- Email: [email protected]
The influence of meteorology and atmospheric transport patterns on the chemical composition of rainfall in south-east England
PII:
Atmospheric Environment Vol. 32, No. 6, pp. 1039—1048, 1998 ( 1998 Elsevier Science Ltd All rights reserved. Printed in Great Britain S1352–2310(97)00365–8 1352—2310/98 $19.00#0.00
THE INFLUENCE OF METEOROLOGY AND ATMOSPHERIC TRANSPORT PATTERNS ON THE CHEMICAL COMPOSITION OF RAINFALL IN SOUTH-EAST ENGLAND I. J. BEVERLAND,*,- J. M. CROWTHER,‡ M.S.N. SRINIVAS° and M.R. HEAL± - Department of Public Health Sciences, University of Edinburgh, Medical School, Teviot Place, Edinburgh, EH8 9AG; U.K.; ‡ Department of Physical Sciences, Glasgow Caledonian University, Glasgow, G4 0BA. U.K.; ° Central Scientific Instruments Organisation, Environmental Monitoring Instruments Division, Sector 30, Chandigarh 160 020, India; and ± Department of Chemistry, University of Edinburgh, King’s Buildings, West Mains Road, Edinburgh, EH9 3JJ, U.K. (First received 27 February 1997 and in final form 21 August 1997. Published March 1998) Abstract—Rainwater composition was examined at event temporal resolution, over a 6 month period, at a site in southeast England. The data were used to assess the overall levels of acidic deposition at the site, and to identify functional relationships between wet deposition and causal meteorological processes. Rainfall-weighted average concentrations were close to those estimated from the U.K. national acid deposition network, but deposition levels were below those suggested by network data because of the unusually dry summer in 1989. The rainfall chemistry data were related to locally recorded wind direction, and to back trajectories calculated with analysed wind field data from U.K. Meteorological Office numerical weather prediction models. The influence of local wind direction during rainfall was significant in terms of observed concentrations and deposition. However back trajectory analysis was a better indicator of the Lagrangian history and pollutant loading of the air masses reaching the site, with clear differences in rainfall composition noted between different transport patterns. A power-law relationship existed between wet deposition and rainfall amount, although the data exhibited considerable scatter around the functional relationship because of the influence of transport pattern. Proportional ionic composition was also influenced by transport pattern with enhanced chloride levels for maritime events. The nitrate : sulphate ratio was inversely related to the time of travel from major anthropogenic source regions. ( 1998 Elsevier Science Ltd. All rights reserved. Key word index: Acid rain, acid deposition, precipitation events, back trajectory analysis, atmospheric transport patterns.
1. INTRODUCTION
Episodes of acid deposition can be highly damaging to biological systems (Smith and Hunt, 1978; Fowler and Cape, 1984a; Dana and Slinn, 1988). Wet deposition episodes are highly dependent on the interaction of anthropogenic emissions and specific meteorological conditions, especially air-mass histories and rainfall amount. For example, it has been noted that a high proportion of episodes in northern Europe are preceded by stagnating anticyclonic conditions over industrialised central or eastern regions (Laurila and Joffre, 1987; Fowler and Cape, 1984b; Smith, 1983). Pollutants accumulated in the air mass near to the
* Correspondence should be addressed to: Dr Iain J. Beverland, Lecturer in Environmental Health, Department of Public Health Sciences, University of Edinburgh, Medical School, Teviot Place, Edinburgh, EH8 9AG, U.K. Tel.: 0131 650 3212. Fax: 0131 650 6909. E-mail: [email protected].
centre of the anticyclone can be advected to the northwest, to be intercepted by precipitation systems over sensitive locations. Most acid-deposition-monitoring data are collected on time scales of 1 d or longer. However, back trajectories, hence air-mass source regions, can change markedly in time periods of less than 12 h (Beverland and Crowther, 1992). Therefore, to gain better understanding of transport mechanisms it is necessary to make measurements at higher temporal resolution. We used a microprocessor-based monitor to make real-time measurements of the conductivity and pH of sequential rain samples (Beverland et al., 1996). The samples were stored on an event (i.e. complete storm), and sometimes sub-event, basis for subsequent laboratory analysis. This system was operated for a 6month period at a site in south east England during which episodes of deposition were identified. We describe the overall level of deposition during this period, identify functional relationships between wet deposition and local meteorological process and
1039
I. J. BEVERLAND et al.
1040
discuss the relative importance of local wind direction, atmospheric transport pattern and rainfall amount.
2. DATA COLLECTION
2.1. Method for sampling wet deposition The microprocessor-based acid rain monitor collected 0.5 mm (15 ml) incremental rain samples via a tipping bucket mechanism. The sample temperature, conductivity and pH were measured by electrodes housed in PTFE chambers. These measurements were logged by the microprocessor together with the time of tipping bucket movement. Automated daily checks on performance of the pH probe in the acid rain monitor were made using a simulated rain material (NIST SRM 2694-II; 0.5]10~3 M; pH"3.56; conductivity"130 kS cm~1). The details of SRM preparation are given by Koch (1986) and Koch et al. (1982). After pH measurement the sample was transferred to a sequential sampling unit. The monitor was programmed to store the samples on an event, or sub-event, basis. Containers of 500 ml polyethylene and 25 ml polypropylene were used to sample event and sub-event rainfall, respectively. The containers were held tightly against a rigid polyethylene sheet to prevent any contamination and to minimise evaporation of the sample. Where possible, rainwater contact surfaces were constructed from polyethylene, polypropylene or polytetrafluoroethylene (PTFE). Rinsing of any surfaces not constructed from these materials indicated negligible effects on ion concentrations. Dry contamination on the collector funnel was minimised by rinsing with distilled water, as close as possible to the start of rainfall. In most cases this was considerably less than 3 h. On a few occasions when rainfall began in the middle of the night the funnel may have been exposed to dry contamination for up to 12 h. However the back trajectories and the sub-event chemistry patterns for these events suggested that this contamination did not unduly influence the event-averaged data for these rainfalls. We have noted these events carefully but have not excluded them from the event-based analysis presented here. Contamination tests were conducted by analysing distilled water passed through the wet-deposition monitor. SO2~ concentrations in this water varied between 0 and 0.6 keq 4l~1 and NO~ was 3 undetectable [cf. unpolluted background levels for the Northern hemisphere of 3.27 and 2.86 keq l~1, respectively (Szepesi and Fekete, 1987)]. The tests demonstrated that, in addition to the rinsing of the collection funnel, regular cleaning of the system was necessary to ensure contamination free samples. Therefore all pipelines and chambers were cleaned with distilled water on a weekly basis. Contamination tests conducted during the field campaign indicated negligible levels of contamination (Beverland et al., 1996). 2.2. Sampling site Continuous rainfall chemistry measurements were made from June to November 1989 at the Meteorological Office Experimental Site at Beaufort Park, Bracknell, Berkshire in south-east England (51 24N 0 45W). With the exception of the high density of road traffic throughout Berkshire there were no significant local sources of atmospheric pollution at this site on the western outskirts of a new town area. Previous studies within U.K. cities suggest that the precipitation composition would not have been significantly influenced by the suburban nature of the sampling location (RGAR, 1987). 2.3. Sampling strategy and laboratory analysis A total of 67 rainfall events were sampled during the field experiment. An event was defined as a minimum rainfall amount of 0.5 mm. A period of not less than 2 h between 0.5 mm rain increments separated events.
