Air parcel back-trajectories were computed with an isentropic method extensively described in Avila et al. (1997). The parcel position was calculated in time steps of 900 s, integrating trajectory segments of 12 h after 48 time steps. Back-trajectories were calculated over a period of 3 days and were computed at four different levels, corresponding to four different potential temperatures (were K refers to Kelvin degrees): 305 K (2000—4000 m), 310 K (3000—5000 m), 315 K (4000—6000 m) and 320 K (5000—9000 m). 3.4. Statistical procedures To conform to the requirement of normality, all variables were log-transformed (except for pH) before statistical analysis. Then, the correlation matrix was calculated, and principal component analysis and cluster analysis were performed. The data set comprised 331 observations and 11 variables (Conductivity, pH, Na>, K>, Ca>, Mg>, NH>, SO\, NO\ and Cl\ and alkalinity).    The statistical procedures were carried out with the Statistica/Mac program (StatSoft, 1991). The applied K-mean clustering technique computed the square Euclidean distances between samples to allocate samples within the clusters. The K-means method produced four different clusters of greatest possible distinction, by minimising the variability within clusters and maximising the variability between clusters. This differentiation is defined by the squared Euclidean distances given in Table 1. Data relationships were also explored using a principal component analysis (PCA), a technique used to reduce a multidimensional system with many correlated variables into a simpler system of uncorrelated variables (the factors or principal components) which explain a high percentage of the variance of the original system. The extracted factors can be interpreted according to the correlation of the variables to the factor, thus enabling the identification of the potential source of variance associated with each factor (Drever, 1982). Here, PCA with Varimax rotation was applied to the entire transformed data set. Three factors were extracted, which explained 87% of the total variance.

4. Results and discussion 4.1. Meteorological approach The average bulk deposition chemistry at La Castanya (Montseny) is characterised by a moderate marine contribution, moderate N concentrations and a dominant role, on an equivalent basis, of Ca> and SO\ (Table 2). On  average, most of the potential acidity associated with strong acid anions is neutralised by base cations. In consequence, there is a net positive alkalinity and the mean pH is not acidic (pH"6.43, Table 2). However, acid rains (pH(4.5) were present at Montseny in 20% of the rain events that comprised 15% of the total precipitation amount. On the other hand, very alkaline events (pH'6.5) were also present in similar proportions (17% of the events, 20% of the precipitation). pH values ranged between 3.98 and 8.01. Ion concentrations also varied widely, with differences between maximum and minimum values ranging by 2—3 orders of magnitude (Table 2). Many studies have pointed out the determinant role of meteorological factors in determining the chemical features of precipitation (Asman et al., 1981; Raynor and Haynes, 1982; Singh et al., 1987). To evaluate this hypothesis, we classified all bulk precipitation observations into the four meteorological classes based on meteorological synoptic charts and back-trajectory analysis. 4.1.1. Mean composition per meteorological class For each class we calculated the geometric mean (as a better descriptor of the central tendency in log-normal distributions as is here the case), and the volumeweighted mean and standard deviation. These results are shown in Table 3, together with maximum and minimum concentrations for each class. Figs. 5 and 6 present the distributions of pH and the total ionic content for the four meteorological classes. Marine rains had lower concentrations for all ions (except for NH>) than rains from the other meteorologi cal classes, and they were moderately acidic, with the mode in the pH distribution at 4.5—5.0 (Fig. 5). In marine rains, the class of maximum ionic content

1670

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

Table 2 Volume-weighted mean concentrations, standard deviation and maximum and minimum values for the main dissolved ions in bulk precipitation at La Castanya (Montseny) collected in the period August 1983—July 1994 (n"333 samples). Conductivity in lS cm\, at 20°C; concentrations in leq l\ Variable

Weighted mean

Weighted S.D.

Max.

Min.

