Water-soluble Ions of Aerosols in Taipei in Spring 2002

1 Research Center for Environmental Changes, Academia Sinica, Taipei, Taiwan, ROC * Corresponding author address: Dr. Shih-Chun Candice Lung, Research Center for Environmental Changes, Academia Sinica, Taiwan, ROC; E-mail: sclung@rcec.sinica.edu.tw Dust originating from arid and semiarid areas in Central Asia can reach Taiwan during springtime. However, aerosols in a metropolitan region, such as Taipei city, also have significant local sources. In this work, the influence of long-range transport and local pollution, on various water-soluble ions in fine and coarse modes, was investigated for several dust events. Concentration differences between dust-event and non-dust-event days were assessed. Field measurements were conducted in Taipei from February to May 2002. Twelve-hour samples of PM2.5 and PM10 were taken. Ion chromatography was used to analyze water-soluble ions (Na, NH4 +, K, Mg , Ca , Cl , NO3 − , and SO4 2− ). The mean concentrations of Na, NH4 , K, Mg 2+ , Ca , Cl, NO3 , SO4 2− , and PM in the fine mode during the entire sampling campaign were 0.29, 2.87, 0.29, 0.029, 0.13, 0.57, 0.92, 2.68, and 28.2 μg m 3 − , respectively. The corresponding mean concentrations in the coarse mode ( PM2.5-10) were 0.75, 0.30, 0.096, 0.14, 0.63, 0.84, 2.33, 1.60, and 34.0 μg m 3 − , respectively. Results of Principle Component Analysis showed that elevated concentrations of fine Na , K , Mg , Ca , and Cl and coarse Na , Mg , Ca , NO3 − , and SO4 2− on dust-event days were a result of long-range transport. A prolonged influence of long-range transport on concentrations of fine and coarse ions, after dust peaked, was also observed. K, Mg , and Ca increased by about 1.7 to 3.8 fold and Na and Cl increased by 45 210% on dust-event days. Fine modes of NH4 + , NO3 − , and SO4 2− were derived from local pollution, while most coarse NO3 − and some coarse SO4 2− were derived from long-range transport. On average, coarse NO3 − and SO4 2− increased by 67% and 97%, respectively, during dust events. <Back to Index>


INTRODUCTION
Asian dust storms (ADS) are common during late winter and spring in East Asia (e.g., Duce et al. 1980;Tsunogai and Kondo 1982;Prospero et al. 1989;Liu and Shiu 2001;Terada et al. 2002).They occur in arid and semi-arid regions during high-speed surface wind conditions (Duce et al. 1980).Under favorable synoptic conditions, significant amounts of Asian dust are transported to Taiwan (Chen & Chen 1987;Liu and Shiu 2001).Young et al. (1997) reported that during a dust event in spring 1995, PM 10 levels, at Taiwan Environmental Protection Administration (TEPA) monitoring stations, jumped dramatically within 3 -4 hours, along the western side of Taiwan.The magnitude and timing of maximum PM 10 levels tended to decrease and be delayed from north-to-south, coast-to-inland, and west-to-east.On the other hand, Young et al. (1997) also reported another high concentration case due to local pollution, only several days after the dust event.It was shown that the accumulation of PM 10 and CO were due to stagnant weather conditions, and a shallow surface inversion layer, which was associated with the returning flow of a high-pressure system.
In recent years, ADS over Taiwan have received considerable attention because of their impact on air quality.High concentrations of mineral elements and high levels of anthropogenic species, such as sulfate and nitrate, have been observed in the outflow of continental Asia (Talbot et al. 1997;Matsumoto et al. 1998;Clarke et al. 2001).Anthropogenic aerosols have reached Korea (Park and Cho 1998), Japan (Ishikawa et al. 1998), Hawaii (Perry et al. 1999) and even western sections of the US mainland (Jaffe et al. 1999(Jaffe et al. & 2003)).It was estimated that the total emissions of SO 2 and NO X from China in 1986 were 18.3 and 6.77 Tg yr 1 − , respectively (Kato & Akimoto 1992).This trend is expected to increase due to rapid increases in energy consumption and economic activity in China.Thus, assessing the influence of transported anthropogenic aerosols on Taiwan is very important.Nonetheless, aerosols in metropolitan regions, such as Taipei, also have significant local sources.Whether the observed pollutants of dust-event days are from local sources or have been picked up and carried to Taiwan as a result of long-range transport is an important scientific question.In this work, influences of long-range transport and local pollution on various watersoluble ions, in fine and coarse modes, on dust-event days in Taipei were investigated.Watersoluble Na + , NH 4 + , K + , Mg 2+ , Ca 2+ , Cl − , NO 3 − , and SO 4 2− were analyzed and Principle Component Analysis (PCA) was used to differentiate between the contributions of long-range transport and local pollution.Concentration differences of water-soluble ions on dust-event and non-dust-event days were also evaluated.