Precipitation samples were collected from the field site as soon as possible after the cessation of rainfall and filtered through 0.4 km Teflon filters. Sixty events were analysed for the anions: chloride ([Cl~]), nitrate ([NO~]) and sulphate 3 (sodium [Na], ([SO2~]) using ion chromatography. Cation 4 magnesium [Mg], and calcium [Ca]) measurements were made for some of these events using an atomic absorption spectrometer. Analytical resources were not sufficient to determine all of the ionic compounds necessary for ion balance quality assurance calculations (e.g. Peden, 1983). However laboratory tests using prepared standards and synthetic rainwater samples indicated that the overall uncertainties (i.e. both precision and accuracy) in our ion chromatographic determinations were less than 10% of typical concentrations. It is unlikely that these uncertainties would have influenced the conclusions of the study. Rainfall-weighted average values were calculated as follows: n xx" +
RA i x i RA 505!i/0
(1)
where xx"volume weighted average (kS cm~1 or keq l~1); n"number of events; RA "the rainfall amount for the ith i event (mm); RA "the total amount of rainfall deposited 505!(mm); x "the conductivity or concentration for the ith i event (kS cm~1 or keq l~1)
3. RESULTS
3.1. Summary of wet deposition statistics The rainfall-weighted average of hydrogen-ion concentration ([H`]) was 37 keq l~1 (Table 1), which was close to the annual average calculated from the U.K. national acid deposition monitoring network data (RGAR, 1987). Due to the proximity of major European sources, the rainfall-weighted pH of 4.4 was significantly lower than ‘‘background’’ or ‘‘unpolluted’’ pH values which generally range between 5.0 and 5.6 (Galloway et al., 1982). The rainfall-weighted averages of [Cl~], [NO~] and [SO2~] concentra3 4 tions were similar to the corresponding averages calculated for the national network data. With the exception of [H`], the rainfall-weighted averages were lower than the corresponding arithmetic averages. This was due to low concentrations towards the end of longer events (Beverland and Crowther, 1992). The opposite trend for [H`] was probably the result of low concentrations of neutralising species in the later stages of events. The coefficients of variation for the event concentration data approached or exceeded 100%, ranging from 92% for [Cl~] to 148% for [NO~] 3 (Table 1). These large values reflect considerable interevent variability. Frequency histograms of the concentration data showed strong positive skewness (skewness statistic ranged between 2.0 and 2.4) suggesting that the underlying distributions were lognormal. The use of histograms to assess closeness to normality is subjective and dependent on the class interval chosen. To make objective assessments of normality the Kolmogorov—Smirnov (K—S) Test was used (Ebdon, 1981). K—S tests on the data indicated that
46 48 102 72 101 181 87 66 Conductivity H` Cl~ NO~ 3 SO2~ 4 Na Mg Ca
25 kS cm~1 37 keq l~1 69 keq l~1 38 keq l~1 63 keq l~1 232 keq l~1 40 keq l~1 59 keq l~1 66 65 60 60 59 26 25 26
Measurement
35 kS cm~1 37 keq l~1 111 keq l~1 49 keq l~1 86 keq l~1 191 keq l~1 55 keq l~1 92 keq l~1
No. of events
Note: Rainfall weighted mean concentrations and wet deposition for southeast England (1981—1985) are given for comparison. (Data obtained from RGAR, 1987).The Beaufort Park SO2~ data 4 was not corrected for marine contribution (because of the low number of [Na] determinations).
— 0.02—0.04 g H m~2 yr~1 — 0.3—0.4 g N m~2 yr~1 0.5—1.0 g S m~2 yr~1 — — — — 0.01 g H m~2 yr~1 — 0.2 g N m~2 yr~1 0.4 g S m~2 yr~1 — — —
— 40 keq l~1 75 keq l~1 31 keq l~ 64 keq l~ — — — — 198 174 281 198 198 167 101
Deposition
131 132 92 148 118 95 159 72
Coefficient of variation for deposition (%) Rainfallweighted mean concentration Arithmetic mean concentration
Standard deviation (s.d)
Coefficient of variation for concentration (arithmetic mean /s.d.) (%)
Table 1. Statistical summary for the event data set
Concentration from U.K. Network
Deposition from U.K. Network
Chemical composition of rainfall in south-east England
1041
the natural logarithm of concentrations gave sample distributions which were significantly closer to normality compared with the un-transformed variables (logarithmic transformation increased K—S a from 0.00—0.02 to 0.20). Similar observations have been made in statistical analyses of other rainfall chemistry data sets (Lindberg, 1982; Singh et al. 1987). Subsequent statistical analyses were therefore carried out on the log transformed concentration data. The total wet depositions for the sampling period were calculated from the product of total rainfall amount and the rainfall-weighted average concentrations (Table 1). H`, NO~ and SO2~ depositions were 3 4 lower than the national network estimates, reflecting the exceptionally dry weather during much of the field experiment. The coefficients of variation for the event deposition data were all greater than 100% and exceeded those for concentration values. The greater inter-event variation results from the additional variability introduced by the rainfall amount, which outweighs the tendency for lower concentrations in higher volume rainfall events. Deposition data tended to be log-normally distributed (K—S a"0.00 for untransformed deposition data, K—S a"0.20 for logarithmic transformed deposition data). 3.2. Relationship between concentration/deposition and wind direction t-Tests were used to determine if significant differences existed between samples from different local, ground-level wind-direction sectors. The tests assumed that the sample distributions to be tested were normal therefore log transformed concentrations % were used (Section 3.1). Ninety-degree wind direction sectors were chosen. The average wind direction was calculated using procedures outlined by Mori (1986, 1987). Subdividing the data set meant that the number of data points in each wind direction sector was quite small which may have violated the assumption of normal sample distributions. To guard against this, the non-parametric Mann Whitney test was also used. The analyses tested the null hypothesis that the mean or median of the log concentrations, c, occurring % when the average wind direction was from sector i was less than, or equal to, the mean/median when the averaged wind direction was from sector j (H : c )c , 0 i j iOj) compared with the alternative, that the mean/median concentration was greater (H : c 'c ). 1 i j Critical values at the 90, 95, 99 and 99.9% confidence levels were used. For each ionic compound, highest concentrations occurred when the wind had a north-easterly component. The concentration differences with respect to the other sectors were significant at 99% and 95% confidence levels for most compounds for the parametric and non-parametric tests, respectively. Chloride was the only exception and showed no significant directional dependence. Sperber (1987) suggests that the average local wind direction is a good indicator of the source regions of air which fuel the storm systems
1042
I. J. BEVERLAND et al.
and contribute to pollutant loading. The initial examinations of the data seem to support this claim. The higher concentrations observed when local wind direction was in the northeast sector may be associated with increased urbanisation in the east from Bracknell to the London conurbation. The event-averaged concentration data were plotted against the average wind direction calculated for the rainfall period. It was difficult to interpret underlying trends from these scatter plots because the number of events changed with the level of the average wind direction. Cleveland and Kleiner (1975) introduced a method, analogous to a moving average, to deal with this scatter plot perception problem. They made use of an existing statistic, the midmean, and introduced two new statistics: the upper and lower semi-midmeans. The midmean is defined as the average of all observations between and including the quartiles. The upper and lower semi-midmeans are the midmeans of all observations above and below the median respectively. Exclusion of the upper and lower quartiles in the calculation of the midmean removes extreme values which may distort the underlying trend. While the plot of the rainfall-weighted mean concentrations against wind direction sector is representative of average conditions, the use of moving statistics gives additional information about the behaviour of the upper and lower tails of the concentration distribution (Cleveland and Kleiner, 1975; Sperber, 1987). For each compound (except [Cl~]) concentrations were greater when accompanying winds were from the northeastern quadrant (Figs 1A—D). For [H`] and [NO~] the lower semi-midmean, the midmean and 3 the upper semi-midmean showed similar patterns when plotted against mean wind direction within the four sectors. The sector mean for [Cl~] appeared to be almost constant. However, this masked some directional dependency which was evident in the moving statistics. Both the midmean, and the upper semi-midmean, showed minimum [Cl~] for the southeasterly sector because of the absence of significant marine sources in this quadrant. The [Cl~] upper semi-midmean showed more variability compared to the other statistics. Peak [Cl~] in the northeasterly quadrants may have resulted from anthropogenic input of HCl to specific events. Similarly, the secondary peak in the northwestern sector was probably caused by high marine [Cl~] input. The moving statistics for [SO2~] showed highest concentrations in the 0—90° 4 sector. However, the upper semi-midmean, and to a lesser extent the mean, were increased in the 90—180° sector by a small number of high concentration episodes. A clear pattern emerged from the moving statistics for the deposition data with the largest depositions in the 0—90° sector (deposition data are not shown graphically). This resulted from association of high concentrations in this sector with moderate or high rainfall amounts.