Cond. H# Na# K# Ca> Mg> NH>  NO  SO\  Cl\ Alkalinity pH

18.0 14.3 22.3 4.0 57.5 9.8 22.9 20.7 46.1 28.4 13.8 6.4

8.6 12.3 17.1 2.6 52.5 6.4 14.7 10.9 22.1 20.9 45.4

144 105 608 67 1164 152 170 241 425 318 608 8.01

3.1 0 1.1 0.1 2.1 0.4 0.8 1.3 4.9 2.5 !105 3.98

Volume-weighted hydrogen concentration calculated from the pH of each sample. This value does not take into account the neutralization due to alkaline samples, thus, it cannot be taken to represent the mean pH.  Mean pH calculated from volume-weighted mean alkalinity, assuming that bicarbonate accounts for most of the alkalinity.

(600—800 leq l\) was much lower than that of the other meteorological groups (Fig. 6). Because the Marine class comprised events of Atlantic and Mediterranean provenance, a Kolmogorov—Smirnov test was applied to test whether the ion chemistry for both origins belonged to the same distribution. This statistic indicated that significant differences existed for the distributions of Ca>, NH> and NO\ (p"0.03, 0.011 and 0.009, respectively).   The arithmetic means of Ca>, NH> and NO\ were 36,   25 and 22 leq l\ in events form an Atlantic provenance (n"156), whereas they were only of 14, 13 and 11 leq l\ for the Mediterranean events (n"14). European rains presented the highest mean conductivity, reflecting the very high concentrations of NH>, NO\   and SO\ and the acidic pH, as hydrogen ions have very  high specific conductivity. This class presented the lowest mean pH. All precipitation in the whole data set with pH(4.0 belonged to this class. The pH value distribution was clearly displaced to the acid side as compared to the other groups (Fig. 5). The high acidity was associated with very high NO\ and SO\ concentrations in this   group. Also, this group presented high concentrations of base cations and marine components. This resulted in European precipitation having a very high ionic content, only lower than some African events (Fig. 6). African rains presented high concentrations for base cations and the marine ions. Concentrations of NH> and 

NO\ were similar to those in the Marine class, but SO\   concentrations were higher. Precipitation was highly alkaline (Fig. 5). Local rains presented intermediate characteristics between Marine and the other two classes: they were more concentrated than Marine rains, but less than European or African rains. They presented a positive alkalinity, but lower than that of the rains of the African class. These chemical characteristics for each class can be interpreted in relation to the principal emission processes at each source area. Low concentrations in oceanic rains would reflect the background concentrations for the northern hemisphere (Galloway et al., 1983). For the Iberian peninsula, low concentrations of excess SO\  and NO\ have been found in air masses arriving from  the Atlantic Ocean to Portugal (Pio et al., 1991) and northern Spain (Ezcurra et al., 1988). The mean values for excess SO\ in these studies were 30—50 leq l\, and for  NO\, 8—10 leq l\, similar to our values for SO\ and   slightly lower than ours for NO\. However, in contrast  to these sites close to the Atlantic coast, Atlantic air masses must cross the Iberian peninsula before arriving at Montseny. Therefore, it appears that there is not an important increase of acid anions during the passage of air masses through Spain. The marine class does not present high concentrations for ions of marine origin (Na>, Cl\ and Mg>), a similar result to the oceanic rains sampled at Valencia (Carratala`, 1993). In fact, this class shows the lowest concentrations for these ions. Sea-salt aerosols might be deposited during transport across Spain. The Kolmogorov— Smirnov test comparing Atlantic and Mediterranean precipitation did not show significant differences for Na and Cl between both types of rain. European events presented an acid pH (pH"4.4) and a very high load of pollutant species (SO\, NO\ and   NH>), a result of the incorporation of anthropogenic  emissions from central or northern Europe. Base cations were also high, which could derive from the incorporation of terrigenous particles or fly ashes from industrial activities of European origin. The high base cation content and high alkalinity of African events have been attributed to dissolution of calcareous dust in rain originated over North Africa (LoyePilot et al., 1986). The high SO\ concentration in this class  is probably due to dissolution of gypsum (CaSO4), and that of sea-related ions to dissolution of halite (NaCl) in the African dusts. Gypsum and halite are very soluble minerals that might readily dissolve in cloud water so that no trace of these minerals has been found in X-ray mineralogical analysis of the dusts collected with the rains (Avila et al., 1997).

