Atmospheric Environment Pergamon Press 1973. Vol. 7,
VISIBILITY,
HUMIDITY
pp. 281-290. Printed in Great Britain.
AND SMOKE IN SYDNEY
M. P. PATERSON* NSW Department
of Health, Division of Industrial Hygiene and Pollution Lidcombe, New South Wales, Australia
Control,
(First received 13 June 1972 and in final form 21 August 1972) Abstract-Meteorological observations taken at 9 a.m. in May at Observatory Hill, overlooking Sydney Cove, in the three periods 1930-1935, 1951-1956 and 1963-1970 have been analyzed. While the mean relative humidity decreased from one period to the next the geometric mean visibility has increased on Sundays in a manner consistent with the calculated regression of log (visibility) on relative humidity. Weekday visibility has increased also, but to an extent less than that which is predicted solely from the relative humidity change. Saturday morning visibility has not increased significantly from one period to the next and has become indistinguishable from weekday morning visibility. Saturday and Sunday relative humidity is statistically indistinguishable from weekday relative humidity in all of the data. Data from direct measurement at ground level of particulate air pollution during the third period are moderately correlated with the log (visibility) while being almost uncorrelated with relative humidity. Estimates from the regression and from comparison of Sunday with weekday visibility suggest independently that the presently observed average smoke concentrations on May mornings lead to a visibility reduction to less than one half of the clean air value. Estimated reductions during the earlier periods, derived from the comparison of Sunday and weekday visibilities, are of the same or slightly smaller order.
1. INTRODUCTION IN THE absence
of a long series of direct observations of pollutant concentrations in an urban atmosphere the trend of pollutant concentrations may be inferred from other related observations. Levels of particulates and visible gases may be estimated from the influence they have exerted on the visible range, a quantity which is estimated regularly as one of the standard surface observations by weather observers. With this single observed quantity it is not possible to distinguish between the different visibilityinfluencing species. With changes in fuels and technology the size and composition of the particulates will change, and it is to be expected that the relationship between concentration of particulates and the effect on visibility also changes. The standard measurement of “smoke”, as discussed in the next section, cannot itself be directly related to mass concentration, number concentration or surface area concentration of particulates in the air. It is not to be expected then, that the effect of pollution on the visible quality of the air can be simply related to that single measurement of smoke concentration. The New South Wales Clean Air Act (1958) provides for control of sources of smoke, sulphur dioxide and other species to maintain within statutory limits either the concentration at the point of emission or the maximum concentration which will be experienced at ground level due to operation of the source. The acceptable levels of smoke emission have been defined on the Ringelmann “shades of grey” scale, estimated by visual comparison with a printed chart. Particulates in flue gases are also determined by isokinetic sampling in the flues of larger installations. By the early nineteen sixties a * Present address: Air Pollution Research Group, Mathematics Science and Technology, London SW 7, England. 281
Department,
Imperial
College of
M. P. PATERSON
282
reduction in smoke emissions from many sources had been effected. Industrial activity and motor transport have both continued to increase in total volume however, and the urban area of Sydney also continues to expand. This study of visibility in Sydney was undertaken to determine whether available information can be used to document changes in the visual air quality of Sydney over a substantial period and estimate from those changes the variation in air pollution prevailing. Analyses have been done on data of two periods before enactment of the Clean Air Act and one period after the Act became effective. An attempt has been made to account explicitly for the role played by water in the atmosphere, known to be the major contributor to the variability of the visible range. The month of May, at the end of the southern Fall, was chosen for the study as the time of greatest measured levels of smoke concentration at ground level. No measurements are available to express the impact of pollution on visual quality, the “murkiness” of the air. 2. THE
OBSERVATIONS
The Commonwealth Bureau of Meteorology has taken the standard set of surface meteorological observations at Observatory Hill, above Sydney Cove, near the centre of Sydney, since the nineteenth century. The measurement of air pollutants, smoke being of specific interest here, by the Air Pollution Control Branch of the New South Wales Department of Health, was commenced at George Street North, 400 yd away, only in 1962. This study has used data from the two sites to determine the magnitude of and trend in visible pollutants concentration and its effect on the atmosphere of central Sydney. The dependent variable is visibility, the distance of the furthest resolvable objects. The independent variables are relative humidity, computed from temperatures measured at a point, and the optical density of non-volatile atmospheric particulates collected on a filter paper, hereafter called “smoke”. The study covers three periods between 1930 and 1970, using May 9 a.m. data. Visibility is the estimated horizontal range of vision averaged around the four quadrants. A good discussion of validity and limitations of visibility estimates can be found in MIDDLETON’S (1952) book, Vision Through the Atmosphere. With the recent erection of tall builidings in central Sydney the visibility observations can no longer include the southeast quadrant. Visibility usually falls between 100 yd and 40 miles, a range of three decimal orders of magnitude. Its precision within a factor of two, onethird of a decimal order of magnitude, is adequate for a statistical study such as this. A cursory glance at the data suggests that the logarithm of the visibility is approximately linearly related to the relative humidity and perhaps also to the smoke density. Relative humidity is a measure of the free energy of water vapour in the air. As such it determines the magnitude of various interactions between the water vapour and deliquescent, hygroscopic, soluble and insoluble particulates which are present in the air. Sea-salt particles, for example, form a saturated solution droplet by taking up water vapour when relative humidity exceeds 73 per cent and the equilibrium size for relative humidity ranging up to 100 per cent is a function of relative humidity. Species such as sulphuric acid will start to change their equilibrium size and thus their effect on light scattering, at relative humidities as low as 20 per cent. Relative humidity has thus been selected for use in this study as an appropriate measure of water vapour in the air.
