- Email: [email protected]
Estimating background precipitation quality from network data
Environmental Pollution 75 (1992) 137 143
Estimating background precipitation quality from network data B. B. Hicks & R. S. Artz National Oceanic and Atmospheric Administration, Air Resources Laboratory, 1325 East West Highway, Silver Spring, Maryland 20910, USA
Assessments of the relative merits of alternative acid-rain control strategies revolve around considerations of potential benefit per unit effort and/or cost. A question that often arises concerns the changes in deposition that would follow if all industrial (or societal) emissions were eliminated, in which case precipitation chemistry would be dominated by emissions from natural sources. Estimates of the 'natural background' of precipitation chemistry can be based on (a) measurements made at distant locations, (b) reducing emissions to zero in numerical simulations, or (c) examinations of existing data. Each alternative is flawed because (a) of the assumption that natural emissions in one location are like those in another, (b) no existing model contains descriptions of chemical processes involving all of the chemical species of importance, and (c) all contemporary data records of relevance are affected by precisely the emissions we wish to reduce. Here, the third alternative is explored in detail, using event precipitation chemistry data from North America. The analysis reveals a background pH level that varies from site to site, but always lies in the range 5.0-5.3.
INTRODUCTION
In general, Lagrangian models lack the capacity to consider chemical reactions in the detail necessary for conditions known to exist over the eastern portion of North America, and hence a new generation of Eulerian models was developed (e.g. Carmichael & Peters, 1984a,b; Chang et al., 1987). These new models are fundamentally different from the assessment models of earlier generations, in that they are designed to be episodic, whereas the dispersion routines that predated them necessarily addressed either ensemble averages or long-term average conditions. All of these models are designed to answer questions about how existing deposition rates will change, if changes are made in emissions. An obvious question then arises about how the reduced deposition would compare to the levels that would result, if all emissions were reduced to zero. In other words, to evaluate the effectiveness of feasible control strategies, it is necessary to consider the limit that is potentially achievable by such strategies and the rate at which alternative control measures approach this limit. First, such limits must be identified. There are three main methods that are appropriate:
Policy decisions regarding the imposition of controls on industrial emissions are hindered by questions of cost and benefit. In general, many of the related concerns revolve around a common focal theme, concerning the question of the effect at sensitive receptors of alternative control strategies imposed on sources far upwind. The US National Acid Precipitation Assessment Program was set up to provide a scientific basis for answering questions of this kind, and has succeeded in developing a number of tools for conducting the assessments that are required. The principal techniques are numerical models, which are required in this particular instance because the interaction among different chemical species and the interplay with a wide variety of physical and biological processes casts doubt on the validity of any simple judgement. In the 1970s; a variety of transport and dispersion models were developed (e.g. short-term Lagrangian models such as that of Eliassen (1978) and longer-term statistical trajectory models such as that of Sheih (1977)) with the intention of guiding the decisionmaking process. At that time the debate was centering on the problem o f sulfur dioxide emissions and the products of their chemical reactions in the atmosphere.
1. The models developed to relate deposition to emissions could be run with industrial emissions, etc. (as appropriate), set to zero. In practice, this approach is flawed because contemporary models do not include the chemicals of natural origin that are known to
Environ. Pollut. 0269-7491/92/$03.50 © 1991 Elsevier Science
Publishers Ltd, England. Printed in Great Britain 137
B. B. Hicks, R. S. Artz
138
the locations of interest, and see whether these distributions reveal indications of lower limits that cannot be exceeded.
dominate deposition in remote areas, as will become evident later. 2. Measurements could be made in remote areas, unaffected by local pollution sources, and the results taken to be representative o f the target areas that are of interest. The problem with this approach is the need to assume that observations at some distant location are representative o f local conditions in some hypothetical set o f circumstances, without the ability to test the assumption. 3. Study the distributions of deposition measured at
Here, the third o f these approaches will be used to assess the likely limits on precipitation quality across the portion of the eastern U S A where an extensive record of appropriate data is available. The intention is not to provide an exhaustive examination of all the data that are available, but rather to demonstrate a possible way to address the problem. 3
504 ° o°
o
~ o
O~o~~~
®
0
o 0
I
0 0 0 o
Oo
°
'
.~
05
ooo$ °
o~
0
o
~P
o
o
8~o5
008
o
0000
o
0° 8 o
o
oo
o
cB
o o ~ ~ o 03
::L
I
~__
o oo
0 0
o
o
oo
o
~
~
5
~
°o oo ~ 7 ~ i ~ _ o ~ o
0
o
o
o
~
o
° o
o °°°
o
0
o o -05
T
I
I
[
l
38
34
I
I
~
I
I
46
I
,
I
-1
-i
I
I
5~
I
]4
I
I
5o~oo~'°
l
38
I
l
42
[
o
I
4.6
o
o
oo
o
[
I
~
I
I
51
58
K+
NO,"
o oo
l
5
-05
o
2
o
o
o o ::L
I
o
,
0
0%%
o
0
o~
o
oO~
°°°
o o ~o
o
o
~I0°
/
oo o
Q oo j9
~oo
od%
00
o
Oo
oo
o
o
08
Q
o
o
~
o
o o
5
$4
o
-05
o
-05
o
o ooo o oo
o I
~
, 4
]
3.8
[
I
I
4.2
I
I
I
4.6
I
5
I
-I
]
5.4
I
.38
34
58
CI
25
o oo
42
4~
58
Na ÷
2~ J
2 -
o "~ ~L
I
o
cD 000 0
o°°°
15
15-~
o~
o ° ot~a~ooo=o ~ : ~ o~3Oo
,o
I
°
' o
~
~
o
°oo° %°
°o%
D
1 4
_°°~i°
o
o
o
0
~
o
o
05 -
0.5 ° 0
0
-o ° LLJO0
0
o
ooo 0
o
°
-o5
-05
-
o
o
oo
o
o -i
-1 34
) 8
• ~
48
pH
~
:)4
~8
I
34
38
I
I
i
42
I
4G
I
I
~
~4
5B
pH
Fig. 1. The relationships between concentrations in event precipitation and pH for the Whiteface Mountain (NY) station of the MAP3S network.