3.3. Inter-species and inter-sector concentration variability Sperber (1987) found the variability of [NO~] to be 3 significantly greater than [SO2~], in all wind direc4 tion sectors, from 6 yr of hourly precipitation data obtained at Long Island, New York, and suggested that SO and SO2~ aerosol are more homogeneously 2 4 mixed in the atmosphere than NO~. Examination of 3 the data set from Beaufort Park indicated general agreement with Sperber’s findings although the increased variability of NO~ was only statistically sig3 nificant (F-test at 95% confidence level) for the northwest wind direction sector and the unclassified data set. Sperber (1987) found variability in ionic concentrations to be less for the sector which represented the major axis of pollutant input. F-tests carried out on the Beaufort Park data did not support this theory. 3.4. ¹he relationship between concentration, deposition and atmospheric transport pattern Back trajectories were calculated, at the 950 mb level, for each rain event using the model described by Maryon and Heasman (1988). Trajectory start times were selected to be as near as possible to the middle of rain events. The trajectories were classified into 6 different transport patterns (Figs 2A—F). Differences in mean concentration and deposition values, for each transport pattern, were examined by using both parametric and non-parametric tests (as described in Section 3.2). There were clear differences between the transport patterns. [NO~] and [SO2~] concentra3 4 tions were greatest in the North Sea and Eastern European transport regimes (Figs 2A and B, respectively) with lowest concentrations in the mixed maritime/continental and southwesterly marine regimes (Figs 2E and F, respectively). A similar pattern was noted for the deposition values. Using the transport pattern classification, significant differences were noted over a larger number of independent variable classes, compared with the wind direction sector classification (e.g. for NO~ significant 3 differences were observed between 5 of the transport patterns but only between 2 of the wind direction sectors). This emphasises the superiority of back trajectory analysis for interpretative purposes. Contingency table analysis showed that the events from the 0—90° wind direction sector were comprised entirely of transport patterns A and B. However the situation was much less clear for the other sectors, and for each individual transport pattern. Therefore, in contrast to our initial findings in Section 3.2, we conclude that local wind direction data, averaged for the duration of an event, was not a particularly good indicator of the Lagrangian history of the air masses supplying pollutants to the precipitation system. 3.5. ¹he significance of rainfall amount In accordance with several other studies (Khwaja and Husain, 1990; Lindberg, 1982; Davies, 1976; Huff
Fig. 1. Moving statistics for event-averaged concentrations. Data cover period from June to November 1989. Statistics are based on wind direction average for the periods during rainfall (then classified and averaged within 90° sectors).
Chemical composition of rainfall in south-east England 1043
1044
I. J. BEVERLAND et al.
Fig. 2. Classified atmospheric transport patterns.
and Stout, 1964; Bleeker et al., 1966) we observed a marked decrease in event-average ionic concentrations as precipitation amount increased. We examined the correlations between: (a) event average concentration against log (rainfall amount); and (b) % log (concentrations) against log (rainfall amount). % % The correlation coefficients were higher for the powerlaw relationship [regression (b)], however in all cases the coefficients were not statistically significant (P(0.05), and we concluded that precipitation volume alone is not useful in explaining event-averaged concentrations. Simplified washout theory suggests exponential decreases of pollutant concentrations in air, and therefore also in rain, with accumulated rainfall amount during the course of individual events. Dawson (1978) found significant negative correlations between the logarithm of the concentration of ionic species and the accumulated rainfall amount. Similar analyses of the sub-event Beaufort Park data showed virtually
no correlation. It might be concluded that the earlier observations of Dawson were fortuitous and sitespecific. Therefore the simplified picture of washout is inadequate on event time scales when advective processes must also be considered (Beverland and Crowther, 1992). Previous studies (Hicks and Shannon, 1979; MacCracken, 1979; Lindberg, 1982; Khwaja and Husain, 1990) have found a power-law relationship between wet deposition levels (D) and rainfall amount (h): D"K1 hK2
(2)
where K1 and K2 are constants. Linear regression between the log transformed variables was used to % test this power-law hypothesis. There was considerable scatter in the relationships derived for the entire data set (Fig. 3). However the regression models were all significant (P(0.001) suggesting the presence of underlying functional relationships (Table 2). One of the main reasons for the
Chemical composition of rainfall in south-east England
1045
Fig. 3. Relationships between deposition and total rainfall amount. (A) Relationship between deposited H` and rainfall amount. (B) Relationship between deposited NO~, and rainfall amount. (Regression lines 3 shown are calculated from data from all events.)
Table 2. Regression analysis of wet deposition rainfall amount Data set Complete data set (67 events) SW marine transport pattern (42 events) Eastern European transport pattern (5 events)
Regression equation log % log % log % log % log % log % log % log % log % log % log % log %
R2 (%)
n
Standard error of slope
47 50 37 44 63 58 63 59 91 93 95 98
65 60 60 59 41 39 39 39 5 5 5 5
0.15 0.08 0.14 0.12 0.15 0.10 0.11 0.11 0.16 0.10 0.13 0.07
(H`) "2.80#1.14 log (rainfall amount) % (Cl~) "4.53#0.61 log (rainfall amount) % (NO~) "3.17#0.83 log (rainfall amount) 3 % (SO2~) "4.01#0.79 log (rainfall amount) 4 % (H`) "2.36#1.19 log (rainfall amount) % (Cl~) "4.44#0.69 log (rainfall amount) % (NO~) "2.54#0.87 log (rainfall amount) 3 % (SO2~) "3.58#0.78 log (rainfall amount) 4 % (H`) "4.98#0.84 log (rainfall amount) % (Cl~) "4.29#0.62 log (rainfall amount) % (NO~) "5.03#0.92 log (rainfall amount) 3 % (SO2~) "5.57#0.84 log (rainfall amount) 4 %
Note: The regression models were all statistically significant (slopeO0, P(0.001).
scattering may have been the influence of the prevailing transport pattern on the rainwater ionic concentration. Regression analyses were conducted on data from individual transport patterns. We restricted these analyses to the larger data sets in the southwesterly marine transport and the eastern European transport patterns (Figs 2F and 2B, respectively). Considerably higher R2 values were observed, demonstrating that much of the scatter in Fig. 3 was due to transport pattern effects (Table 2). The power-law exponents calculated for [NO~] 3 and [SO2~] were not significantly different for dif4 ferent transport patterns, suggesting that physical scavenging processes were responsible for the underlying functional relationship (Table 2). In contrast, the intercept [or log (K1)] values were significantly % different emphasising the importance of transport pattern in modelling studies utilising these types of
relationships. Significant differences were noted in the exponents calculated for different compounds. Hicks and Shannon (1979) found power-law exponents of 0.5 and 0.6 for the deposition of radioactive isotopes from nuclear bomb testing and sulphur deposition respectively. Lindberg (1982) calculated exponents of 0.72$0.08 for SO2~ and 0.86$0.08 for 4 H`. Lindberg reported that the sulphate value of 0.72 was not statistically different from the 0.6 exponent calculated by Hicks and Shannon, however there appeared to be evidence of a larger exponent for H`. Khwaja and Husain (1990) found exponents of 0.71, 0.55 and 0.72 for SO2~, NO~ and Cl~ respectively. 4 3 The exponents calculated from the Beaufort Park data were higher than the findings of the studies mentioned above (Table 2). The observed differences were statistically significant (P(0.005) for the data set as a whole and for most of the individual transport patterns.
1046
I. J. BEVERLAND et al.
The present study emphasises that transport pattern, site and species dependence must be considered if power-law relationships are used in acid-deposition models. 3.5. Proportional anion composition of event rainfall data In accordance with our findings in the previous section Davies (1988) suggested that much of the variation in precipitation composition is caused by variations in the overall concentration due to dilution effects. Influences other than dilution can be examined using a triangular diagram based on the primary components. Figure 4 shows the relative proportions of the anions NO~, SO2~ and Cl~. For the Beaufort 3 4 Park data there was a clear distinction between events with continental and marine transport patterns (Cl~ below 20% of total anions and Cl~ above 30%, respectively). The highly polluted events, with trajectories looping out over the North Sea (Fig. 2A), exhibited intermediate proportions.
The marine origin events appeared to lie along a line of increasing proportion of Cl~. The events with the highest proportion of Cl~ were the autumnal storms associated with gale force winds. The high wind speeds during air-mass passage over the oceans would have enhanced the formation and atmospheric mixing of chloride rich aerosol and rapid passage over essentially rural parts of England would have provided little opportunity for entrainment of anthropogenic nitrogen and sulphur species. Davies (1988) and Skartveit (1982) found that the [NO~] : [SO2~] ratio in precipitation decreased with 3 4 increasing travel time of polluted air. This was attributed to faster oxidation of nitrogen oxides compared to sulphur dioxide, resulting in nitric acid becoming available for removal before sulphuric acid (Rodhe et al., 1981). The estimated travel times from source regions during some of the continental origin events were noted for trajectories which were judged to have passed over major sources (i.e. the 6 labelled events in Fig. 4). In agreement with the earlier studies,
Fig. 4. Relative proportions of the anions NO~, SO2~ and Cl~ in rainfall (event-averaged). NO~ 3 4 3 proportions are read from the left-hand side edge using the lines which slope downward to the right. Cl~ proportions are read from the bottom edge using the lines which slope upwards to the right and SO2~ 4 proportions are read from the right-hand side edge using the horizontal lines (Arrows next to NO~, SO2~ 3 4 and Cl~ labels indicate how to read the proportion of these anions). Transport times (in days) are plotted to the immediate right of 6 events. These events had trajectories which passed over major source regions.