4.1.2. Extreme values The comparison of maximum and minimum values for each class indicates that the wide range of concentrations

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

1671

Table 3 Mean values for the chemical variables in the four meteorological classes, where xg indicates the geometric mean and xw and sdw indicate the volume-weighted mean and volume-weighted standard deviation, respectively. Maximum and minimum values for each meteorological class are shown. Mean pH calculated from volume-weighted mean alkalinity for meteorological classes with positive alkalinity (Local and African), and from volume-weighted mean H for classes with negative alkalinity (Marine and European). Ion concentrations in leq l\ and conductivity in lS cm\, at 20°C Cond Marine (n"170) xg 14.3 xw 13.7 sdw 5.4 max 39.8 min 3.1 Local (n"61) xg 19.5 xw 18.7 sdw 7.1 max 69.0 min 3.9 European (n"34) xg 40.7 xw 35.5 sdw 14.2 max 81.1 min 17.1 African (n"66) xg 28.8 xw 26.7 sdw 13.7 max 144 min 10.1

pH

4.8 6.9 4.2

5.5 7.1 4.1

4.4 5.8 4.0

7.2 8.1 4.2

H

10.0 16.3 11.1 70.8 0.12 4.4 12.7 11.4 87.0 0.1 34.5 41.3 23.4 105 1.5 0.3 3.5 5.5 67.6 0

Alk

!14.8 12.3

1.7 26.7

!41.0 23.6

82.7 94.2

Na

K

9.9 15.4 11.2 105 1.05

2.5 2.7 1.5 29.1 0.05

22.4 23.5 15.6 208 2.1

14.1 20.7 14.4 103 2.1

4.4 4.4 2.1 51.1 0.4

26.9 32.1 25.9 188 3.3 31.6 42.4 23.6 170 6.1

observed in the whole data set (Table 2) is maintained in these groupings, typically ranging 2 orders of magnitude in Table 3. Therefore, one could conclude that the meteorological classification used is not clearly discriminating the chemical signal of the different provenance. However, maximum and minimum values are extreme cases that are not greatly represented in each class. These extreme values could derive from (1) misinterpretation of the provenance due to the approximate identification through synoptic meteorological charts, (2) mixing of air masses during the rain event, or (3) chemical variability of the precipitation within each provenance. Although the first two points cannot be ruled out, meteorological charts were carefully screened and, as explained, mixing cases were discarded. In support to point (3), we have observed, for example, that some of the events of undoubtedly African provenance did not present the typical chemical signature described for African ‘‘red’’ rains of basic pH, high alkalinity and high base cation concentrations (Loye-Pilot et al., 1986, 1990; Roda` et al., 1993; Avila et al., 1997). However, rains from an African tra-