Estimating background precipitation quality from network data SOURCES OF DATA The present intent is to derive information on likely background pH levels by examining the statistical distributions of concentrations of different chemical species in precipitation samples. For this purpose, it is necessary to consider wet deposition data only. Open-bucket collections of atmospheric deposition are not suitable; these 'bulk' data are influenced by dry depositing particles and the products of gas exchange between the air and all exposed surfaces, so that the linkage between different chemical species is artificially complicated. There are two kinds of wet-only network in operation in North America at this time (Hales et al., 1987). One is a routine operation employing weekly sampling, the other is a research network, using 'event' collection (often actually daily collection). The databases generated by these two kinds of operation compare well, although some differences arise in interpreting the results. The two networks serve two different purposes--the weekly sampling protocol permits economical operation of a larger number of stations, as needed to detect trends with time and spatial irregularities, whereas the event data are needed to explore the reasons for the precipitation chemistry that is observed. The present analysis will be of 'event' data only; these provide the best information on the chemical characteristics of precipitation as it falls, and hence provide an opportunity to relate precipitation quality to meteorological conditions. In both regards, weekly data are deficient.
139
US Department of Energy). These data, and all of the North American information used in the present analysis, were derived from the deposition database assembled under the auspices of the US National Acid Precipitation Assessment Program (Watson & Olsen, 1984). Figure 1 presents plots of concentrations of different species in event precipitation versus reported pH (laboratory measurement). Note that in all such plots and analyses the concentrations are considered logarithmically, because concentrations display a nearly lognormal distribution, much like the precipitation itself. In general, the individual plots of Fig. 1 show the large scatter that is to be expected, with clear indications of strong correlations when concentrations are high. At low concentrations, however, the distribution becomes far more disordered. Inspection of the data reveals that much of the scatter is due to those few data points with the lightest rain. Correlation between concentrations of any substance in precipitation and the amount of rain that falls are well known; there is a significant dilution effect involved, which is aggravated by sample evaporation when very small quantities are collected. In the analysis that follows, all incomplete data sets have been excluded. In practice, this amounts to omission of all events with precipitation less than about 2 mm. The remaining data reveal a far reduced role of precipitation amount as a variable influencing precipitation quality.
ISO~-I VERSUS INO~I ESTIMATING 'BACKGROUND' pH
The linear portion of the [SO2-] versus pH plot shown If all industrial emissions of sulfur dioxide were suddenly stopped, what would be the consequence on rainfall pH? This question underlies all discussions of the potential benefits of emission control strategies. Without knowing what limiting situation is achievable, it is not possible to assess how close to this achievable limit alternative control strategies can bring us. The question just posed carries the tacit assumption that sulfur dioxide is the main culprit. As will be seen later, this is not necessarily the case. A parallel question can be posed, regarding emissions of nitrogen oxides. In practice, sulfate and nitrate in precipitation are highly correlated. Here, the influence of NO x emission reductions will be considered as a matter independent of the SO 2 control issue. The questions pertain to areas that are affected by pollution, by sulfur and nitrogen oxides emitted somewhere upwind. Wet deposition is the simplest measurement on which to base any conclusions that may be derived; it also has the largest database. Consider, for example, the data obtained at Whiteface Mountain, NY, as part of the MAP3S program (the Multistate Atmospheric Power Production Pollution Study, of the
in Fig. 1 illustrates a part of the overall behavior pattern, which can be discussed with some confidence--a reduction in sulfate concentrations in precipitation will have an accompanying change in pH associated with it. However, it should be noted that [NO~] could be interpreted in the same way. The reason is that [SO42-] and [NO~] are highly correlated in air, and even more so in precipitation. An interesting first step is, therefore, to explore the comparative roles of [SO42-] and [NO~] at different locations. To this end, a stepwise regression method has been used, as follows: • A simple linear regression identifies which of the concentration variables explains most of the variance in pH. • A multiple regression technique is then used repetitively to order the contributions of the other variables in further reducing the residual (so far unexplained) variance. The sequence is stopped when no remaining variable explains a statistically significant portion of the residual variance. Figure 2 shows how the results are ordered, on a map of the eastern USA. The results are tabulated to show
B B. Hicks, R. S. Artz
140
WHITEFACE S042-
-0.67
NO 3Ca 2÷
-0.41 +0.35
ITHACA S042-
-0.65
NO 3Ca 2+
-0.34 +0.25
BROOKHAVEN NO 3-
-0.82
S042-
-0.40
NH4 ÷
+0.31
STATE COLLEGE N O 3-
-0.77
S042-
-0.39
Ca 2+
+0.29
BONDVILLE
OXFORD
OAK RIDGE
BIRDWOOD
SO42-
-0.50
S042-
-0.63
S042-
-0.82
N O 3-
-0.78
LEWES NO 3-
-0.83
Ca 2+
+0.59
NH4 +
+0.34
Ca 2+
+0.20
SO# 2-
-0.43
S042-
-0.40
NH4 +
+0.48
N O 3-
-0.37
N O 3-
-0.33
Ca 2+
+0.33
Ca 2+
+0.33
Fig. 2. The distribution of stations considered in the present analysis, and the results of a stepwise correlation analysis in which the effects of chemicals that influence pH are eliminated sequentially, in the orders shown for each site.
the order in which site-specific data indicates that imposition of controls would most effectively enable a modification of local pH. The numbers shown are correlation coefficients. For Oak Ridge, TN, for example, the data indicate that the most efficient means to modify local precipitation is to control sulfur. Once the effect of sulfate on pH is eliminated, then the residual variability in pH is best reduced by controlling calcium, but in this case (even if such controls were feasible) the effect of reduction in Ca would be to increase pH. Controls on nitrogen oxides would appear to be a relatively low priority, in this particular case. Clearly, the situation in the mid-Atlantic region is different--there, nitrogen appears to be a more effective modifying agent than sulfur. It is clear that a focus on sulfur is indeed appropriate for most locations, but there is evidence that for some locations [SO4 2-] is not dominant. These stations are in the far east, downwind of major urban areas.