Visibility, Humidity and Smoke in Sydney
283
Smoke in the atmosphere defies precise definition. For purposes of m~surement it may be defined as the sum of natural and man-made particles collected on a filter paper when a sample of air is drawn through it. The man-made particles constitute the majority in the air in and around cities. In the past the industrial and domestic use of fossil fuels has contributed the bulk of such particles to the urban atmosphere. In many cities today this dominance has been replaced by particles from motor vehicles, both those emitted immediately as solid and liquid aerosol and those which condense from products of chemical reactions in the atmosphere, The measure of smoke is the coefficient of Haze, COH, the unit being the amount required to reduce light transmission through the filter by 1 per cent. Concentration is expressed in COH’s per thousand linear feet. In some studies it has been found that 1 COH/lOOO linear feet is equivalent to 20-30 pg of suspended solids per cubic metre of air. Such a concentration cannot be determined directly by light beam extinction in the atmosphere due to the substantial role played by water vapour. 3. WHERE
IS SYDNEY?
Sydney is a city of three million on the southeast coast of Australia. The city lies across the east-facing mouth of the Nepean River basin. In May the area is under the influence of the high pressure belt, with westerlies prevailing aloft. The clear skies result in a high incidence of nocturnal radiation temperature inversions. These dominate the surface weather of the basin, with drainage of cold air in the inversion layer determining local winds throughout the basin. Under inversion conditions Sydney and suburbs experience the westerly drift of air flowing out of the basin at an average speed of six knots measured at Observatory Hill at 9 a.m. The visible top of the surface haze layer has been estimated by Pender to range from 400 to 800 ft at Sydney at this time of year. Topographic contours substantially enclosing the basin extend only to 1000 ft m.s.1. Surface heating during the morning generally results in breakup of the surface inversion between 9 a.m. and noon. This can be seen in the substantial clearing of the surface layers and the reduction of ground level pollutant concentrations measured in Sydney during the late morning. The inversion is re-established toward sunset. Sydney’s situation has at times been compared in air pollution potential with that of Los Angeles. This ~ompa~son is somewhat misleading. Apart from the disparity in total pollution source strength, in Sydney’s favour, the source distribution is significantly different. The Sydney sources are placed at the outlet of the basin, where drainage wind speeds are a maximum. At present only scattered sources are found in the more sluggish air of the body of the basin. Substantial sources will be excluded from the upper reaches of the basin by effective planning. Los Angeles already has an even source distribution throughout its basin. In Sydney’s case other winds do not return the surface drainage flow into the basin and the natural drainage of the inversion layer discourages extended residence of air in the basin. The problem of Los Angeles is compounded by the greater height of the surrounding moun~ins. 4. HUMIDITY
AND
VISIBILITY
Do the days of the week differ from each other in visibility? Comparisons have been made between the distribution of visibility for each day of the week and the distri-
M. P. PATERSON
284
bution of visibility for all days of the week. Separate comparisons have been made for each of the three periods considered in this study. FIGURE 1 shows the geometric mean 9 a.m. visibility in May by day-of-the-week for each period. Each mean has been determined from 26 or more observations. Sunday visibilities appear to have been consistently higher than weekday values. Saturdays appear to have enjoyed a lesser difference than Sunday.
1
! ~
on.
Sue
Wed.
Th u.
FI
FIG. 1. Geometric mean visibility by day of week for the three periods, 193&1935, and 1963-1970, from May 9 a.m. observations at Observatory Hill.