Ca

Mg

NH 

NO 

SO 

Cl

5.4 6.2 3.6 42.1 0.4

16.6 20.1 12.6 105 0.9

17.0 16.8 8.4 99.8 1.3

36.3 35.4 15.4 137 4.9

14.8 20.0 13.2 127 2.5

53.1 53.1 28.8 397 8.0

10.2 10.5 4.9 84.0 0.5

24.5 30.5 16.5 143 0.8

26.9 26.7 11.0 131 2.5

59.9 57.7 22.6 212 7.3

19.5 26.3 17.9 166 3.7

6.4 5.4 3.0 19.0 1.0

83.2 70.9 42.6 233 12.6

19.0 16.6 9.9 59.7 4.3

49.0 57.2 28.3 170 1.9

57.5 49.1 22.8 118 16.5

120 103 40.0 213 51.9

36.3 38.5 33.1 318 6.4

6.9 7.3 5.3 67.3 1.0

155 162 115 1164 7.4

20.9 18.7 9.1 152 3.7

18.2 17.7 10.9 108 1.0

29.5 21.4 10.4 241 5.6

70.8 56.8 25.0 425 15.6

39.8 52.8 51.0 280 7.0

jectory would only present this distinctive signature if they came across a dust cloud in their path over the African continent at the appropriate elevation of air circulation. The uplift of dust is more frequent under certain temperature and humidity conditions, notably produced in spring and autumn (Morales, 1979). At our site, six African trajectories (occurring in winter and spring) had a typical Marine fingerprint of low pH and very dilute concentrations. This would indicate that the atmosphere was dust-free at the time of the air mass transport. On the other hand, some meteorological situations classified as Marine (based in the criteria from Fig. 1) presented the African characteristics of basic pH and high base cation load. This was the case of seven events (occurring in April, May and June) with pH values higher than 6.0 and Ca> concentration higher than 100 leq l\. In that case, we interpreted that these air mass trajectories, although defined as marine by being traced from the Atlantic Ocean, might have impinged upon African dust clouds moving westwards over the

1672

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

Fig. 5. pH distribution for the meteorological classes.

Atlantic, so they have scavenged African materials. The transport of African dust over the North Atlantic has been extensively described (Prospero and Nees, 1986; Prospero, 1996; Duce et al., 1991). Although the above facts induce high variability within each class, we have maintained the meteorological classification because the high number of observations per class allows for a good definition of the central trend in each class. 4.2. Statistical approach To determine the sources for the variability of the data other approaches can be used, for example based on the statistical treatment of the data by means of multivariate techniques. This method has been particularly useful for the identification of pollution sources in studies of air quality (Sanchez Gomez and Ramos Martin, 1987; Crawley and Sievering, 1986; Thomas, 1986) and will be here applied for a comparison with the meteorological classification. The main interest of the statistical approach is in obtaining groups of events based uniquely on the numerical values of the rain composition, and try to determine if there is a correspondence between these and the meteorological classes.

To that purpose, a clustering analysis was applied as a classification tool and principal component analysis was performed to help interpret the groupings. The cluster analysis discriminated four groups separated by distances shown in Table 1. The cluster size and average composition is shown in Table 4. Principal component analysis extracted three factors that explained respectively 56.2, 20.5 and 10.5% of the total variance (Table 5). Component one was positively correlated to most ions (SO\, NO\ , NH>, K>, Ca> and Mg>), and there   fore can be interpreted as a size factor. The second component presented very high loadings for pH, alkalinity and Ca>, therefore representing an alkalinity—acidity axis. The third component represented clearly the marine contribution, with high loadings of Cl\, Na> and Mg> (Table 5). Very similar results in the interpretation of the principal components have been reported for the precipitation chemistry in the Iberian peninsula in the works of Ezcurra et al. (1988), Camarero and Catalan (1993), Codina and Lorente (1996) and Escarre´ et al. (1998). For our study, the scores of the individual samples in the space defined by the two main components are shown in Fig. 7. This space is defined by an axis representing the total chemical load (factor 1) and an axis representing the alkaline—acid character of the precipitation (factor 2).

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

1673

Fig. 6. Total ionic content distribution for the meteorological classes.

Table 4 Mean values for the chemical variables in the four clusters, where xg indicates the geometric mean and xw and sdw indicate the volume-wighted mean and volume-wighted standard deviation, respectively. The sum for the total ions (volume-weighted) is given in the last column. Mean pH calculated from volume-weighted mean alkalinity for clusters with positive alkalinity (3), and from volumeweighted mean H for classes with negative alkalinity (1,2 and 4). Concentrations in leq L\ and conductivity in lS cm\, at 20°C Cond Cluster xg xw sdw Cluster xg xw sdw Cluster xg xw sdw Cluster xg xw sdw

1 (n"86) 14.8 14.8 4.6 2 (n"96) 12.8 12.6 4.7 3 (n"96) 25.9 24.0 12.4 4 (n"53) 39.1 36.2 9.5