DISCUSSION Figure 1 presents concentration versus pH scatter plots for sulfate, nitrate, chloride, ammonium, sodium, cal-
cium, magnesium and potassium, respectively. The strength of the low-pH linear behavior decreases in the sequence of this listing. A useful interpretation of the overall behavior is as follows: 1. As shown in Fig. 2, the dominant chemical species affecting precipitation acidity is sulfate, but nitrate appears to be nearly as important. In a second category lie calcium and ammonium. There appears to be no relationship for any of the other species that are considered. 2. Consider the situation for sulfate alone. On some occasions, the air at Whiteface Mountain, for example, is largely unaffected by emissions from upwind sources, and on these occasions the precipitation will have characteristics that are indicative of background conditions at this location. In concept, accordingly, the data of Fig. l can be considered to represent two regimes, one which represents a background level in which sulfate is not at high enough levels to lower the pH, a n d the other in which sulfate concentrations are directly influencing pH. The first regime represents a background condition, the second represents the additional influence of sulfur emissions on this background level. The interaction between these two regimes identifies some measure of the level of sulfate in precipitation
Estimating background precipitation quafity from network data
141
Table 1. Values of 'background pH' derived from observations of event precipitation chemistry at sites of the multistate atmospheric power production pollution study (MAP3S) (for comparison, the result obtained using data for a remote site of the global precipitation chemistry network (see Table 2) is also included.)
1.0
0.8
06
"= o.4 .natocns / / ~
~ N O 3~ (SO4 ~ =
)
.~ >" o2
pH
Fig. 3. A plot of the square of the correlation coefficient yielded by a linear regression of pH on chemical concentration in precipitation over increasing pH ranges extending up to the value that is plotted. The curve for sulfate shows a maximum correlation for a pH of about 5-2, which corresponds to a situation in which further reduction in sulfate does not markedly increase pH. Also shown is the behavior for nitrate, computed after allowance is made for the influence of sulfate.
that we should not expect to drop below, even if all of the important upwind industrial emissions are eliminated. 3. A method is needed to derive an objective determination of a measure of 'background' from distributions like that illustrated in Fig. 1. Figure 3 presents the results of a sequential correlation analysis, designated to assess the amount of variance in pH that is directly attributable to sulfate concentration over ranges of sequentially increasing pH, starting with the most acidic data. The diagram shows that the amount of total variance in pH explained increases steadily until the pH is about 5.2. Subsequent values belong to a different r e g i m e - - t h e y detract from the overall correlation, not improve upon it. Hence, the value pH 5-2 can be interpreted as an objective evaluation o f the precipitation acidity that could be used as a target for emission control scenarios for this particular location (Brookhaven, Long Island, NY). 4. Now consider sulfate and nitrate, together with the other chemical species. Sulfate and nitrate are clearly correlated. In essence, dirty rain contains a lot of chemicals of all kinds. It is for this reason that ammonia displays a relationship with pH that appears to be contrary to the expectation based on the correct observation that ammonium is alkaline. 5. A variogram (such as is shown in Fig. 3) for nitrate is revealing, but cannot be easily interpreted because of the underlying relationship between sulfate and nitrate concentrations in rain. In short, nitrate concentrations cannot be assumed to be independent of sulfate concentrations. Consequently, the curve for nitrate that is drawn in Fig. 3 represents the residual variance accounted for by variations in nitrate, after the contributions directly due to sulfate are eliminated
Whiteface Mountain, NY Ithaca, NY State College, PA Charlottesville, VA Brookhaven, NY Lewes, DE Oxford, OH Bondville, IL
5.3 5-0 5.0 5.0 5.2 5.2 5.0 5.0
Katherine, Australia
5.0
from consideration. That is, the nitrate data are derived from a partial correlation analysis, conducted in the same stepwise manner as for sulfate, but computed to reveal the proportion of residual variance explained (instead of the proportion of the total variance). A similar interpretation of this second variogram is now feasible--that controls on emissions of NO x independently of sulfur may not necessarily cause a strong reduction in precipitation acidity, even though for the site now considered, NO~ is marginally more correlated with pH than is SO42- (see Fig. 2). Table I lists the conclusions drawn for all of the monitoring sites considered here--Whiteface Mountain (NY), Ithaca (NY), State College (PA), Birdwood (VA), Lewes (DE), Oxford (OH) and Brookhaven (NY). Also shown is the consequence of an identical analysis of event precipitation data collected at a remote site-Katherine, Australia. The pH value that is revealed is much the same as the value yielded for the USA stations.
C H A R A C T E R I S T I C S O F 'BACKGROUND' PRECIPITATION Alternative methods for revealing 'background' characteristics of precipitation at measurement sites have been tested, and yield similar results. At this time, there appears to be no reason to select any one of the alternative methods relative to any other. The examination of precipitation acidity given above permits identification of a value that is indicative of the 'background pH' that could theoretically be achieved in industrialized areas if stringent emission controls were imposed. Once identified, this separation point can then be used to judge the likely reduction in the concentrations of different chemical species that could be achieved if the 'ultimate and absolute' controls were imposed. This separation-point determination can be refined, however, by using information generated from
142
B. B. Hicks, R. S. Artz
the other methods. Vong (1990), for example, considers precipitation chemistry for stations upwind of major sources, and concludes that 'background' concentrations of sulfate in precipitation are likely to be in the range 2-16 ~ equiv litre -l, and the average background pH is probably about 5.3. A certain amount of natural variability in pH between stations is to be expected, even in the absence of industrial pollution, since pH is affected by many trace chemicals in addition to those on which controls are feasible. Among these contributing chemicals are carbon dioxide, volcanically produced sulfur compounds, nitrogen oxides (produced from lightning, as well as by biological reactions in plants and soils), seasalt compounds (containing e.g., sodium, chloride and sulfate) and many organic compounds. Whatever the cause, it is clear that there are large temporal and spatial variations in the chemistry of precipitation, even at remote locations, and that interpolating among existing wet deposition measurement sites is not a clear-cut exercise. In all areas of the world, rainfall contains trace quantities of chemicals scavenged from the atmosphere. Carbon dioxide is dissolved, for example, until the solution attains a pH of about 5-6 (a H + concentration of c. 2.5 p~ equiv litre-l). In marine areas, the dominant chemicals are the same as are found in high concentrations in sea water, such as sulfate. Data obtained in operations of the Global Precipitation Chemistry Network (Table 2) are particularly revealing. For Amsterdam island, if all measured sulfate were in the form of sulfuric acid, the volume-weighted mean pH would be no higher than c. 4.5. In all areas affected by thunderstorms, nitrate concentrations are partially due to lightning; in the northeastern USA it was estimated that the lightning-generating NO x was 24% as high as anthropogenic NO x during the hour with the greatest frequency of lightning, although this value was typically only 2.5% on average (Pierce et al., 1991); much of the remainder is due to emission from soils. (In contrast, as an average across the USA, about 45% of NO x emissions are due to transportation sources, and about 30% due to electric utilities; relatively little is from natural sources.)