1951-1956
To test the significance of the differences in distribution the rank sum test as described by HOEL (1954) has been applied to the daily data on the hypothesis that day-of-the-week distributions of visibility do not differ significantly from the overall distribution in any one period. No weekday distribution, Monday-Friday, showed differences from the overall distribution for the period significant at the 2 per cent level on a two-sided test. Significance levels of the Saturday and Sunday distribution differences from overall distributions, all representing greater visibility, are given in TABLE
I. TABLE 1. SIGNIFICANCELEVELSON THE RANK SUM T~sr FOR THE DIFFERENCESIN EACH OF THREE PERIODS BETWEEN SATURDAYDISTRIBUTION OF GEOMETRIC MEAN VISIBILITYAND ALL-DAYSDISTRIBIJTION FOR THE PERIOD AND BETWEEN SUNDAY DISTRIBUTION OF GEOMETRIC MEAN VISIBILITYAND THE ALL-DAYS DISTRIBIJTION FOR THE PERIOD 1930-1935 Saturday Sunday
2.2% 1.1%
1951-1956 8.2% 1.2%
1963-1970 20% 0.004 %
In each of the three periods the difference between Sunday 9 a.m. visibility and the overall distribution has been significant at the 2 per cent level. Differences between Saturday 9 a.m. visibility and the overall distribution have decreased in significance with each successive period. In the last two periods the Saturday visibility has been statistically indistinguishable from the overall distribution.
Visibility, Humidity and Smoke in Sydney
285
Inspection of FIG. 1 shows Monday morning visibility in each period to fall below the visibility on Tuesday and Wednesday. There is no evidence here of persistence into Monday of the clearer atmosphere of Sunday. Indeed some additional factor to opposite effect may be present. Such a factor might be the Monday morning start-up of plant which operates through the working week. Such a start-up can result in considerable emission of smoke until the combustion becomes efficient. If we assume that Sunday air quality, measured by smoke content, has not improved substantially over the forty years of the study an explanation is needed for the increase from one period to the next in the Sunday 9 a.m. geometric mean visibility. Relative humidity is one naturally varying factor known to have an effect on visual range. Natural dust and smoke is another but is not conspicuously present on the New South Wales coast in May. Mean 9 a.m. relative humidities for the three periods, and within each period by day-of-the-week, have been calculated. They are displayed in FIG. 2 as the abscissa for the day-of-the-week point, with visibility shown as the ordinate on a logarithmic scale.
Relative
humidity
FICA2. Mean relative humidity (%) is shown on the ordinate and geometric mean visibility on the abscissa. Each point represents one day of the week (M-Monday, etc.). Points in the tirst period are circled and those in the second are enclosed in squares. The lines (i), (ii) and (iii) show regressions of log (visibility) on RH for the three periods, 1930-1935, 1951-1956 and 19634970.
It will be seen that relative humidity has fallen from one period to the next. Within each period the Saturday and Sunday mean relative humidities are indistinguishable from weekday values. Mean humidity has fallen from 78.9 per cent in the first period to 75.0 per cent in the second and 705 per cent in the third period. The relative humidity data presented here cannot be taken as certain evidence of a significant climatic change between 1930 and 1970. Data on the annual average relative humidity between 1920 and 1970 shows a similar relationship between averages for the three periods studied but no conspicuous overall trend to lower relative humidity. On the other hand, the COMMONWEALTH BUREAUOF METEOROLOGY (1965) and EDGAR and
M.P. PATERSON
286
LINACRE(1972) have pointed out that the frequency of fogs in Sydney has decreased in the past 40 or more years. The fog frequency is a measure only of one extreme end of the distribution of relative humidities. SETTREE(private communication) has reported a reduction in fog frequency on Jervis Bay, 110 miles south of Sydney, over the past 40 y, He also notes other indications of climatic change at Jervis Bay. Bureau of Meteorology publications of the 1930’s recorded long-term mean May 9 a.m. relative humidities higher than those now recorded. DEACON(1953) and DAS (1956) have adduced trends in temperature, rainfall and surface pressure gradients as evidence of a secular trend in climate of the eastern half of the Australian continent. The 9 a.m. daily data has been analyzed for each period to determine the linear regression of log (visibility) on relative humidity. The regression coefficients are displayed in TABLE2 and the regressions drawn onto FIG. 2. The regressions have not been tested for non-linearity, though a visual inspection shows no discernible curvature. FIGURE2 shows that over the range of humidities observed, roughly 45-100 per cent, regressions for the first and second periods show a high degree of coincidence while the third regression has comparable slope but is displaced to visibilities which are lower by 20 per cent. On the basis of these regressions the improvement in Sunday visibility, from 6.3 to 9.0 and then to Il.3 miles in successive periods is consistent with the humidity change from 79.7 to 75.9 per cent and then to 70.7 per cent. Comparable improvements in weekday visibility have failed to occur between the second and third periods. The S.D. of the relative humidity values observed in each of the three periods was about 14 per cent. The geometric standard deviation of visibility was in each case a factor of about three. TABLE 2. LINEARREGRESSION COEFFICIENTS FORTHREE PERIODS, FITTEDTO THE MAY 9 a.m. METEOROLOGICAL DATAFROMOBSERVATORY HILL, SYD~Y, IN THE FORM:LOG(VJXBIJJTY)= a0 -+-aI r.h.