Sampling Strategy
Samplings of aerosols were conducted on the rooftop of the Central Weather Bureau located in central Taipei city.The sampling height was about 20 meters above ground.Twelvehour samples were taken from approximately 8 am to 8 pm (daytime sample, D) and from 8 pm to 8 am (night-time sample, N).From February to April 2002, sampling was conducted based on dust-event forecasts.Samples were collected one day before forecasted dust events and up until one day after the events had ceased.Since there were no dust storms in May 2002, samples were taken in the last week of the month.
One set of Partisol ® Model 2300 Speciation Sampler (RP2300, Rupprecht & Patashnick Co., Inc., Albany, NY, USA) was used to collect PM 2.5 and PM 10 samples.RP2300 consists of four channels, two of which were used to collect PM 2.5 and PM 10 samples, for elemental carbon/organic carbon (EC/OC) analysis.Two other channels were used to collect PM 2.5 and PM 10 samples, using 47 mm Teflon filters (P/N R2PL047, PALL Corporation, Ann Arbor, Michigan), followed by Nylon and citric-acid-coated quartz filters.After the samples were obtained, the Teflon filters were cut in half, for metal and ion analysis.Nylon and quartz filters were used to adsorb easily volatilized particulate nitrate and ammonium, respectively.In addition, RP2300 consists of honeycomb denuders that adsorb acidic gases to avoid artifacts in filter analysis.The results of EC/OC and metal analysis are presented in two accompanying papers (Chou et al. 2004;Hsu et al. 2004).A Gilian calibrator (Gilibrator 2, Sensioyne Gilian, Clearwater, FL) was used to calibrate flow rates of the four channels to approximately 10 l min 1 − .
These rates were then controlled by four independent mass flow controllers and a microprocessor ensuring constant volumetric flow rates.

Extraction and Analysis
Filters were equilibrated in a 20 ±1°C and 35 ± 5% RH environment for 48 hours, before a Mettler micro-balance (MX5, Mettler-Toledo AG, Grelfensee, Switzerland) was used to weigh them.A Static Eliminator (Allfield Enterprise Corp., Taipei, Taiwan) was used to neutralize electrostatic charges, during weighing.Each filter was weighed twice, and the acceptable variation was 3 µg.
After weighing, 30 min sonication, with 4 cc of de-ionized water, was used to extract the Teflon filters.AS12A and CS12A columns, with the conductivity detector of a Dionex DX120 Ion Chromatograph (Dionex Corporation, Sunnyvale, CA, USA), were used for anion and cation analysis, respectively.The eluent was 0.3 mM NaHCO 3 and 2.7 mM Na CO 2 3 for anion and 16 mM CH O S 4 3 for cation analysis, with 1 ml min 1 − .The same method was used to extract and analyse Nylon and quartz filters.
Standards were purchased from Merck (Merck KgaA, Darmstadt, Germany) and sevenpoint calibration curves were prepared for each batch of samples.The r-values of calibration curves were all greater than 0.995.Method detection limits (MDL) of ionic species, in µg per milliliter, were Laboratory blanks were used to assess contamination.The blank concentrations of ionic species were less than corresponding MDLs.Concentrations of non-detects were taken as zero.

Data Analysis
Concentrations of ionic species in the coarse mode ( PM 2.5-10 ) were obtained by subtracting those of PM 2.5 (the fine mode) from the corresponding concentrations of PM 10 .The Statistical Analysis System (SAS version 8.2, SAS Institute Inc., Cary, NC) was used for statistical analysis.The Spearman correlation coefficients were used to assess associations between the two variables.The Wilcoxon rank-sum test was used to compare ionic concentrations of dust event days with those of non-dust event days.Principal Component Analysis (PCA) was used to determine if patterns of water-soluble ionic profiles, in fine and coarse modes, on days with dust events were different from days with no dust events.PCA has been used in environmental studies to extract information from multivariate datasets (Ozeki et al. 1995;Shine et al. 1995;Liu et al. 1996).The data was standardized beforehand to have zero mean and unit variance, so that certain ionic species, with high concentrations, would not affect PCA results.Only a component with eigen value greater than 1 accounts for a meaningful amount of variance, and is worthy of being retained.The retained components, with oblique rotation, were used in this study.A detailed description of PCA can be found in Kleinbaum et al. (1987).