In all areas within or downwind of vegetated areas (including many remote polar and marine areas) organic acids may account for up to c. 65% of free acidity (Keene et al., 1983) and even in the northeastern USA, organic acids may account for c. 16% of free acidity in MAP3S samples (Keene & Galloway, 1984). In addition, sulfur dioxide is exuded from surface fissures, and especially from volcanoes, leading to the presence of sulfate in precipitation. Other sulfur species, most notably hydrogen sulfide and dimethyl sulfide, are emitted by biological processes from wetlands and directly from the ocean. The products of atmospheric reactions involving all of these substances are potentially acidifying, and add to the natural background acidity of precipitation in remote areas. As a group, if organic acids are routinely measured as a part of the sampling protocol, the mean pH of precipitation at remote stations is rarely much higher than about 5-1, unless large amounts of neutralizing materials (soil) are present. In the context of the analysis presented in this paper, this translates to a natural background pH of c. 5.1 to 5.6 at MAP3S stations in the absence of organic acid measurements.
CONCLUSIONS AND R E C O M M E N D A T I O N S There is no technique that will give a precise and robust answer to the question of the consequences on precipitation acidity if stringent control measures are imposed. In the past, data on background precipitation chemistry obtained at remote locations selected to approximate specific target areas of special interest (such as the eastern USA) have been used as guidance. Here, an alternative path has been followed, making use of event precipitation chemistry data collected in the target area. The results obtained are quite similar to the conclusions drawn from examination of data from remote sites: depending on the location, the pH that could be achieved if all industrial sulfur emissions to air were curtailed would be in the range 5-0-5-3. Even in this case, there would be a remaining level of sulfate in the precipitation, amounting to about 5/~ equiv litre -1.
Table 2 a. Examples of differences in the composition of precipitation at different remote locations--a remote island (Amsterdam Island, Indian Ocean), a remote inland area (Katherine, Northern Territory, Australia) and a high elevation inland area (Lijiang, China); (concentrations are volume-weighted mean values, expressed in p equiv litre -I. Note that each of the values listed is an approximation to precipitation quality and that these data refer to different periods, with different sampling protocols. Close comparison is not warranted.)
Site
SO2-
NO 3
NH~f
HCOO-
CH3COO-
H÷
pH
Amsterdam Island Amsterdam Island (excluding sea salt) Katherine Lijiang
29.2 4.8 3.9 5.3
1.6 1.6 4.0 1.6
2.4 2.4 2.9 4.7
2.1 2.1 7.3 3.4
0.8 0.8 2.9 2.3
8.3 8.3 18.3 9.7
5.08 5.08 4.74 5.01
a Data from Galloway & Gaudry (1984) and Galloway, J. N. (1989, unpublished).
Estimating background precipitation quality from network data The present analysis confirms that sulfur is not always the optimal species to control, if the desire is to reduce precipitation acidity. In particular, data obtained in the USA indicate that controls on emissions of sulfur species may not be as effective as would be controls on N O x emissions, if such were possible, for a sizeable part of the mid-Atlantic coastal region.
ACKNOWLEDGEMENTS This work was carried out as a component of Task G r o u p II (Atmospheric Chemistry) of the National Acid Precipitation Assessment Program, under the sponsorship of the National Oceanic and Atmospheric Administration and the US Environmental Protection Agency.
REFERENCES Carmichael, G. R. & Peters, L. K. (1984a). An Eulerian transport/transformation/removal model for SO 2 and sulfate, I, model development. Atmos. Environ., 18, 937-51. Carmichael, G. R. & Peters, L. K. (1984b). An Eulerian transport/transformation/removal model for SO 2 and sulfate, II, model calculation of SO x transport in the eastern United States. Atmos. Environ., 18, 953-67. Chang, J. S., Brost, R. A., Isaksen, I. S. A., Middleton, P., Stockwell, W. R. & Walcek, C. J. 0987). A three-dimen-
143
sional Eulerian acid deposition model: physical concepts and formulation. J. Geophys. Res., 92, 14681-700. Eliassen, A. (1978). The OECD study of long range transport of air pollutants: long range transport modelling. Atmos. Environ., 12, 479-87. Galloway, J. N. & Gaudry, A. (1984). The composition of precipitation on Amsterdam Island, Indian Ocean. Atmos. Environ., 18, 2649 56. Hales, J. M., Hicks, B. B. & Miller, J. M. (1987). The role of research measurement networks as contributors to Fedral assessments of acid deposition. Bull. Am. Meteorol. Soc., 68, 216-25. Keene, W. C. & Galloway, J. N. (1984). Organic acidity in precipitation of North America. Atmos. Environ., 18, 249197. Keene, W. C., Galloway, J. N. & Holden, J. D. (1983). Measurement of weak organic acidity in precipitation from remote areas of the world. J. Geophys. Res., 88, 5122-30. Pierce, T. E., Coventry, D. H., Novak, J. H. & Van Meter, A. R. (1991). Estimating lightning-generated NO x emissions for regional air pollution models. Proceedings 7th AMSA WMA conference on Applications of Air Pollution Meteorology, New Orleans, 14-18 Jan. 1991, pp. 160-3. Sheih, C.-M. (1977). Application of statistical trajectory model to the simulation of sulfur pollution over the northeastern united States. Atmos. Environ., II, 173-178. Vong, R. J. (1990). Mid-latitude northern hemisphere background sulfate concentration in rainwater. Atmos. Environ., 2,4, 1007-18. Watson, C. R. & Olsen, A. R. (1984). Acid Deposition system (ADS) for Statistical Reporting: System Design and User's Code Manual. US Environmental Protection Agency Report EPA-600-8-84-023, Research Triangle Park, NC.