Chemical composition of rainfall in south-east England
lower [NO~] : [SO2~] ratios were associated with 3 4 increased travel time from anthropogenic source regions. In particular the two events with high ('40%) NO~ proportions exhibited slow passage over south3 east England, within 12 h of rainfall. The high density of vehicular traffic in the home counties may have caused the elevated proportion of NO~. 3
4. CONCLUSIONS
f The rainfall-weighted average concentrations for the 6 month field experiment at Beaufort Park were close to those estimated from the U.K. national acid deposition network. The deposition levels were somewhat below those suggested by network data because of the unusually dry summer of 1989. f The event temporal resolution attainable by the acid-rain monitor enabled a detailed study of the relationships between wet deposition, wind direction and atmospheric transport patterns. The influence of local wind direction during rainfall was significant in terms of observed concentrations and deposition. However back trajectory analysis was a better indicator of the Lagrangian history and pollutant loading of the air masses reaching the site. Pronounced differences were observed between transport patterns, with significantly enhanced deposition in the North Sea and Eastern European categories. The use of moving statistics was useful for interpretation of variations of acid deposition with changes in wind direction. f Concentration of pollutant species tended to be inversely related to precipitation amount. However other factors were also significant, and precipitation amount alone cannot be used to explain event average concentrations. Virtually no correlation was observed between log-transformed sub-event concentrations and rainfall amount, emphasising that simplified scavenging theory fails to account for the important influence of advection. f A power-law relationship existed between wet deposition and rainfall amount. The data exhibited considerable scatter around the functional relationship. Much of this scatter was attributed to the influence of transport pattern. The power-law exponents calculated were significantly greater than those reported in similar studies in North America. Significant differences were noted in the exponents calculated for different ionic compounds. f A triangular diagram of the principal anionic compounds provided a useful graphical interpretation of the influence of transport pattern on rainwater composition. Enhanced [Cl~] proportions were evident for maritime events. The [NO~] : [SO2~] 3 4 ratio was inversely related to the time of travel from major anthropogenic source regions. Acknowledgements—The authors wish to thank Dr F. B. Smith of the U.K. Meteorological Office and Miss G. Taylor
1047
of the University of Edinburgh for helpful comments. We are grateful for technical support from Messrs J. A. Broadfoot, M. A. Wallace, J. Revie, and R. W. Weston of the University of Strathclyde; and for assistance from Mr D. H. O¨Ne´ill and Ms J. P. Brown of the University of Edinburgh in the preparation of this manuscript. Financial and technical support from the U.K. Natural Environment Research Council and the U.K. Meteorological Office are acknowledged.
REFERENCES
Beverland, I. J. and Crowther, J. M. (1992) On the interpretation of event and sub-event rainfall chemistry. Environmental Pollution 75, 163—174. Beverland, I. J., Crowther, J. M. and Srinivas, M. S. N. (1996) Development and operation of a microprocessor-based acid rain monitor. Atmospheric Environment 30, 3611—3622. Bleeker, W., Dansgaard, W. and Lablans, W. (1966) Some remarks on simultaneous measurements of particulate contaminants including radioactivity and isotopic composition of precipitation. ¹ellus 18, 773—785. Cleveland, W. S. and Kleiner, B. (1975) A graphical technique for enhancing scatterplots with moving statistics. ¹echnometrics 17, 447—453. Dana, M. T. and Slinn, W. G. N. (1988) Acidic deposition distribution and episode statistics from the MAP3S network database. Atmospheric Environment 22, 1469—1474. Davies, T.D. (1976) Precipitation scavenging of sulphur dioxide in an industrial area. Atmospheric Environment 10, 879—890. Davies, T. D. (1988) Chemical composition of snow in the remote Scottish Highlands. In: Acid Deposition of High Elevation Sites, eds. M. H. Unsworth and D. Fowler, pp. 517—539. Kluwer Academic Publishers, Dordrecht, The Netherlands. Dawson, G. A. (1978) Ionic composition of rain during sixteen convective showers. Atmospheric Environment 12, 1991—1999. Ebdon, D. (1981) Statistics in Geography: A Practical Approach. Basil Blackwell, Oxford. Fowler, D. and Cape, J. N. (1984a) The contamination of rain samples by dry deposition on rain collectors. Atmospheric Environment 18, 183—189. Fowler, D. and Cape, J. N. (1984b) On the episodic nature of wet deposited sulphate and acidity. Atmospheric Environment 18, 1859—1866. Galloway, J. N., Likens, G. E., Keene, W. C. and Miller, J. M. (1982) The composition of precipitation in remote areas of the World. Journal of Geophysical Research 87, 8771—8786. Hicks, B. B. and Shannon, J. D. (1979) A method for modelling the deposition of sulphur by precipitation over regional scales. Journal Applied Meteorology 18, 1415—1420. Huff, F.A. and Stout, G. E. (1964) Distribution of radioactive rainout in convective rainfall. Journal of Applied Meteorology 3, 707—717. Khwaja, H. A. and Husain, L. (1990) Chemical characterisation of acid precipitation in Albany, New York. Atmospheric Environment 24A, 1869—1882. Koch, W. F. (1986) Standard reference materials: methods and procedures used at the National Bureau of Standards to prepare, analyse and certify SRM2694, simulated rainwater, and recommendations for use, U.S. Department of Commerce/National Bureau of Standards, Washington D.C. U.S.A. Koch, W. F., Marinenko, G. and Stolz, J. W. (1982) Simulated Reference Materials, Vol. IV, National Measurements Laboratory, Inorganic analytical research division, Washington D.C., U.S.A.
1048
I. J. BEVERLAND et al.
Laurila, T. and Joffre, S. M. (1987) Meteorological analysis of wet deposition in Finland. In Acid Rain: Scientific and ¹echnical Advances, ed. R. Perry, pp. 167—174. Selper Publishing, London. Lindberg, S. E. (1982) Factors influencing trace metal, sulphate and hydrogen ion concentrations in rain. Atmospheric Environment 16, 1701—1709. MacCracken, M. C. (1979) MAP3S progress report for F½1977 and F½1978, DOE/EV-0040, US Department of Energy, Washington D.C., U.S.A. Maryon, R. H. and Heasman, C. C. (1988) The accuracy of plume trajectories forecast using the U.K. Meteorological Office operational forecasting models and their sensitivity to calculation schemes. Atmospheric Environment 22, 259—272. Mori, Y. (1986) Evaluation of several ‘single-pass’ estimators of the mean and the standard deviation of wind direction. Journal of Climate and Applied Meteorology 25, 1387—1397. Mori, Y. (1987) Methods for estimating the mean and the standard deviation of wind direction. Journal of Climate and Applied Meteorology 26, 1282—1284. Peden, M. E. (1983) Sampling, analytical and quality assurance protocols for the national atmospheric deposition program. In Sampling and Analysis of Rain, ed. S. A. Campbell, AS¹M S¹P 823, pp. 72—83.
RGAR (1987) Acid deposition in the U.K. 1981—1985. Second report of the U.K. Review Group on Acid Rain, Warren Spring Laboratory, Stevenage, U.K. Rodhe, H., Crutzen, P. and Vanderpol, A. (1981) Formation of sulphuric and nitric acids in the atmosphere during long range transport. ¹ellus 33, 132—141. Singh, B., Nobert, M. and Zwack, P. (1987) Rainfall acidity as related to meteorological parameters in Northern Quebec. Atmospheric Environment 21, 825—842. Skartveit, A. (1982) Wet scavenging of sea-salts and acid compounds in a rainy, coastal area. Atmospheric Environment 16, 2715—2724. Smith, F. B. (1983) Conditions pertaining to high acid concentrations in rain. In Proceedings of ».D.I. Conference on: Acid precipitation — origin and effects, Lindau, Germany. Smith, F. B. and Hunt, R. D. (1978) Meteorological aspects of the transport of pollution over long distances. Atmospheric Environment 12, 461—467. Sperber, K. R. (1987) The concentration and deposition of nitrate, sulphate and ammonium as a function of wind direction from precipitation samples. Atmospheric Environment 21, 2629—2641. Szepesi, D. J. and Fekete, K. E. (1987) Background levels of air and precipitation quality for Europe. Atmospheric Environment 21, 1623—1630.