Concentration Distribution
Sixty-three pairs of PM 2.5 / PM 10 samples were collected: four in February, thirty-two in March, eighteen in April, and nine in May.Since the filters of one pair of samples, from February, were misplaced, during sampling, sixty-two pairs of valid samples were considered.The mean concentrations of Na + , NH 4 , and PM in the fine mode, during the entire sampling campaign, were 0.29, 2.87, 0.29, 0.029, 0.13, 0.57, 0.92, 2.68, and 28.2 µg m 3 − , respectively.The corresponding mean concentrations in the coarse mode ( PM 2.5-10 ) were 0.75, 0.30, 0.096, 0.14, 0.63, 0.84, 2.33, 1.60, and 34.0 µg m 3 − , respectively.On average, ionic species measured (summed as total ions) accounted for 45% of the fine particle mass and 25% of the coarse mass.Sulfate was the predominant ion (28%) in the fine mode, while nitrate was the most significant ion (9%) in the coarse mode.Detailed concentrations and percentages of each ionic species are listed in Table 1.
The time series of ionic concentrations in fine and coarse modes are shown in Figs.1a -h.The coarse mode to fine mode ratios (C/F) of PM and various ionic species are also presented in Figs.1a -h.The results indicate that Na + , Mg 2+ , Ca 2+ , Cl − , and NO 3 − were mostly in coarse PM, while NH 4 + , K + , and SO 4 2− were predominantly distributed in fine PM.In an accompanying paper, dust events were identified based on mass concentrations and C/F ratios of PM and Al (Chou et al. 2004).Spearman correlation analysis was used to determine if C/F ratios of various ions changed along with C/F ratios of PM.The results show that C/F ratios of chloride and sulfate ions correlated with C/F ratios of PM (Table 2).C/F ratios of Cl − also   correlated with Na + , K + , and Ca 2+ ; and those of SO 4 2− correlated with Na + , K + , and Cl − .In addition, Na + , K + , Mg 2+ , and Ca 2+ inter-correlated.Since certain ions were predomi- nantly distributed in the fine or coarse mode, C/F ratios showed quite a few missing values (Figs.1a -h).Hence, correlation coefficients obtained from C/F ratios might lead to an underestimate of the true association amongst ions and PM.

Origins of Water-Soluble Ions
In order to further assess the relationship of the ionic species, PCA was used to extract patterns of ionic profiles in fine and coarse modes, during the entire sampling period.The results show that only the first two and three components displayed eigen values greater than 1 for fine and coarse modes, respectively.Thus, these components were retained for rotation.Results of rotated component patterns are listed in Table 3.An ion was said to load on a given component if the factor loading for that component was 0.4 or greater.It should be noted that scattered precipitation was observed on several sampling days.PCA was conducted by excluding these days and the results were very similar to those in Table 3.Therefore, PCA results, based on all sampling days, were retained, to retain more information.Similarities between these two sets of PCA results also indicated the robustness of the data analysis method.
The first component of the fine mode was comprised of Na + , K + , Mg 2+ , Ca 2+ , and Cl − .These ions could have come from either seawater or soil.Na + was conventionally assumed to be entirely from seawater.Ratios of K + , Mg 2+ , Ca 2+ , and Cl − to Na + were compared to seawater composition (Table 4).Since, Table 4 shows that the ratios of K + and Ca 2+ to Na + were much higher than those in seawater, these ions must have had an additional source(s), such as soil.Ions from two distinct sources (seawater and soil) were both included in the first component, indicating that these ions were probably transported to Taipei, under the same synoptic meteorological conditions.Therefore, this component could be taken as an indicator of long-range transport.Since, the second component consisted of pollutants (NH 4 + , NO 3 − , and SO 4 2− ), it could be used to represent local pollution.The two components accounted for 59% and 18%, of variance in the fine mode, respectively.