Estimating background precipitation quality from network data B. B. Hicks & R. S. Artz National Oceanic and Atmospheric Administration, Air Resources Laboratory, 1325 East West Highway, Silver Spring, Maryland 20910, USA
Assessments of the relative merits of alternative acid-rain control strategies revolve around considerations of potential benefit per unit effort and/or cost. A question that often arises concerns the changes in deposition that would follow if all industrial (or societal) emissions were eliminated, in which case precipitation chemistry would be dominated by emissions from natural sources. Estimates of the 'natural background' of precipitation chemistry can be based on (a) measurements made at distant locations, (b) reducing emissions to zero in numerical simulations, or (c) examinations of existing data. Each alternative is flawed because (a) of the assumption that natural emissions in one location are like those in another, (b) no existing model contains descriptions of chemical processes involving all of the chemical species of importance, and (c) all contemporary data records of relevance are affected by precisely the emissions we wish to reduce. Here, the third alternative is explored in detail, using event precipitation chemistry data from North America. The analysis reveals a background pH level that varies from site to site, but always lies in the range 5.0-5.3.
INTRODUCTION
In general, Lagrangian models lack the capacity to consider chemical reactions in the detail necessary for conditions known to exist over the eastern portion of North America, and hence a new generation of Eulerian models was developed (e.g. Carmichael & Peters, 1984a,b; Chang et al., 1987). These new models are fundamentally different from the assessment models of earlier generations, in that they are designed to be episodic, whereas the dispersion routines that predated them necessarily addressed either ensemble averages or long-term average conditions. All of these models are designed to answer questions about how existing deposition rates will change, if changes are made in emissions. An obvious question then arises about how the reduced deposition would compare to the levels that would result, if all emissions were reduced to zero. In other words, to evaluate the effectiveness of feasible control strategies, it is necessary to consider the limit that is potentially achievable by such strategies and the rate at which alternative control measures approach this limit. First, such limits must be identified. There are three main methods that are appropriate:
Policy decisions regarding the imposition of controls on industrial emissions are hindered by questions of cost and benefit. In general, many of the related concerns revolve around a common focal theme, concerning the question of the effect at sensitive receptors of alternative control strategies imposed on sources far upwind. The US National Acid Precipitation Assessment Program was set up to provide a scientific basis for answering questions of this kind, and has succeeded in developing a number of tools for conducting the assessments that are required. The principal techniques are numerical models, which are required in this particular instance because the interaction among different chemical species and the interplay with a wide variety of physical and biological processes casts doubt on the validity of any simple judgement. In the 1970s; a variety of transport and dispersion models were developed (e.g. short-term Lagrangian models such as that of Eliassen (1978) and longer-term statistical trajectory models such as that of Sheih (1977)) with the intention of guiding the decisionmaking process. At that time the debate was centering on the problem o f sulfur dioxide emissions and the products of their chemical reactions in the atmosphere.
1. The models developed to relate deposition to emissions could be run with industrial emissions, etc. (as appropriate), set to zero. In practice, this approach is flawed because contemporary models do not include the chemicals of natural origin that are known to
Environ. Pollut. 0269-7491/92/$03.50 © 1991 Elsevier Science
Publishers Ltd, England. Printed in Great Britain 137
B. B. Hicks, R. S. Artz
138
the locations of interest, and see whether these distributions reveal indications of lower limits that cannot be exceeded.
dominate deposition in remote areas, as will become evident later. 2. Measurements could be made in remote areas, unaffected by local pollution sources, and the results taken to be representative o f the target areas that are of interest. The problem with this approach is the need to assume that observations at some distant location are representative o f local conditions in some hypothetical set o f circumstances, without the ability to test the assumption. 3. Study the distributions of deposition measured at
Here, the third o f these approaches will be used to assess the likely limits on precipitation quality across the portion of the eastern U S A where an extensive record of appropriate data is available. The intention is not to provide an exhaustive examination of all the data that are available, but rather to demonstrate a possible way to address the problem. 3
504 ° o°
o
~ o
O~o~~~
®
0
o 0
I
0 0 0 o
Oo
°
'
.~
05
ooo$ °
o~
0
o
~P
o
o
8~o5
008
o
0000
o
0° 8 o
o
oo
o
cB
o o ~ ~ o 03
::L
I
~__
o oo
0 0
o
o
oo
o
~
~
5
~
°o oo ~ 7 ~ i ~ _ o ~ o
0
o
o
o
~
o
° o
o °°°
o
0
o o -05
T
I
I
[
l
38
34
I
I
~
I
I
46
I
,
I
-1
-i
I
I
5~
I
]4
I
I
5o~oo~'°
l
38
I
l
42
[
o
I
4.6
o
o
oo
o
[
I
~
I
I
51
58
K+
NO,"
o oo
l
5
-05
o
2
o
o
o o ::L
I
o
,
0
0%%
o
0
o~
o
oO~
°°°
o o ~o
o
o
~I0°
/
oo o
Q oo j9
~oo
od%
00
o
Oo
oo
o
o
08
Q
o
o
~
o
o o
5
$4
o
-05
o
-05
o
o ooo o oo
o I
~
, 4
]
3.8
[
I
I
4.2
I
I
I
4.6
I
5
I
-I
]
5.4
I
.38
34
58
CI
25
o oo
42
4~
58
Na ÷
2~ J
2 -
o "~ ~L
I
o
cD 000 0
o°°°
15
15-~
o~
o ° ot~a~ooo=o ~ : ~ o~3Oo
,o
I
°
' o
~
~
o
°oo° %°
°o%
D
1 4
_°°~i°
o
o
o
0
~
o
o
05 -
0.5 ° 0
0
-o ° LLJO0
0
o
ooo 0
o
°
-o5
-05
-
o
o
oo
o
o -i
-1 34
) 8
• ~
48
pH
~
:)4
~8
I
34
38
I
I
i
42
I
4G
I
I
~
~4
5B
pH
Fig. 1. The relationships between concentrations in event precipitation and pH for the Whiteface Mountain (NY) station of the MAP3S network.