pH

H

5.0

7.6 11.3 7.7

4.7

Alk

Na

K

Ca

!9.1 7.0

10.8 18.1 12.3

3.5 3.6 1.5

35.6 35.2 14.8

18.1 20.5 10.9

!20.0 11.3

6.8 11.2 7.3

1.8 2.1 1.2

12.8 13.0 6.6

7.2

0.3 1.3 1.8

98.0 84.0

28.0 36.4 22.3

6.3 6.6 4.5

4.4

32.7 38.0 20.9

!37.8 21.0

32.3 45.6 28.4

6.2 6.3 2.7

134 142 98 79.1 72.4 33.2

Mg

NH 

NO 

7.2 8.3 3.5

22.1 26.4 14.2

21.4 21.9 8.0

3.4 4.1 1.8

11.5 15.2 8.2

18.5 16.9 8.6 20.5 20.3 9.0

SO 

Cl

Sum

43.5 43.7 14.5

15.7 23.1 14.5

191

13.2 13.0 6.1

30.5 30.8 12.2

10.6 14.8 8.4

125

20.6 22.3 15.7

28.2 22.4 11.0

67.0 55.6 26.0

35.4 45.6 28.8

446

48.5 53.3 21.5

54.1 50.8 16.9

110 102 29.3

41.3 58.9 36.9

448

1674

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

Cluster 2, on the lower left of the scatterplot, contains samples of low ionic content and an acidic character. Cluster 3, on the right side of the graph, includes samples of alkaline character and high ionic concentrations. Cluster 4, on the upper left, includes acid samples of high ionic content. Cluster 1 occupies a central position, indicating neutral pHs and medium to low ionic total concentrations. The mean composition for each cluster is given in Table 4, and, as above, can be summarised by considering the total ionic content and the pH. The correspondence between the cluster groups and the meteorological classification is given in Table 6. It can be seen that Local rains are well represented in clusters 1, 3 and 4, but a better correspondence occurs for Table 5 Correlation coefficients of the three principal components and the ion concentrations Variable

Factor 1

Factor 2

Factor 3

pH Alk. Na K Ca Mg NH  NO  SO  Cl

!0.040 0.018 0.192 0.566 0.568 0.530 0.835 0.911 0.852 0.218

0.958 0.940 0.172 0.423 0.646 0.365 !0.161 0.040 0.101 0.150

0.134 0.183 0.945 0.349 0.382 0.708 0.052 0.277 0.396 0.947

the other origins. For example, most of the Marine rains belong to cluster 2, most of African rains belong to cluster 3 and European rains to cluster 4. Clusters characterised by low pH contain the observations from Marine (cluster 2) and European origin (cluster 4). Marine samples were very diluted and acidic (Table 3), so it seems that neutralising base cations of terrigenous origin are scarcely incorporated into the air masses during transport over Spain. The most acidic samples, however, were from a European provenance, associated to a high load of SO\, NO\ and NH>.    4.3. Seasonal patterns Most precipitation events occurred in spring and autumn (Table 7). During winter, there was a maximum difference between the meteorological provenances, with Marine air masses accounting for 76% of the rains. African trajectories contributed to similar percentages (20—25%) during spring, summer and autumn, but decreased markedly during winter. Most of the events with a European trajectory occurred in spring and summer. Marine rains were the most frequent class, accounting for 52% of the total precipitation events, followed by the African and Local rains with 20 and 18% of total, respectively. The European class only contained 10% of total events. Fig. 8 shows the annual percent contribution of the different air mass origins for the 11-yr period of study. The most important feature is the decrease in the frequency of European events. For the first 5 years of study (1983—1988), the average annual frequency of European

Fig. 7. Scores of the samples represented in the space defined by the two first components from the Analysis of Principal Components. Samples identified by cluster.

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

1675

Table 6 Contingency table of the precipitation samples by meteorological classification and by cluster grouping Local

Marine

African

European

Total

Cluster 1 low conc, neutral pH Cluster 2 low conc, low pH Cluster 3 high conc, high pH Cluster 4 high conc, low pH