Atmospheric Environment Vol. 32, No. 6, pp. 1039—1048, 1998 ( 1998 Elsevier Science Ltd All rights reserved. Printed in Great Britain S1352–2310(97)00365–8 1352—2310/98 $19.00#0.00
THE INFLUENCE OF METEOROLOGY AND ATMOSPHERIC TRANSPORT PATTERNS ON THE CHEMICAL COMPOSITION OF RAINFALL IN SOUTH-EAST ENGLAND I. J. BEVERLAND,*,- J. M. CROWTHER,‡ M.S.N. SRINIVAS° and M.R. HEAL± - Department of Public Health Sciences, University of Edinburgh, Medical School, Teviot Place, Edinburgh, EH8 9AG; U.K.; ‡ Department of Physical Sciences, Glasgow Caledonian University, Glasgow, G4 0BA. U.K.; ° Central Scientific Instruments Organisation, Environmental Monitoring Instruments Division, Sector 30, Chandigarh 160 020, India; and ± Department of Chemistry, University of Edinburgh, King’s Buildings, West Mains Road, Edinburgh, EH9 3JJ, U.K. (First received 27 February 1997 and in final form 21 August 1997. Published March 1998) Abstract—Rainwater composition was examined at event temporal resolution, over a 6 month period, at a site in southeast England. The data were used to assess the overall levels of acidic deposition at the site, and to identify functional relationships between wet deposition and causal meteorological processes. Rainfall-weighted average concentrations were close to those estimated from the U.K. national acid deposition network, but deposition levels were below those suggested by network data because of the unusually dry summer in 1989. The rainfall chemistry data were related to locally recorded wind direction, and to back trajectories calculated with analysed wind field data from U.K. Meteorological Office numerical weather prediction models. The influence of local wind direction during rainfall was significant in terms of observed concentrations and deposition. However back trajectory analysis was a better indicator of the Lagrangian history and pollutant loading of the air masses reaching the site, with clear differences in rainfall composition noted between different transport patterns. A power-law relationship existed between wet deposition and rainfall amount, although the data exhibited considerable scatter around the functional relationship because of the influence of transport pattern. Proportional ionic composition was also influenced by transport pattern with enhanced chloride levels for maritime events. The nitrate : sulphate ratio was inversely related to the time of travel from major anthropogenic source regions. ( 1998 Elsevier Science Ltd. All rights reserved. Key word index: Acid rain, acid deposition, precipitation events, back trajectory analysis, atmospheric transport patterns.
1. INTRODUCTION
Episodes of acid deposition can be highly damaging to biological systems (Smith and Hunt, 1978; Fowler and Cape, 1984a; Dana and Slinn, 1988). Wet deposition episodes are highly dependent on the interaction of anthropogenic emissions and specific meteorological conditions, especially air-mass histories and rainfall amount. For example, it has been noted that a high proportion of episodes in northern Europe are preceded by stagnating anticyclonic conditions over industrialised central or eastern regions (Laurila and Joffre, 1987; Fowler and Cape, 1984b; Smith, 1983). Pollutants accumulated in the air mass near to the
* Correspondence should be addressed to: Dr Iain J. Beverland, Lecturer in Environmental Health, Department of Public Health Sciences, University of Edinburgh, Medical School, Teviot Place, Edinburgh, EH8 9AG, U.K. Tel.: 0131 650 3212. Fax: 0131 650 6909. E-mail: [email protected].
centre of the anticyclone can be advected to the northwest, to be intercepted by precipitation systems over sensitive locations. Most acid-deposition-monitoring data are collected on time scales of 1 d or longer. However, back trajectories, hence air-mass source regions, can change markedly in time periods of less than 12 h (Beverland and Crowther, 1992). Therefore, to gain better understanding of transport mechanisms it is necessary to make measurements at higher temporal resolution. We used a microprocessor-based monitor to make real-time measurements of the conductivity and pH of sequential rain samples (Beverland et al., 1996). The samples were stored on an event (i.e. complete storm), and sometimes sub-event, basis for subsequent laboratory analysis. This system was operated for a 6month period at a site in south east England during which episodes of deposition were identified. We describe the overall level of deposition during this period, identify functional relationships between wet deposition and local meteorological process and
1039
I. J. BEVERLAND et al.
1040
discuss the relative importance of local wind direction, atmospheric transport pattern and rainfall amount.
2. DATA COLLECTION
2.1. Method for sampling wet deposition The microprocessor-based acid rain monitor collected 0.5 mm (15 ml) incremental rain samples via a tipping bucket mechanism. The sample temperature, conductivity and pH were measured by electrodes housed in PTFE chambers. These measurements were logged by the microprocessor together with the time of tipping bucket movement. Automated daily checks on performance of the pH probe in the acid rain monitor were made using a simulated rain material (NIST SRM 2694-II; 0.5]10~3 M; pH"3.56; conductivity"130 kS cm~1). The details of SRM preparation are given by Koch (1986) and Koch et al. (1982). After pH measurement the sample was transferred to a sequential sampling unit. The monitor was programmed to store the samples on an event, or sub-event, basis. Containers of 500 ml polyethylene and 25 ml polypropylene were used to sample event and sub-event rainfall, respectively. The containers were held tightly against a rigid polyethylene sheet to prevent any contamination and to minimise evaporation of the sample. Where possible, rainwater contact surfaces were constructed from polyethylene, polypropylene or polytetrafluoroethylene (PTFE). Rinsing of any surfaces not constructed from these materials indicated negligible effects on ion concentrations. Dry contamination on the collector funnel was minimised by rinsing with distilled water, as close as possible to the start of rainfall. In most cases this was considerably less than 3 h. On a few occasions when rainfall began in the middle of the night the funnel may have been exposed to dry contamination for up to 12 h. However the back trajectories and the sub-event chemistry patterns for these events suggested that this contamination did not unduly influence the event-averaged data for these rainfalls. We have noted these events carefully but have not excluded them from the event-based analysis presented here. Contamination tests were conducted by analysing distilled water passed through the wet-deposition monitor. SO2~ concentrations in this water varied between 0 and 0.6 keq 4l~1 and NO~ was 3 undetectable [cf. unpolluted background levels for the Northern hemisphere of 3.27 and 2.86 keq l~1, respectively (Szepesi and Fekete, 1987)]. The tests demonstrated that, in addition to the rinsing of the collection funnel, regular cleaning of the system was necessary to ensure contamination free samples. Therefore all pipelines and chambers were cleaned with distilled water on a weekly basis. Contamination tests conducted during the field campaign indicated negligible levels of contamination (Beverland et al., 1996). 2.2. Sampling site Continuous rainfall chemistry measurements were made from June to November 1989 at the Meteorological Office Experimental Site at Beaufort Park, Bracknell, Berkshire in south-east England (51 24N 0 45W). With the exception of the high density of road traffic throughout Berkshire there were no significant local sources of atmospheric pollution at this site on the western outskirts of a new town area. Previous studies within U.K. cities suggest that the precipitation composition would not have been significantly influenced by the suburban nature of the sampling location (RGAR, 1987). 2.3. Sampling strategy and laboratory analysis A total of 67 rainfall events were sampled during the field experiment. An event was defined as a minimum rainfall amount of 0.5 mm. A period of not less than 2 h between 0.5 mm rain increments separated events.