Estimating background precipitation quality from network data SOURCES OF DATA The present intent is to derive information on likely background pH levels by examining the statistical distributions of concentrations of different chemical species in precipitation samples. For this purpose, it is necessary to consider wet deposition data only. Open-bucket collections of atmospheric deposition are not suitable; these 'bulk' data are influenced by dry depositing particles and the products of gas exchange between the air and all exposed surfaces, so that the linkage between different chemical species is artificially complicated. There are two kinds of wet-only network in operation in North America at this time (Hales et al., 1987). One is a routine operation employing weekly sampling, the other is a research network, using 'event' collection (often actually daily collection). The databases generated by these two kinds of operation compare well, although some differences arise in interpreting the results. The two networks serve two different purposes--the weekly sampling protocol permits economical operation of a larger number of stations, as needed to detect trends with time and spatial irregularities, whereas the event data are needed to explore the reasons for the precipitation chemistry that is observed. The present analysis will be of 'event' data only; these provide the best information on the chemical characteristics of precipitation as it falls, and hence provide an opportunity to relate precipitation quality to meteorological conditions. In both regards, weekly data are deficient.
139
US Department of Energy). These data, and all of the North American information used in the present analysis, were derived from the deposition database assembled under the auspices of the US National Acid Precipitation Assessment Program (Watson & Olsen, 1984). Figure 1 presents plots of concentrations of different species in event precipitation versus reported pH (laboratory measurement). Note that in all such plots and analyses the concentrations are considered logarithmically, because concentrations display a nearly lognormal distribution, much like the precipitation itself. In general, the individual plots of Fig. 1 show the large scatter that is to be expected, with clear indications of strong correlations when concentrations are high. At low concentrations, however, the distribution becomes far more disordered. Inspection of the data reveals that much of the scatter is due to those few data points with the lightest rain. Correlation between concentrations of any substance in precipitation and the amount of rain that falls are well known; there is a significant dilution effect involved, which is aggravated by sample evaporation when very small quantities are collected. In the analysis that follows, all incomplete data sets have been excluded. In practice, this amounts to omission of all events with precipitation less than about 2 mm. The remaining data reveal a far reduced role of precipitation amount as a variable influencing precipitation quality.
ISO~-I VERSUS INO~I ESTIMATING 'BACKGROUND' pH
The linear portion of the [SO2-] versus pH plot shown If all industrial emissions of sulfur dioxide were suddenly stopped, what would be the consequence on rainfall pH? This question underlies all discussions of the potential benefits of emission control strategies. Without knowing what limiting situation is achievable, it is not possible to assess how close to this achievable limit alternative control strategies can bring us. The question just posed carries the tacit assumption that sulfur dioxide is the main culprit. As will be seen later, this is not necessarily the case. A parallel question can be posed, regarding emissions of nitrogen oxides. In practice, sulfate and nitrate in precipitation are highly correlated. Here, the influence of NO x emission reductions will be considered as a matter independent of the SO 2 control issue. The questions pertain to areas that are affected by pollution, by sulfur and nitrogen oxides emitted somewhere upwind. Wet deposition is the simplest measurement on which to base any conclusions that may be derived; it also has the largest database. Consider, for example, the data obtained at Whiteface Mountain, NY, as part of the MAP3S program (the Multistate Atmospheric Power Production Pollution Study, of the
in Fig. 1 illustrates a part of the overall behavior pattern, which can be discussed with some confidence--a reduction in sulfate concentrations in precipitation will have an accompanying change in pH associated with it. However, it should be noted that [NO~] could be interpreted in the same way. The reason is that [SO42-] and [NO~] are highly correlated in air, and even more so in precipitation. An interesting first step is, therefore, to explore the comparative roles of [SO42-] and [NO~] at different locations. To this end, a stepwise regression method has been used, as follows: • A simple linear regression identifies which of the concentration variables explains most of the variance in pH. • A multiple regression technique is then used repetitively to order the contributions of the other variables in further reducing the residual (so far unexplained) variance. The sequence is stopped when no remaining variable explains a statistically significant portion of the residual variance. Figure 2 shows how the results are ordered, on a map of the eastern USA. The results are tabulated to show
B B. Hicks, R. S. Artz
140
WHITEFACE S042-
-0.67
NO 3Ca 2÷
-0.41 +0.35
ITHACA S042-
-0.65
NO 3Ca 2+
-0.34 +0.25
BROOKHAVEN NO 3-
-0.82
S042-
-0.40
NH4 ÷
+0.31
STATE COLLEGE N O 3-
-0.77
S042-
-0.39
Ca 2+
+0.29
BONDVILLE
OXFORD
OAK RIDGE
BIRDWOOD
SO42-
-0.50
S042-
-0.63
S042-
-0.82
N O 3-
-0.78
LEWES NO 3-
-0.83
Ca 2+
+0.59
NH4 +
+0.34
Ca 2+
+0.20
SO# 2-
-0.43
S042-
-0.40
NH4 +
+0.48
N O 3-
-0.37
N O 3-
-0.33
Ca 2+
+0.33
Ca 2+
+0.33
Fig. 2. The distribution of stations considered in the present analysis, and the results of a stepwise correlation analysis in which the effects of chemicals that influence pH are eliminated sequentially, in the orders shown for each site.
the order in which site-specific data indicates that imposition of controls would most effectively enable a modification of local pH. The numbers shown are correlation coefficients. For Oak Ridge, TN, for example, the data indicate that the most efficient means to modify local precipitation is to control sulfur. Once the effect of sulfate on pH is eliminated, then the residual variability in pH is best reduced by controlling calcium, but in this case (even if such controls were feasible) the effect of reduction in Ca would be to increase pH. Controls on nitrogen oxides would appear to be a relatively low priority, in this particular case. Clearly, the situation in the mid-Atlantic region is different--there, nitrogen appears to be a more effective modifying agent than sulfur. It is clear that a focus on sulfur is indeed appropriate for most locations, but there is evidence that for some locations [SO4 2-] is not dominant. These stations are in the far east, downwind of major urban areas.