24

53

7

2

86

5

87

1

3

96

18

20

57

1

96

14

10

1

28

53

Total

61

170

66

34

331

Table 7 Seasonal percentages for meteorological provenance

Summer Autumn Winter Spring

Local

Marine

African

European No. of cases

29 21 9 13

32 48 76 57

24 24 8 20

15 7 7 10

79 96 59 97

events was 14%; for the last 5 yr (1989—1994), this frequency declined to 6%. In a previous work on the time trends of the precipitation chemistry, Avila (1996) found that at Montseny SO\ concentrations had decreased by  36% and the median pH had increased from 4.7 to 5.5 between 1983 and 1994. This was interpreted as a result of the reduction in 30% of the SO emissions in Spain  and Europe for the last decade (Agren, 1994). However, the meteorological evidence presented here of decreasing European air fluxes to Montseny can also help explain this declining SO\ trend and the increase of pH.  Recently, modifications in the wind circulation and precipitation patterns have been predicted for the Mediterranean region linked to climate change (Giorgi et al., 1992). The decreasing transport of European air masses over NE Spain reported here on 11 yr seems to point out to such a modification, but longer time sequences are necessary to confirm this finding. Also, some evidence exists for an increase of the influence of African air masses over South Europe, based on long term records in the chemistry of ice cores from the Alps (De Angelis and Gaudichet, 1991; Wagenbach and Weiss, 1989) and from long meteorological records in the Pyrenees (Dessens and Van Dinh, 1990). If these changes were confirmed, they would have important implications for the precipitation chemistry in southern Europe leading to a decrease of the rain acidity.

Fig. 8. Annual percentages of the types of rain classified by meteorological origin for the period 1983—1994.

5. Conclusions The meteorological provenance and the chemical composition were determined for 331 bulk precipitation samples collected at Montseny (NE, Spain). The main objective of the study was to identify the source regions of acid rain and pollutants. Multivariate techniques (cluster analysis and principal component analysis) supported the classification of rains into four principal groups: (1) Marine, (2) Local, (3) European and (4) African. Each meteorological class had a distinct composition with increasing concentrations of all ions following the above order, except for SO\, NO\ and NH> which had the    highest concentrations in the European class. Acid rain was associated to precipitation of Atlantic (mean pH"4.8) and European origin (mean pH"4.4). Alkaline rain (mean pH"7.2) was associated to African trajectories. Local events produced neutral rains (mean pH"5.5). Although Atlantic air masses need cross the Iberian peninsula before arriving at Montseny, precipitation of Atlantic origin had low ion contents, similar to reported values for oceanic rain sampled at the Atlantic

1676

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677

border in Portugal (Pio et al., 1991) and north Spain (Ezcurra et al., 1988). Marine rains accounted for 52% of the events, African and Local for 20 and 18%, respectively, while European events accounted for 10% of the cases. During the 11 year period there was a decreasing trend for the frequency of European events.

Acknowledgements This research has been funded by the following projects: CAYCIT 2129/83, CICYT AMB92-0349, ENCORE, and DM2E. The collaboration of the Departament d’Agricultura, Ramaderia i Pesca de la Generalitat de Catalunya and Departament de Fı´ sica de l’Aire de la Universitat de Barcelona is fully acknowledged. We thank F. Roda` for review of the manuscript, and technical personnel from CREAF for field and laboratory assistance.

References Agren, C., 1994. New figures presented. Acid News 5, 14—15. Asman, W.H.A., Slanina, J., Baard, J.H., 1981. Meteorological interpretation of the chemical composition of rainwater at one measuring site. Water, Air, and Soil Pollution 16, 159—175. Avila, A., 1996. Time trends in the precipitation chemistry at a montane site in northeastern Spain for the period 1983—1994. Atmospheric Environment 30, 1363—1373. Avila, A., Queralt-Mitjans, I., Alarco´n, M., 1997. Mineralogical composition of African dust delivered by red rains over northeastern Spain. Journal of Geophysical Research 102, (D18), 21,977—21,996. Avila, A., Alarco´n, M., Queralt-Mitjans, I., 1998. The chemical composition of dust transported in red rains — its contribution to the biogeochemical cycle of a holm oak forest in Catalonia (Spain). Atmospheric Environment 32, 179—191. Bergametti, G., Gomes, L., Remoudaki, E., Desbois, M., Martin, D., Buat-Me´nard, P., 1989. Present transport and deposition patterns of African dusts to the north-western Mediterranean. In: Leinen, M., Sarnthein, M. (Eds.), Paleoclimatology and Paleometeorology: Modern and Past Patterns of Global Atmospheric Transport. pp. 227—252. Camarero, L., Catalan, J., 1993. Chemistry of bulk precipitation in the central and eastern Pyrenees, northeast Spain. Atmospheric Environment 27, 83—94. Carratala`, A., 1993. Caracterizacio´n quı´ mica de la precipitacio´n en la Comunidad Valenciana. Ph.D. Thesis. Universidad de Alicante, 169pp. Cheng, R.J., Hwu, J.R., Kim, J.T., Leu, S.M., 1987. Deterioration of marble structures. The role of acid rain. Analytical Chemistry 59A, 104—106. Clavero, P.L., Raso, J.M., 1979. Cata´logo de tipos sino´pticos para un estudio clima´tico del Este de la penı´ nsula Ibe´rica y Islas Baleares. Aportaciones en homenaje al geo´grafo Salvador Llobet, Departamento de Geografı´ a, 63—86.