With respect to the coarse mode, the first component included ions mainly from sea-salt ( Na + ), soil ( Mg 2+ , Ca 2+ ) and anthropogenic ( NO 3 − , SO 4 2− ) sources.For the same reason as the fine mode, this component could be used as an indicator of long-range transport.This component showed that long-range transport not only brought mineral particles, but also carried anthropogenic aerosols to Taipei, at least in the coarse mode.Since, the second component mainly consisted of ions from sea-salt, it was used to represent the influence of sea-spray.The third component, similar to the second component for the fine mode, was viewed as an indication of local pollution.These three components could explain around 57%, 16%, and 13%, of variance in the coarse mode, respectively.
According to the factor loading (Table 3), long-range transport would bring fine Na + , K + , Mg 2+ , Ca 2+ , and Cl − , and coarse Na + , Mg 2+ , Ca 2+ , NO 3 − , and Five and eight measurements of fine and coarse modes, respectively, were found to have positive PC1 values, but were not from days identified as dust events (marked in the figures).In the fine mode, three out of the five "exceptions" occurred in samples taken directly after the dust events (case 2, 3, and 4).The spikes in the graph suggest that peaks of fine water-soluble ions, carried by long-range transport, arrived later than the mineral dusts.In fact, one spike was also observed at the end of Case 1 and just happened to be included in the dust event as the last sample.This spike again came later than the PM peak.The other two exceptions occurred on 4/01D and 5/27D and warranted further investigation.It should be noted that 3/31D-3/31N was a suspected dust case, and while it was discussed and excluded from dust-event cases by Chou et al. (2004), it was included as a dust case by Hsu et al. (2004).It was found here that the sample right after this suspected case (4/01D) showed a positive influence from long-range transport, as in the other strong dust-event cases, identified.
With regard to the coarse mode, one exception occurred on 3/31D, i.e., the suspected dust-event case, as stated above.One and five exceptions were found on the following days of Cases 2 and 3, respectively; thus indicating the prolonged influence of long-range transport.Moreover, the peak of PC1 in Cases 1, 3, and 4 occurred 12 hrs after the peak of PM.This is consistent with the findings of Hsu et al. (2004), where peak values of Pb occurred 12 to 48 hrs, after the dust peaks.Besides the prolonged effect, the other two exceptions in the coarse mode occurred 12 and 24 hrs, before dust-event Cases 3 and 2, respectively.In addition, the graph indicates that PC1, in both fine and coarse modes, varied during the dust events.These results were consistent with the findings of Talbot et al. (1997) and provided evidence that individual dust-laden air masses were not well mixed.Consequently, the arriving time of various particulate compounds varied.Lin et al. (2004) analyzed long-range transport cases from November 1999 to May 2000 and from November 2000 to May 2001, with hourly concentrations of air pollutants such as PM 10 , CO, and SO 2 , from four EPA monitoring stations in Taiwan.They also pointed out that hourly correlations of mineral dust and air pollutants were very poor, as opposed to good correlations, based on seasonal variation, as observed by Prospero et al. (2003).
In general, identification of strong dust-event cases presented minor discrepancies between analysis, based on PM/Al, and those, based on long-range transport components, of water-soluble ions.Since, ADS carried mostly coarse PM and crustal elements such as Al, classification of dust-event days in this work complied with that of the four strong cases identified by Chou et al. (2004).Since, none of the weak dust-event days, identified by Chou et al. (2004), had positive PC1, only samples, from the four strong dust-event cases, were considered in the following data analysis, while other samples were viewed as non-dust-event samples.

Comparison Between Dust-event and Non-dust-event Days
Concentrations of water-soluble ions, in fine and coarse modes on dust-event and nondust-event days, are presented in Tables 5a, b.Non-sea-salt (nss) SO 4 2− was calculated by excluding contributions of seawater.Ratios of dust-event to non-dust-event samples are also shown in Tables 5a, b.The Wilcoxon rank sum test was used to compare concentrations of the two periods.The significant differences, at p = 0.05 and p = 0.1, [the chances (p) that these two groups came from the same population were less than 0.05 and 0.1, respectively] are indicated in Table 5.Of all the ionic species in the fine and coarse modes, only concentrations of fine NO 3 − and coarse NH 4 + decreased during dust periods.They were both from local sources.
Mg 2+ and Ca 2+ both increased by up to threefold, with most in both fine and coarse modes.SO 4 2− and nss-SO 4 2− in the fine mode (from local sources) were roughly the same in both periods; however, their concentrations in the coarse mode increased by about twofold during dust-event periods, due to long-range transport.Na + and Cl − in the coarse mode, presumably from seawater, increased by about twofold.