Precipitation samples were collected from the field site as soon as possible after the cessation of rainfall and filtered through 0.4 km Teflon filters. Sixty events were analysed for the anions: chloride ([Cl~]), nitrate ([NO~]) and sulphate 3 (sodium [Na], ([SO2~]) using ion chromatography. Cation 4 magnesium [Mg], and calcium [Ca]) measurements were made for some of these events using an atomic absorption spectrometer. Analytical resources were not sufficient to determine all of the ionic compounds necessary for ion balance quality assurance calculations (e.g. Peden, 1983). However laboratory tests using prepared standards and synthetic rainwater samples indicated that the overall uncertainties (i.e. both precision and accuracy) in our ion chromatographic determinations were less than 10% of typical concentrations. It is unlikely that these uncertainties would have influenced the conclusions of the study. Rainfall-weighted average values were calculated as follows: n xx" +
RA i x i RA 505!i/0
(1)
where xx"volume weighted average (kS cm~1 or keq l~1); n"number of events; RA "the rainfall amount for the ith i event (mm); RA "the total amount of rainfall deposited 505!(mm); x "the conductivity or concentration for the ith i event (kS cm~1 or keq l~1)
3. RESULTS
3.1. Summary of wet deposition statistics The rainfall-weighted average of hydrogen-ion concentration ([H`]) was 37 keq l~1 (Table 1), which was close to the annual average calculated from the U.K. national acid deposition monitoring network data (RGAR, 1987). Due to the proximity of major European sources, the rainfall-weighted pH of 4.4 was significantly lower than ‘‘background’’ or ‘‘unpolluted’’ pH values which generally range between 5.0 and 5.6 (Galloway et al., 1982). The rainfall-weighted averages of [Cl~], [NO~] and [SO2~] concentra3 4 tions were similar to the corresponding averages calculated for the national network data. With the exception of [H`], the rainfall-weighted averages were lower than the corresponding arithmetic averages. This was due to low concentrations towards the end of longer events (Beverland and Crowther, 1992). The opposite trend for [H`] was probably the result of low concentrations of neutralising species in the later stages of events. The coefficients of variation for the event concentration data approached or exceeded 100%, ranging from 92% for [Cl~] to 148% for [NO~] 3 (Table 1). These large values reflect considerable interevent variability. Frequency histograms of the concentration data showed strong positive skewness (skewness statistic ranged between 2.0 and 2.4) suggesting that the underlying distributions were lognormal. The use of histograms to assess closeness to normality is subjective and dependent on the class interval chosen. To make objective assessments of normality the Kolmogorov—Smirnov (K—S) Test was used (Ebdon, 1981). K—S tests on the data indicated that
46 48 102 72 101 181 87 66 Conductivity H` Cl~ NO~ 3 SO2~ 4 Na Mg Ca
25 kS cm~1 37 keq l~1 69 keq l~1 38 keq l~1 63 keq l~1 232 keq l~1 40 keq l~1 59 keq l~1 66 65 60 60 59 26 25 26
Measurement
35 kS cm~1 37 keq l~1 111 keq l~1 49 keq l~1 86 keq l~1 191 keq l~1 55 keq l~1 92 keq l~1
No. of events
Note: Rainfall weighted mean concentrations and wet deposition for southeast England (1981—1985) are given for comparison. (Data obtained from RGAR, 1987).The Beaufort Park SO2~ data 4 was not corrected for marine contribution (because of the low number of [Na] determinations).
— 0.02—0.04 g H m~2 yr~1 — 0.3—0.4 g N m~2 yr~1 0.5—1.0 g S m~2 yr~1 — — — — 0.01 g H m~2 yr~1 — 0.2 g N m~2 yr~1 0.4 g S m~2 yr~1 — — —
— 40 keq l~1 75 keq l~1 31 keq l~ 64 keq l~ — — — — 198 174 281 198 198 167 101
Deposition
131 132 92 148 118 95 159 72
Coefficient of variation for deposition (%) Rainfallweighted mean concentration Arithmetic mean concentration
Standard deviation (s.d)
Coefficient of variation for concentration (arithmetic mean /s.d.) (%)
Table 1. Statistical summary for the event data set
Concentration from U.K. Network
Deposition from U.K. Network
Chemical composition of rainfall in south-east England
1041
the natural logarithm of concentrations gave sample distributions which were significantly closer to normality compared with the un-transformed variables (logarithmic transformation increased K—S a from 0.00—0.02 to 0.20). Similar observations have been made in statistical analyses of other rainfall chemistry data sets (Lindberg, 1982; Singh et al. 1987). Subsequent statistical analyses were therefore carried out on the log transformed concentration data. The total wet depositions for the sampling period were calculated from the product of total rainfall amount and the rainfall-weighted average concentrations (Table 1). H`, NO~ and SO2~ depositions were 3 4 lower than the national network estimates, reflecting the exceptionally dry weather during much of the field experiment. The coefficients of variation for the event deposition data were all greater than 100% and exceeded those for concentration values. The greater inter-event variation results from the additional variability introduced by the rainfall amount, which outweighs the tendency for lower concentrations in higher volume rainfall events. Deposition data tended to be log-normally distributed (K—S a"0.00 for untransformed deposition data, K—S a"0.20 for logarithmic transformed deposition data). 3.2. Relationship between concentration/deposition and wind direction t-Tests were used to determine if significant differences existed between samples from different local, ground-level wind-direction sectors. The tests assumed that the sample distributions to be tested were normal therefore log transformed concentrations % were used (Section 3.1). Ninety-degree wind direction sectors were chosen. The average wind direction was calculated using procedures outlined by Mori (1986, 1987). Subdividing the data set meant that the number of data points in each wind direction sector was quite small which may have violated the assumption of normal sample distributions. To guard against this, the non-parametric Mann Whitney test was also used. The analyses tested the null hypothesis that the mean or median of the log concentrations, c, occurring % when the average wind direction was from sector i was less than, or equal to, the mean/median when the averaged wind direction was from sector j (H : c )c , 0 i j iOj) compared with the alternative, that the mean/median concentration was greater (H : c 'c ). 1 i j Critical values at the 90, 95, 99 and 99.9% confidence levels were used. For each ionic compound, highest concentrations occurred when the wind had a north-easterly component. The concentration differences with respect to the other sectors were significant at 99% and 95% confidence levels for most compounds for the parametric and non-parametric tests, respectively. Chloride was the only exception and showed no significant directional dependence. Sperber (1987) suggests that the average local wind direction is a good indicator of the source regions of air which fuel the storm systems
1042
I. J. BEVERLAND et al.
and contribute to pollutant loading. The initial examinations of the data seem to support this claim. The higher concentrations observed when local wind direction was in the northeast sector may be associated with increased urbanisation in the east from Bracknell to the London conurbation. The event-averaged concentration data were plotted against the average wind direction calculated for the rainfall period. It was difficult to interpret underlying trends from these scatter plots because the number of events changed with the level of the average wind direction. Cleveland and Kleiner (1975) introduced a method, analogous to a moving average, to deal with this scatter plot perception problem. They made use of an existing statistic, the midmean, and introduced two new statistics: the upper and lower semi-midmeans. The midmean is defined as the average of all observations between and including the quartiles. The upper and lower semi-midmeans are the midmeans of all observations above and below the median respectively. Exclusion of the upper and lower quartiles in the calculation of the midmean removes extreme values which may distort the underlying trend. While the plot of the rainfall-weighted mean concentrations against wind direction sector is representative of average conditions, the use of moving statistics gives additional information about the behaviour of the upper and lower tails of the concentration distribution (Cleveland and Kleiner, 1975; Sperber, 1987). For each compound (except [Cl~]) concentrations were greater when accompanying winds were from the northeastern quadrant (Figs 1A—D). For [H`] and [NO~] the lower semi-midmean, the midmean and 3 the upper semi-midmean showed similar patterns when plotted against mean wind direction within the four sectors. The sector mean for [Cl~] appeared to be almost constant. However, this masked some directional dependency which was evident in the moving statistics. Both the midmean, and the upper semi-midmean, showed minimum [Cl~] for the southeasterly sector because of the absence of significant marine sources in this quadrant. The [Cl~] upper semi-midmean showed more variability compared to the other statistics. Peak [Cl~] in the northeasterly quadrants may have resulted from anthropogenic input of HCl to specific events. Similarly, the secondary peak in the northwestern sector was probably caused by high marine [Cl~] input. The moving statistics for [SO2~] showed highest concentrations in the 0—90° 4 sector. However, the upper semi-midmean, and to a lesser extent the mean, were increased in the 90—180° sector by a small number of high concentration episodes. A clear pattern emerged from the moving statistics for the deposition data with the largest depositions in the 0—90° sector (deposition data are not shown graphically). This resulted from association of high concentrations in this sector with moderate or high rainfall amounts.