Codina, B., Lorente, J., 1996. Rainwater composition in the Barcelona area. Fresenius Environmental Bulletin 5, 412—417. Colin, J.L., Renard, D., Lescoat, V., Jaffrezo, J.L., 1989. Relationship between rain and snow acidity and air mass trajectory in eastern France. Atmospheric Environment 23, 1487—1498. Crawley, J., Sievering, H., 1986. Factor analysis of the MAP3S/RAINE precipitation chemistry network: 1976— 1980. Atmospheric Environment 20, 1001—1013. DeAngelis, M., Gaudichet, A., 1991. Saharan dust deposition over Mont Blanc (French Alps) during the last 30 years. Tellus 43B, 61—75. Dessens, J., van Dinh, P., 1990. Frequent Saharan dust outbreaks north of the Pyrenees: a sign of climatic change? Weather 9, 327—333. Diaz-Caneja, N., Bonet, A., Gutierrez, I., Martinez, A., Villar, E., 1989. The chemical composition of rainfall in a city of northern Spain. Water, Air, and Soil Pollution 43, 277—291. Drablos, D., Tollan, 1980. Ecological impact of acid precipitation. Proceedings of an International Conference, NIVA, Oslo. Drever, J.I., 1982. The Geochemisty of Natural Waters. Prentice-Hall, Englewood Cliffs, NJ. Duce, R.A., Liss, P.S., Merrill, J.T., Atlas, E.L., Buat-Menard, P., Hicks, B.B., Miller, J.M., Prospero, J.M., Arimoto, R., Chirch, T.M., Ellis, W., Galloway, J.N., Hansen, L., Jickells, T.D., Knap, A.h., Reinhardt, K.H., Schneider, B., Soudine, A., Tokos, J.J., Tsugonai, S., Wollast, R., Ahou, M., 1991. The atmospheric input of trace species to the world ocean. Global Biogeochemical Cycles 5, 193—259. Escarre´, A., Carratala´, A., Avila, A., Bellot, J., Pin ol, J., Milla´n, M., 1998. Precipitation chemistry and air pollution. In: Roda`, F., Retana, J., Gracia, C., Bellot, J. (Eds.), Ecology of Mediterranean Evergreen Oak Forests. Springer, Berlin (in press). Ezcurra, A., Casado, H., Lacaux, J.P., Garcia, C., 1988. Relationship between meteorological situations and acid rain in Spanish Basque country. Atmospheric Environment 22, 2779—2786. Eriksson, E., 1953. Composition of atmospheric precipitation. Tellus 4, 215—232. Galloway, J.N., Knap, A.H., Church, T.M., 1983. The composition of western Atlantic precipitation using shipboard collectors. Journal of Geophysical Research 88, 859—864. Gascoyne, M., Patrick, C.K., 1981. Variation in rainwater chemistry and its relation to synoptic conditions, at a site in north-west England. International Journal of Environmental Studies 17, 209—214. Giorgi, F., Marinucci, M.R., Visconti, G., 1992. A 2;CO climate  change over Europe generated using a Limited Area Model nested in a General Circulation Model. II: climate change scenario. Journal of Geophysical Research, 97, 10,011—10,028. Golterman, H.L., Clymo, R.S., Ohnstad, M.A.M., 1978. Methods for Physical and Chemical Analysis of Fresh Waters, IBP Handbook Vol. 8. Blackwell, Oxford. Hutchinson, T.C., Havas, M., 1980. Effects of Acid Precipitation on Terrestrial Ecosystems. Plenum Press, New York. Loye-Pilot, M.D., Martin, J.M., Morelli, J., 1986. Influence of Saharan dust on the rain acidity and atmospheric input to the Mediterranean. Nature 321, 427—428. Loye-Pilot, M.D., Martin, J.M., Morelli, J., 1990. Atmospheric input of inorganic nitrogen to the western Mediterranean. Biogeochemistry 9, 117—134.