DISCUSSION Figure 1 presents concentration versus pH scatter plots for sulfate, nitrate, chloride, ammonium, sodium, cal-
cium, magnesium and potassium, respectively. The strength of the low-pH linear behavior decreases in the sequence of this listing. A useful interpretation of the overall behavior is as follows: 1. As shown in Fig. 2, the dominant chemical species affecting precipitation acidity is sulfate, but nitrate appears to be nearly as important. In a second category lie calcium and ammonium. There appears to be no relationship for any of the other species that are considered. 2. Consider the situation for sulfate alone. On some occasions, the air at Whiteface Mountain, for example, is largely unaffected by emissions from upwind sources, and on these occasions the precipitation will have characteristics that are indicative of background conditions at this location. In concept, accordingly, the data of Fig. l can be considered to represent two regimes, one which represents a background level in which sulfate is not at high enough levels to lower the pH, a n d the other in which sulfate concentrations are directly influencing pH. The first regime represents a background condition, the second represents the additional influence of sulfur emissions on this background level. The interaction between these two regimes identifies some measure of the level of sulfate in precipitation
Estimating background precipitation quafity from network data
141
Table 1. Values of 'background pH' derived from observations of event precipitation chemistry at sites of the multistate atmospheric power production pollution study (MAP3S) (for comparison, the result obtained using data for a remote site of the global precipitation chemistry network (see Table 2) is also included.)
1.0
0.8
06
"= o.4 .natocns / / ~
~ N O 3~ (SO4 ~ =
)
.~ >" o2
pH
Fig. 3. A plot of the square of the correlation coefficient yielded by a linear regression of pH on chemical concentration in precipitation over increasing pH ranges extending up to the value that is plotted. The curve for sulfate shows a maximum correlation for a pH of about 5-2, which corresponds to a situation in which further reduction in sulfate does not markedly increase pH. Also shown is the behavior for nitrate, computed after allowance is made for the influence of sulfate.
that we should not expect to drop below, even if all of the important upwind industrial emissions are eliminated. 3. A method is needed to derive an objective determination of a measure of 'background' from distributions like that illustrated in Fig. 1. Figure 3 presents the results of a sequential correlation analysis, designated to assess the amount of variance in pH that is directly attributable to sulfate concentration over ranges of sequentially increasing pH, starting with the most acidic data. The diagram shows that the amount of total variance in pH explained increases steadily until the pH is about 5.2. Subsequent values belong to a different r e g i m e - - t h e y detract from the overall correlation, not improve upon it. Hence, the value pH 5-2 can be interpreted as an objective evaluation o f the precipitation acidity that could be used as a target for emission control scenarios for this particular location (Brookhaven, Long Island, NY). 4. Now consider sulfate and nitrate, together with the other chemical species. Sulfate and nitrate are clearly correlated. In essence, dirty rain contains a lot of chemicals of all kinds. It is for this reason that ammonia displays a relationship with pH that appears to be contrary to the expectation based on the correct observation that ammonium is alkaline. 5. A variogram (such as is shown in Fig. 3) for nitrate is revealing, but cannot be easily interpreted because of the underlying relationship between sulfate and nitrate concentrations in rain. In short, nitrate concentrations cannot be assumed to be independent of sulfate concentrations. Consequently, the curve for nitrate that is drawn in Fig. 3 represents the residual variance accounted for by variations in nitrate, after the contributions directly due to sulfate are eliminated
Whiteface Mountain, NY Ithaca, NY State College, PA Charlottesville, VA Brookhaven, NY Lewes, DE Oxford, OH Bondville, IL
5.3 5-0 5.0 5.0 5.2 5.2 5.0 5.0
Katherine, Australia
5.0
from consideration. That is, the nitrate data are derived from a partial correlation analysis, conducted in the same stepwise manner as for sulfate, but computed to reveal the proportion of residual variance explained (instead of the proportion of the total variance). A similar interpretation of this second variogram is now feasible--that controls on emissions of NO x independently of sulfur may not necessarily cause a strong reduction in precipitation acidity, even though for the site now considered, NO~ is marginally more correlated with pH than is SO42- (see Fig. 2). Table I lists the conclusions drawn for all of the monitoring sites considered here--Whiteface Mountain (NY), Ithaca (NY), State College (PA), Birdwood (VA), Lewes (DE), Oxford (OH) and Brookhaven (NY). Also shown is the consequence of an identical analysis of event precipitation data collected at a remote site-Katherine, Australia. The pH value that is revealed is much the same as the value yielded for the USA stations.
C H A R A C T E R I S T I C S O F 'BACKGROUND' PRECIPITATION Alternative methods for revealing 'background' characteristics of precipitation at measurement sites have been tested, and yield similar results. At this time, there appears to be no reason to select any one of the alternative methods relative to any other. The examination of precipitation acidity given above permits identification of a value that is indicative of the 'background pH' that could theoretically be achieved in industrialized areas if stringent emission controls were imposed. Once identified, this separation point can then be used to judge the likely reduction in the concentrations of different chemical species that could be achieved if the 'ultimate and absolute' controls were imposed. This separation-point determination can be refined, however, by using information generated from
142
B. B. Hicks, R. S. Artz
the other methods. Vong (1990), for example, considers precipitation chemistry for stations upwind of major sources, and concludes that 'background' concentrations of sulfate in precipitation are likely to be in the range 2-16 ~ equiv litre -l, and the average background pH is probably about 5.3. A certain amount of natural variability in pH between stations is to be expected, even in the absence of industrial pollution, since pH is affected by many trace chemicals in addition to those on which controls are feasible. Among these contributing chemicals are carbon dioxide, volcanically produced sulfur compounds, nitrogen oxides (produced from lightning, as well as by biological reactions in plants and soils), seasalt compounds (containing e.g., sodium, chloride and sulfate) and many organic compounds. Whatever the cause, it is clear that there are large temporal and spatial variations in the chemistry of precipitation, even at remote locations, and that interpolating among existing wet deposition measurement sites is not a clear-cut exercise. In all areas of the world, rainfall contains trace quantities of chemicals scavenged from the atmosphere. Carbon dioxide is dissolved, for example, until the solution attains a pH of about 5-6 (a H + concentration of c. 2.5 p~ equiv litre-l). In marine areas, the dominant chemicals are the same as are found in high concentrations in sea water, such as sulfate. Data obtained in operations of the Global Precipitation Chemistry Network (Table 2) are particularly revealing. For Amsterdam island, if all measured sulfate were in the form of sulfuric acid, the volume-weighted mean pH would be no higher than c. 4.5. In all areas affected by thunderstorms, nitrate concentrations are partially due to lightning; in the northeastern USA it was estimated that the lightning-generating NO x was 24% as high as anthropogenic NO x during the hour with the greatest frequency of lightning, although this value was typically only 2.5% on average (Pierce et al., 1991); much of the remainder is due to emission from soils. (In contrast, as an average across the USA, about 45% of NO x emissions are due to transportation sources, and about 30% due to electric utilities; relatively little is from natural sources.)