3.3. Inter-species and inter-sector concentration variability Sperber (1987) found the variability of [NO~] to be 3 significantly greater than [SO2~], in all wind direc4 tion sectors, from 6 yr of hourly precipitation data obtained at Long Island, New York, and suggested that SO and SO2~ aerosol are more homogeneously 2 4 mixed in the atmosphere than NO~. Examination of 3 the data set from Beaufort Park indicated general agreement with Sperber’s findings although the increased variability of NO~ was only statistically sig3 nificant (F-test at 95% confidence level) for the northwest wind direction sector and the unclassified data set. Sperber (1987) found variability in ionic concentrations to be less for the sector which represented the major axis of pollutant input. F-tests carried out on the Beaufort Park data did not support this theory. 3.4. ¹he relationship between concentration, deposition and atmospheric transport pattern Back trajectories were calculated, at the 950 mb level, for each rain event using the model described by Maryon and Heasman (1988). Trajectory start times were selected to be as near as possible to the middle of rain events. The trajectories were classified into 6 different transport patterns (Figs 2A—F). Differences in mean concentration and deposition values, for each transport pattern, were examined by using both parametric and non-parametric tests (as described in Section 3.2). There were clear differences between the transport patterns. [NO~] and [SO2~] concentra3 4 tions were greatest in the North Sea and Eastern European transport regimes (Figs 2A and B, respectively) with lowest concentrations in the mixed maritime/continental and southwesterly marine regimes (Figs 2E and F, respectively). A similar pattern was noted for the deposition values. Using the transport pattern classification, significant differences were noted over a larger number of independent variable classes, compared with the wind direction sector classification (e.g. for NO~ significant 3 differences were observed between 5 of the transport patterns but only between 2 of the wind direction sectors). This emphasises the superiority of back trajectory analysis for interpretative purposes. Contingency table analysis showed that the events from the 0—90° wind direction sector were comprised entirely of transport patterns A and B. However the situation was much less clear for the other sectors, and for each individual transport pattern. Therefore, in contrast to our initial findings in Section 3.2, we conclude that local wind direction data, averaged for the duration of an event, was not a particularly good indicator of the Lagrangian history of the air masses supplying pollutants to the precipitation system. 3.5. ¹he significance of rainfall amount In accordance with several other studies (Khwaja and Husain, 1990; Lindberg, 1982; Davies, 1976; Huff
Fig. 1. Moving statistics for event-averaged concentrations. Data cover period from June to November 1989. Statistics are based on wind direction average for the periods during rainfall (then classified and averaged within 90° sectors).
Chemical composition of rainfall in south-east England 1043
1044
I. J. BEVERLAND et al.
Fig. 2. Classified atmospheric transport patterns.
and Stout, 1964; Bleeker et al., 1966) we observed a marked decrease in event-average ionic concentrations as precipitation amount increased. We examined the correlations between: (a) event average concentration against log (rainfall amount); and (b) % log (concentrations) against log (rainfall amount). % % The correlation coefficients were higher for the powerlaw relationship [regression (b)], however in all cases the coefficients were not statistically significant (P(0.05), and we concluded that precipitation volume alone is not useful in explaining event-averaged concentrations. Simplified washout theory suggests exponential decreases of pollutant concentrations in air, and therefore also in rain, with accumulated rainfall amount during the course of individual events. Dawson (1978) found significant negative correlations between the logarithm of the concentration of ionic species and the accumulated rainfall amount. Similar analyses of the sub-event Beaufort Park data showed virtually
no correlation. It might be concluded that the earlier observations of Dawson were fortuitous and sitespecific. Therefore the simplified picture of washout is inadequate on event time scales when advective processes must also be considered (Beverland and Crowther, 1992). Previous studies (Hicks and Shannon, 1979; MacCracken, 1979; Lindberg, 1982; Khwaja and Husain, 1990) have found a power-law relationship between wet deposition levels (D) and rainfall amount (h): D"K1 hK2
(2)
where K1 and K2 are constants. Linear regression between the log transformed variables was used to % test this power-law hypothesis. There was considerable scatter in the relationships derived for the entire data set (Fig. 3). However the regression models were all significant (P(0.001) suggesting the presence of underlying functional relationships (Table 2). One of the main reasons for the
Chemical composition of rainfall in south-east England
1045
Fig. 3. Relationships between deposition and total rainfall amount. (A) Relationship between deposited H` and rainfall amount. (B) Relationship between deposited NO~, and rainfall amount. (Regression lines 3 shown are calculated from data from all events.)
Table 2. Regression analysis of wet deposition rainfall amount Data set Complete data set (67 events) SW marine transport pattern (42 events) Eastern European transport pattern (5 events)
Regression equation log % log % log % log % log % log % log % log % log % log % log % log %
R2 (%)
n
Standard error of slope
47 50 37 44 63 58 63 59 91 93 95 98
65 60 60 59 41 39 39 39 5 5 5 5
0.15 0.08 0.14 0.12 0.15 0.10 0.11 0.11 0.16 0.10 0.13 0.07
(H`) "2.80#1.14 log (rainfall amount) % (Cl~) "4.53#0.61 log (rainfall amount) % (NO~) "3.17#0.83 log (rainfall amount) 3 % (SO2~) "4.01#0.79 log (rainfall amount) 4 % (H`) "2.36#1.19 log (rainfall amount) % (Cl~) "4.44#0.69 log (rainfall amount) % (NO~) "2.54#0.87 log (rainfall amount) 3 % (SO2~) "3.58#0.78 log (rainfall amount) 4 % (H`) "4.98#0.84 log (rainfall amount) % (Cl~) "4.29#0.62 log (rainfall amount) % (NO~) "5.03#0.92 log (rainfall amount) 3 % (SO2~) "5.57#0.84 log (rainfall amount) 4 %
Note: The regression models were all statistically significant (slopeO0, P(0.001).
scattering may have been the influence of the prevailing transport pattern on the rainwater ionic concentration. Regression analyses were conducted on data from individual transport patterns. We restricted these analyses to the larger data sets in the southwesterly marine transport and the eastern European transport patterns (Figs 2F and 2B, respectively). Considerably higher R2 values were observed, demonstrating that much of the scatter in Fig. 3 was due to transport pattern effects (Table 2). The power-law exponents calculated for [NO~] 3 and [SO2~] were not significantly different for dif4 ferent transport patterns, suggesting that physical scavenging processes were responsible for the underlying functional relationship (Table 2). In contrast, the intercept [or log (K1)] values were significantly % different emphasising the importance of transport pattern in modelling studies utilising these types of
relationships. Significant differences were noted in the exponents calculated for different compounds. Hicks and Shannon (1979) found power-law exponents of 0.5 and 0.6 for the deposition of radioactive isotopes from nuclear bomb testing and sulphur deposition respectively. Lindberg (1982) calculated exponents of 0.72$0.08 for SO2~ and 0.86$0.08 for 4 H`. Lindberg reported that the sulphate value of 0.72 was not statistically different from the 0.6 exponent calculated by Hicks and Shannon, however there appeared to be evidence of a larger exponent for H`. Khwaja and Husain (1990) found exponents of 0.71, 0.55 and 0.72 for SO2~, NO~ and Cl~ respectively. 4 3 The exponents calculated from the Beaufort Park data were higher than the findings of the studies mentioned above (Table 2). The observed differences were statistically significant (P(0.005) for the data set as a whole and for most of the individual transport patterns.
1046
I. J. BEVERLAND et al.
The present study emphasises that transport pattern, site and species dependence must be considered if power-law relationships are used in acid-deposition models. 3.5. Proportional anion composition of event rainfall data In accordance with our findings in the previous section Davies (1988) suggested that much of the variation in precipitation composition is caused by variations in the overall concentration due to dilution effects. Influences other than dilution can be examined using a triangular diagram based on the primary components. Figure 4 shows the relative proportions of the anions NO~, SO2~ and Cl~. For the Beaufort 3 4 Park data there was a clear distinction between events with continental and marine transport patterns (Cl~ below 20% of total anions and Cl~ above 30%, respectively). The highly polluted events, with trajectories looping out over the North Sea (Fig. 2A), exhibited intermediate proportions.
The marine origin events appeared to lie along a line of increasing proportion of Cl~. The events with the highest proportion of Cl~ were the autumnal storms associated with gale force winds. The high wind speeds during air-mass passage over the oceans would have enhanced the formation and atmospheric mixing of chloride rich aerosol and rapid passage over essentially rural parts of England would have provided little opportunity for entrainment of anthropogenic nitrogen and sulphur species. Davies (1988) and Skartveit (1982) found that the [NO~] : [SO2~] ratio in precipitation decreased with 3 4 increasing travel time of polluted air. This was attributed to faster oxidation of nitrogen oxides compared to sulphur dioxide, resulting in nitric acid becoming available for removal before sulphuric acid (Rodhe et al., 1981). The estimated travel times from source regions during some of the continental origin events were noted for trajectories which were judged to have passed over major sources (i.e. the 6 labelled events in Fig. 4). In agreement with the earlier studies,
Fig. 4. Relative proportions of the anions NO~, SO2~ and Cl~ in rainfall (event-averaged). NO~ 3 4 3 proportions are read from the left-hand side edge using the lines which slope downward to the right. Cl~ proportions are read from the bottom edge using the lines which slope upwards to the right and SO2~ 4 proportions are read from the right-hand side edge using the horizontal lines (Arrows next to NO~, SO2~ 3 4 and Cl~ labels indicate how to read the proportion of these anions). Transport times (in days) are plotted to the immediate right of 6 events. These events had trajectories which passed over major source regions.