A. Avila, M. Alarco´ n / Atmospheric Environment 33 (1999) 1663—1677 Miller, J.M., Galloway, J.N., Likens, G.E., 1978. Origin of air masses producing acid precipitation at Ithaca, New York. A preliminary report. Geophysical Research Letters 5, 757—760. Morales, C. (Ed.), 1979. Saharan Dust: Mobilization, Transport, Deposition. Wiley and Sons, Chichester, 297pp. Munn, E.R., Rodhe, H., 1971. On the meteorological interpretation of the chemical composition of monthly precipitation samples. Tellus 23(1), 1—13. Pio, C.A., Salgueiro, M.L., Nunes, T.V., 1991. Season and airmass trajectory effects on rainwater quality at the southwestern European border. Atmospheric Environment 25, 2259—2266. Popp, C.J., Ohline, R.W., Brandvold, D.K., Brandvold, L.A., 1984. Nature of precipitation and atmospheric particulates in central and northern New Mexico. In: Hicks, B.B. (Ed.), Deposition both Wet and Dry Acid Precipitation Series. Ann Arbor Science, Ann Arbor, pp. 79—85. Prospero, J.M., 1996. Atmospheric transport of particles to the Ocean. In: Ittekkott, V.S., Honjo, S., Depetris, P.J. (Eds.), Particle Flux in the Ocean, SCOPE Report 57, Wiley, New York, pp. 19—52. Prospero, J.M., Nees, R.T., 1986. Impact of the north African drought and El Nin o on mineral dust in the Barbados trade wind. Nature 320, 735—738. Raynor, G.S., Hayes, J.V., 1982. Concentrations of some ionic species in central Long Island, New York, precipitation in

1677

relation to meteorological variables. Water, Air, and Soil Pollution 17, 309—335. Roda`, F., Bellot, J., Avila, A., Escarre´, A., Pin ol, J., Terradas, J., 1993. Saharan dust and the atmospheric inputs of elements and alkalinity to Mediterranean ecosystems. Water, Air, and Soil Pollution 66, 277—288. Samara, C., Tsitouridou, R., Balafoutis, C., 1992. Chemical composition of rain in Thessaloniki, Greece, in relation to meteorological conditions. Atmospheric Environment 26, 359—367. Sanchez Gomez, M.L., Ramos Martin, M.C., 1987. Application of cluster analysis to identify sources of airborne particles. Atmospheric Environment 21, 1521—1527. Sanusi, A., Wortham, H., Millet, M., Mirabel, P., 1996. Chemical composition of rainwater in eastern France. Atmospheric Environment 30, 59—71. Singh, B., Nobert, M., Zwack, P., 1987. Rainfall acidity as related to meteorological parameters in northern Quebec. Atmospheric Environment 21, 825—842. StatSoft, 1991. Statistica/Mac. Thomas, W., 1986. Principal component analysis of trace substance concentrations in rainwater samples. Atmospheric Environment 20, 995—1000. Wagenbach, D., Geis, K., 1989. The mineral dust record in a high alpine glacier (Colle Gnifetti, Swiss Alps). In: Leinen, M., Sarnthein, M. (Eds.), Paleoclimatology and Paleometeorology: Modern and Past Patterns of Global Atmospheric Transport, pp. 543—564.

Suggest Documents