The percentages of water-soluble ions in fine and coarse PM, on dust-event and non-dustevent days, are tabulated in Tables 6a, b.Total ions accounted for about 45% of mass in PM 2.5 and 27% in PM 2.5-10 during non-dust-event days.During dust events, total ions had roughly the same percentage in the fine mode, although fine K + , Mg 2+ , and Ca 2+ showed an obvious increase (p < 0.05) due to long-range transport.However, the percentage of total ions was reduced by about 25% in the coarse mode.This result was to be expected, since dust-laden air masses carry mostly mineral elements (Hsu et al. 2004).Ca 2+ was the only ion in the coarse mode, which had a higher percentage during dust events, but the difference was not significant (p > 0.1).
Equivalent ratios of various cations and anions are shown in Table 7.It should be kept in mind that, while hydrogen ions are an important component in the aerosol ion balance (Tanner et al. 1981), they were not measured.In both fine and coarse modes, equivalent ratios of total cations and total anions increased, during dust events (p < 0.1).This finding suggests that particulate matters in both fine and coarse modes were more alkaline on dust-event days, as expected.The ratios of NH 4 + to nss-SO 4 2− , NO 3 − , and the sum of NO 3 − and SO 4 2− , in the fine mode, were almost the same in both periods, but they were much lower in the coarse mode, during dust events.On the other hand, equivalent ratios of the sum of Ca 2+ and NH 4 + to the sum of nss-SO 4 2− and NO 3 − increased in both fine and coarse modes, indicating that Ca 2+ was a significant contributor to the ion balance, during dust events.This phenomenon was also observed on dust-event days in Hsinzhuang and Taichung in 2002 (Lin 2002;Ye 2002).The important role of Asian dust in the neutralization of acid rain was also observed in Japan (Ishikawa et al. 1998) and demonstrated with a numerical model (Terada et al. 2002).
Ratios of Cl − to sea-salt Cl − (ss Cl − ) are shown in Table 7. Cl − depletion (%) was defined as 100% minus Cl − /ss Cl − (%).On average, Cl − depletion decreased during dust events; however, the change was not statistically significant.In the fine mode, almost all Cl − was from sea-salt on dust-event days, as the ratio was close to 1. On non-dust-event days, higher percentages of Cl − over sea-salt Cl − indicated contributions from other sources.In the coarse mode, 41 -47% depletion was observed on dust-event and non-dust-event days.

DISCUSSIONS
Nss SO 4 2− is an important pollutant.Its concentrations in the fine mode were not much different between dust and non-dust periods, while those in the coarse mode showed a significant increase during dust periods.The same phenomenon was observed on dust-event days in Taichung in Spring 2002 (Ye 2002).The results of PCA show that fine nss-SO 4 2− was from local sources, and coarse nssSO 4 2− was partly affected by long-range transport, on dust-event days.Based on equivalent ratios, it was found that fine nss-SO 4 2− was mostly in the form of (NH ) SO  (Sievering et al. 1992;Chameides and Stelson 1992).Ma et al. (2001) examined single particles, using Scanning Electron Microscope (SEM) coupled with an energy dispersive X-ray microanalyzer (EDX), and found that almost all of the single particles, sampled on ADS days, contained sulfur and chlorine.This result indicated that S was released into the atmosphere in gaseous form and deposited onto coarse sea-salt particles, to form sulfuric acid during long-range transport.In addition, the simulation results, of a three-dimensional regional-scale atmospheric chemistry model, showed that chemical conversions of SO 2 to sulfate, in the presence of mineral aerosols, may contribute 20 -40% of total sulfate production (Xiao et al. 1997).Thus, increases in coarse sulfate Table 7. Equivalent ratios of various cations and anions and percentage of Cl − compared to sea-salt (ss) Cl − (%), on dust events and non-dustevent days in Taipei, Spring 2002; D: dust-event, N-D: non-dust-event.
Note: *: the difference between the two periods was significant at 0.1 level; **: significant at 0.05 level.
concentrations, observed on dust-event days in Taipei, were probably due to acidic gases from the Asian continent, which reacted with coarse sea-salt and soil particles, during transport.