Chemical composition of rainfall in south-east England
lower [NO~] : [SO2~] ratios were associated with 3 4 increased travel time from anthropogenic source regions. In particular the two events with high ('40%) NO~ proportions exhibited slow passage over south3 east England, within 12 h of rainfall. The high density of vehicular traffic in the home counties may have caused the elevated proportion of NO~. 3
4. CONCLUSIONS
f The rainfall-weighted average concentrations for the 6 month field experiment at Beaufort Park were close to those estimated from the U.K. national acid deposition network. The deposition levels were somewhat below those suggested by network data because of the unusually dry summer of 1989. f The event temporal resolution attainable by the acid-rain monitor enabled a detailed study of the relationships between wet deposition, wind direction and atmospheric transport patterns. The influence of local wind direction during rainfall was significant in terms of observed concentrations and deposition. However back trajectory analysis was a better indicator of the Lagrangian history and pollutant loading of the air masses reaching the site. Pronounced differences were observed between transport patterns, with significantly enhanced deposition in the North Sea and Eastern European categories. The use of moving statistics was useful for interpretation of variations of acid deposition with changes in wind direction. f Concentration of pollutant species tended to be inversely related to precipitation amount. However other factors were also significant, and precipitation amount alone cannot be used to explain event average concentrations. Virtually no correlation was observed between log-transformed sub-event concentrations and rainfall amount, emphasising that simplified scavenging theory fails to account for the important influence of advection. f A power-law relationship existed between wet deposition and rainfall amount. The data exhibited considerable scatter around the functional relationship. Much of this scatter was attributed to the influence of transport pattern. The power-law exponents calculated were significantly greater than those reported in similar studies in North America. Significant differences were noted in the exponents calculated for different ionic compounds. f A triangular diagram of the principal anionic compounds provided a useful graphical interpretation of the influence of transport pattern on rainwater composition. Enhanced [Cl~] proportions were evident for maritime events. The [NO~] : [SO2~] 3 4 ratio was inversely related to the time of travel from major anthropogenic source regions. Acknowledgements—The authors wish to thank Dr F. B. Smith of the U.K. Meteorological Office and Miss G. Taylor
1047
of the University of Edinburgh for helpful comments. We are grateful for technical support from Messrs J. A. Broadfoot, M. A. Wallace, J. Revie, and R. W. Weston of the University of Strathclyde; and for assistance from Mr D. H. O¨Ne´ill and Ms J. P. Brown of the University of Edinburgh in the preparation of this manuscript. Financial and technical support from the U.K. Natural Environment Research Council and the U.K. Meteorological Office are acknowledged.
REFERENCES
Beverland, I. J. and Crowther, J. M. (1992) On the interpretation of event and sub-event rainfall chemistry. Environmental Pollution 75, 163—174. Beverland, I. J., Crowther, J. M. and Srinivas, M. S. N. (1996) Development and operation of a microprocessor-based acid rain monitor. Atmospheric Environment 30, 3611—3622. Bleeker, W., Dansgaard, W. and Lablans, W. (1966) Some remarks on simultaneous measurements of particulate contaminants including radioactivity and isotopic composition of precipitation. ¹ellus 18, 773—785. Cleveland, W. S. and Kleiner, B. (1975) A graphical technique for enhancing scatterplots with moving statistics. ¹echnometrics 17, 447—453. Dana, M. T. and Slinn, W. G. N. (1988) Acidic deposition distribution and episode statistics from the MAP3S network database. Atmospheric Environment 22, 1469—1474. Davies, T.D. (1976) Precipitation scavenging of sulphur dioxide in an industrial area. Atmospheric Environment 10, 879—890. Davies, T. D. (1988) Chemical composition of snow in the remote Scottish Highlands. In: Acid Deposition of High Elevation Sites, eds. M. H. Unsworth and D. Fowler, pp. 517—539. Kluwer Academic Publishers, Dordrecht, The Netherlands. Dawson, G. A. (1978) Ionic composition of rain during sixteen convective showers. Atmospheric Environment 12, 1991—1999. Ebdon, D. (1981) Statistics in Geography: A Practical Approach. Basil Blackwell, Oxford. Fowler, D. and Cape, J. N. (1984a) The contamination of rain samples by dry deposition on rain collectors. Atmospheric Environment 18, 183—189. Fowler, D. and Cape, J. N. (1984b) On the episodic nature of wet deposited sulphate and acidity. Atmospheric Environment 18, 1859—1866. Galloway, J. N., Likens, G. E., Keene, W. C. and Miller, J. M. (1982) The composition of precipitation in remote areas of the World. Journal of Geophysical Research 87, 8771—8786. Hicks, B. B. and Shannon, J. D. (1979) A method for modelling the deposition of sulphur by precipitation over regional scales. Journal Applied Meteorology 18, 1415—1420. Huff, F.A. and Stout, G. E. (1964) Distribution of radioactive rainout in convective rainfall. Journal of Applied Meteorology 3, 707—717. Khwaja, H. A. and Husain, L. (1990) Chemical characterisation of acid precipitation in Albany, New York. Atmospheric Environment 24A, 1869—1882. Koch, W. F. (1986) Standard reference materials: methods and procedures used at the National Bureau of Standards to prepare, analyse and certify SRM2694, simulated rainwater, and recommendations for use, U.S. Department of Commerce/National Bureau of Standards, Washington D.C. U.S.A. Koch, W. F., Marinenko, G. and Stolz, J. W. (1982) Simulated Reference Materials, Vol. IV, National Measurements Laboratory, Inorganic analytical research division, Washington D.C., U.S.A.
1048
I. J. BEVERLAND et al.
Laurila, T. and Joffre, S. M. (1987) Meteorological analysis of wet deposition in Finland. In Acid Rain: Scientific and ¹echnical Advances, ed. R. Perry, pp. 167—174. Selper Publishing, London. Lindberg, S. E. (1982) Factors influencing trace metal, sulphate and hydrogen ion concentrations in rain. Atmospheric Environment 16, 1701—1709. MacCracken, M. C. (1979) MAP3S progress report for F½1977 and F½1978, DOE/EV-0040, US Department of Energy, Washington D.C., U.S.A. Maryon, R. H. and Heasman, C. C. (1988) The accuracy of plume trajectories forecast using the U.K. Meteorological Office operational forecasting models and their sensitivity to calculation schemes. Atmospheric Environment 22, 259—272. Mori, Y. (1986) Evaluation of several ‘single-pass’ estimators of the mean and the standard deviation of wind direction. Journal of Climate and Applied Meteorology 25, 1387—1397. Mori, Y. (1987) Methods for estimating the mean and the standard deviation of wind direction. Journal of Climate and Applied Meteorology 26, 1282—1284. Peden, M. E. (1983) Sampling, analytical and quality assurance protocols for the national atmospheric deposition program. In Sampling and Analysis of Rain, ed. S. A. Campbell, AS¹M S¹P 823, pp. 72—83.
RGAR (1987) Acid deposition in the U.K. 1981—1985. Second report of the U.K. Review Group on Acid Rain, Warren Spring Laboratory, Stevenage, U.K. Rodhe, H., Crutzen, P. and Vanderpol, A. (1981) Formation of sulphuric and nitric acids in the atmosphere during long range transport. ¹ellus 33, 132—141. Singh, B., Nobert, M. and Zwack, P. (1987) Rainfall acidity as related to meteorological parameters in Northern Quebec. Atmospheric Environment 21, 825—842. Skartveit, A. (1982) Wet scavenging of sea-salts and acid compounds in a rainy, coastal area. Atmospheric Environment 16, 2715—2724. Smith, F. B. (1983) Conditions pertaining to high acid concentrations in rain. In Proceedings of ».D.I. Conference on: Acid precipitation — origin and effects, Lindau, Germany. Smith, F. B. and Hunt, R. D. (1978) Meteorological aspects of the transport of pollution over long distances. Atmospheric Environment 12, 461—467. Sperber, K. R. (1987) The concentration and deposition of nitrate, sulphate and ammonium as a function of wind direction from precipitation samples. Atmospheric Environment 21, 2629—2641. Szepesi, D. J. and Fekete, K. E. (1987) Background levels of air and precipitation quality for Europe. Atmospheric Environment 21, 1623—1630.