In all areas within or downwind of vegetated areas (including many remote polar and marine areas) organic acids may account for up to c. 65% of free acidity (Keene et al., 1983) and even in the northeastern USA, organic acids may account for c. 16% of free acidity in MAP3S samples (Keene & Galloway, 1984). In addition, sulfur dioxide is exuded from surface fissures, and especially from volcanoes, leading to the presence of sulfate in precipitation. Other sulfur species, most notably hydrogen sulfide and dimethyl sulfide, are emitted by biological processes from wetlands and directly from the ocean. The products of atmospheric reactions involving all of these substances are potentially acidifying, and add to the natural background acidity of precipitation in remote areas. As a group, if organic acids are routinely measured as a part of the sampling protocol, the mean pH of precipitation at remote stations is rarely much higher than about 5-1, unless large amounts of neutralizing materials (soil) are present. In the context of the analysis presented in this paper, this translates to a natural background pH of c. 5.1 to 5.6 at MAP3S stations in the absence of organic acid measurements.
CONCLUSIONS AND R E C O M M E N D A T I O N S There is no technique that will give a precise and robust answer to the question of the consequences on precipitation acidity if stringent control measures are imposed. In the past, data on background precipitation chemistry obtained at remote locations selected to approximate specific target areas of special interest (such as the eastern USA) have been used as guidance. Here, an alternative path has been followed, making use of event precipitation chemistry data collected in the target area. The results obtained are quite similar to the conclusions drawn from examination of data from remote sites: depending on the location, the pH that could be achieved if all industrial sulfur emissions to air were curtailed would be in the range 5-0-5-3. Even in this case, there would be a remaining level of sulfate in the precipitation, amounting to about 5/~ equiv litre -1.
Table 2 a. Examples of differences in the composition of precipitation at different remote locations--a remote island (Amsterdam Island, Indian Ocean), a remote inland area (Katherine, Northern Territory, Australia) and a high elevation inland area (Lijiang, China); (concentrations are volume-weighted mean values, expressed in p equiv litre -I. Note that each of the values listed is an approximation to precipitation quality and that these data refer to different periods, with different sampling protocols. Close comparison is not warranted.)
Site
SO2-
NO 3
NH~f
HCOO-
CH3COO-
H÷
pH
Amsterdam Island Amsterdam Island (excluding sea salt) Katherine Lijiang
29.2 4.8 3.9 5.3
1.6 1.6 4.0 1.6
2.4 2.4 2.9 4.7
2.1 2.1 7.3 3.4
0.8 0.8 2.9 2.3
8.3 8.3 18.3 9.7
5.08 5.08 4.74 5.01
a Data from Galloway & Gaudry (1984) and Galloway, J. N. (1989, unpublished).
Estimating background precipitation quality from network data The present analysis confirms that sulfur is not always the optimal species to control, if the desire is to reduce precipitation acidity. In particular, data obtained in the USA indicate that controls on emissions of sulfur species may not be as effective as would be controls on N O x emissions, if such were possible, for a sizeable part of the mid-Atlantic coastal region.
ACKNOWLEDGEMENTS This work was carried out as a component of Task G r o u p II (Atmospheric Chemistry) of the National Acid Precipitation Assessment Program, under the sponsorship of the National Oceanic and Atmospheric Administration and the US Environmental Protection Agency.
REFERENCES Carmichael, G. R. & Peters, L. K. (1984a). An Eulerian transport/transformation/removal model for SO 2 and sulfate, I, model development. Atmos. Environ., 18, 937-51. Carmichael, G. R. & Peters, L. K. (1984b). An Eulerian transport/transformation/removal model for SO 2 and sulfate, II, model calculation of SO x transport in the eastern United States. Atmos. Environ., 18, 953-67. Chang, J. S., Brost, R. A., Isaksen, I. S. A., Middleton, P., Stockwell, W. R. & Walcek, C. J. 0987). A three-dimen-
143
sional Eulerian acid deposition model: physical concepts and formulation. J. Geophys. Res., 92, 14681-700. Eliassen, A. (1978). The OECD study of long range transport of air pollutants: long range transport modelling. Atmos. Environ., 12, 479-87. Galloway, J. N. & Gaudry, A. (1984). The composition of precipitation on Amsterdam Island, Indian Ocean. Atmos. Environ., 18, 2649 56. Hales, J. M., Hicks, B. B. & Miller, J. M. (1987). The role of research measurement networks as contributors to Fedral assessments of acid deposition. Bull. Am. Meteorol. Soc., 68, 216-25. Keene, W. C. & Galloway, J. N. (1984). Organic acidity in precipitation of North America. Atmos. Environ., 18, 249197. Keene, W. C., Galloway, J. N. & Holden, J. D. (1983). Measurement of weak organic acidity in precipitation from remote areas of the world. J. Geophys. Res., 88, 5122-30. Pierce, T. E., Coventry, D. H., Novak, J. H. & Van Meter, A. R. (1991). Estimating lightning-generated NO x emissions for regional air pollution models. Proceedings 7th AMSA WMA conference on Applications of Air Pollution Meteorology, New Orleans, 14-18 Jan. 1991, pp. 160-3. Sheih, C.-M. (1977). Application of statistical trajectory model to the simulation of sulfur pollution over the northeastern united States. Atmos. Environ., II, 173-178. Vong, R. J. (1990). Mid-latitude northern hemisphere background sulfate concentration in rainwater. Atmos. Environ., 2,4, 1007-18. Watson, C. R. & Olsen, A. R. (1984). Acid Deposition system (ADS) for Statistical Reporting: System Design and User's Code Manual. US Environmental Protection Agency Report EPA-600-8-84-023, Research Triangle Park, NC.