As for NO 3 − , which is another important pollutant, concentrations in the fine mode were reduced on dust-event days, while concentrations in the coarse mode increased significantly, during dust events.This behavior was also observed on dust-event days in Taichung in Spring 2002 (Ye 2002).PCA results revealed that fine NO 3 − was mainly derived from local pollution, while coarse NO 3 − was from long-range transport, during dust events.Atmospheric nitrate is formed by the transportation of gaseous nitric acid (or alternatively of NO 2 , NO 3 , or N O 2 5 ) to the particle surface and its subsequent reaction with other gases or particulate compounds.Fine particulate nitrate is mostly formed by a reaction between nitric acid and ammonia vapor, on particle surfaces, to give NH NO 4 3 (s, aq).This reaction requires low temperatures and an abundant amount of NH 3 , more than is necessary to neutralize H SO 2 4 particles (Matsumoto and Tanaka 1996).Atmospheric coarse nitrate can be formed by heterogeneous reactions, between HNO 3 (g) and NaCl (s, aq), on the surface of sea-salt particles.While particulate NaNO 3 is relatively non-volatile, NH NO 4 3 may be converted back to gas.Released particulate NH NO 4 3 may be dissociated to gaseous HNO 3 , which in turn may be transferred to the coarse mode, through reactions with sea-salt particles, during transport over the ocean (Matsumoto et al. 1998;Zhuang et al. 1999).On dust-event days in Taipei, mean NO X concentrations obtained from Guting station, which is located 1.35 km from the CWB sampling site, were 36 ppm compared to 43 ppm, on non-dust-event days.Less fine nitrate was observed, during dust events in Taipei, and was probably due to smaller NO X .On the other hand, elevated coarse particulate nitrate was probably produced during transportation, by heterogeneous reactions of HNO 3 (g) (from the Asian Continent), on the surface of sea-salt and soil particles.
Concentrations of NH 4 + changed slightly on dust-event days, but the changes were not statistically significant.PCA results show that both fine and coarse NH 4 + were from local pollution.NH 4 + is mainly produced by reactions between gaseous NH 3 and acidic sulfate particles.Dust carrying more alkaline coarse particles might prevent these reactions.Decreases in coarse NH 4 + concentration, during the four strong dust-event cases in Taipei, were insignificant.Na + and Cl − ions are the main composition of seawater.Concentrations in both fine and coarse Na + were increased by about twofold on dust-event days.Since, dust-laden air masses need to pass over the ocean before reaching Taiwan, higher concentrations of Na + on dust- event days were anticipated.Concentrations of fine and coarse Cl − were also increased by 1.5 to 2 fold, during dust periods.Comparisons with seawater composition showed that Cl − depletion occurred in the coarse mode.Since reactions of NaCl, of sea-salt particles (mostly in the coarse mode), and acidic gases such as SO 2 and NO X , during transport, can cause chlorine depletion (Sinvering et al. 1992;Kerminen et al. 1997, Zhuang et al. 1999), it was not surprising to see the depletion of coarse Cl − .
With regard to K + , Mg 2+ , and Ca 2+ , concentrations, in both fine and coarse modes, increased during dust events.In addition, C/F ratios of K + , Mg 2+ , and Ca 2+ were correlated with the C/F ratio of Na + .The results of PCA showed that K + , Mg 2+ , and Ca 2+ of dust-event days, were mainly from long-range transport.Since K + , Mg 2+ , and Ca 2+ are important crustal elements of Asian dust, this result was to be expected.The increase of these three elements was also observed in Taichung on dust-event days in 2002 (Ye 2002).
SO 4 2− to Taipei.Since the first principle component (PC1) of both fine and coarse modes represented long-range transport, a positive value of this PC indicated that concentrations, of the aforementioned ions for corresponding samples, were affected by long-range transport.Time series of C/F ratios of Table 3. Rotated component patterns from Principal Component Analysis of the (a) fine and (b) coarse mode.*: values greater than 0.4.Table 4. Weight ratios of ionic species to Na + , during sampling periods in Taipei, Spring 2002.Note: ionic ratios in seawater composition was obtained from Riley & Chester (1971) PM and the values of the first PC are shown in Figs.2a, b.Four strong dust events (namely, 2/11D to 2/12N, 3/6D to 3/9N, 3/18D to 3/19N, and 4/8N to 4/10D) and two weak ones (3/ 23D-3/23N and 4/17N-4/18D), during the sampling periods identified by Chou et al. (2004), are marked in the graph.For both fine and coarse modes, most samples of strong dust events had positive PC1 values, except for those of the fine mode in the fourth case (4/8N to 4/10D), which was relatively weak, in terms of its effects on water-soluble ions.