Con.% term Slope Corr. coefft. Observations 5.
HUMIDITY,
aa al Y n VISIBILITY
1930-193s
1951-1956
1963-1970
2.55 -0.0239 -0.704 186
2.81 -0.0275 -0.697 186
2-46 -a0242 -0.712 246
AND
SMOKE
IN
THE
THIRD
PERIOD
Smoke concentrations data have been available from 1966 onward to supplement the visibility and relative humidity data. The data consist of 2-h average smoke concentrations measured between 8 and 10 a.m. For this period an average smoke concentration of l-64 units prevailed, compared with a 2-4 p.m. average of O-86 units and daily average of l-20 units. FIGURE 3 shows the mean concentrations for S-10 a.m., 2-4 p.m. and 24 h by day of the week. The 24-h period runs from midnight to midnight. A planar regression of the form
log (visibility) = a0 i_ a, COH + a2 RH has been fitted by the method of least squares to the 9 a.m. data. A total of 1.53data points in the period 1966-1970 have been used, these being points for which all three variables are known. The calculated regression is log (visibility) = 2-69 - O-15 COH - 0.024 RH
287
Visibility, Humidity and Smoke in Sydney
The analysis of variance yields rCOH, rRH. rCOH.
losWs) 1oPWS) RH
=
-
0.55
r2 COH.
=
-
0.73
r2RH,
040
r2COH.
. COH.
RH
=
r2 log(vis)
=
=
O-30
log(viS)
loo(vts)
=
0.54
RH
=
0.16
0.62
Thus, of the variation of log (visibility), 30 per cent can be accounted for by variation in COH alone, 54 per cent by variation in RH alone and 62 per cent by the combination of the variables. Observe the low correlation between relative humidity and smoke, with r2 of only O-16. 2.oL
1.5 -
I.0
-
0.5
-
If.
Sun.
I I ml.
it.
Sun.
D”.
FIG. 3. Mean May smoke haze concentrations at George Street North, expressed in COH units per 1000 linear feet, by day of week. Period 19661970. Solid black bar--l-l0 a.m. White bar-2-4 p.m. Vertically bisected bar-24 h to 9 a.m.
The coefficients in the regression a, and u2, are both significantly different from zero at the 1 per cent level. On the estimated values of the coefficients an increase of one COH unit in smoke density produces a decrease of visible range by 32 per cent and an increase of 1 per cent in relative humidity produces a decrease of visible range by 5 per cent. A 7-5 per cent change in relative humidity will affect visible range as much as a 1 COH unit change in smoke density. At a mean 9 a.m. smoke density of 1.64 COH units the visibility reduction by smoke is estimated to be 42 per cent, equal to the reduction which will result from an 11 per cent increase in relative humidity. The relative humidity variation is thus seen to account for most of the variation in visibility. The smoke concentration accounts for only a small additional percentage leaving some 38 per cent of the variation in log (visibility) unexplained. Some arises from errors in each of the three observations which have gone into the calculation. Instrumental errors and error in distance estimation may be less significant than the sampling errors resulting from the characterization of mean relative humidity over one to several thousand square miles by the values at a single central point. This sampling error should be most significant in the measurement of smoke concentration, known to be highly inhomogeneous in such a situation close to the major sources of smoke. The remaining factor unaccounted for is the visually obscuring trace components of the air which are not represented on the filter paper in the smoke monitor. These may be coloured gases or volatile liquid or solid particles. Nitrogen dioxide is such a
M. P. PATERSON
288
gas observed in atmospheres polluted by motor car exhausts. This factor has been discussed by HORVATH(1971). The photochemical smog which results from car exhaust pollutants in suitable conditions also contains particles condensed from low vapour pressure species which are chemical reaction products, and which may be re-evaporated after collection on the filter. It is unlikely though, that the photochemical smog reactions would be far advanced by 9 a.m. in May at 34”S, the latitude of Sydney. Some of the variation perhaps arises from the application of a mathematical form of regression which is not strictly appropriate to the phenomenon to light attenuation in the atmosphere with interaction between water vapour and particles. The frequently used form (see for example RODHEet al., 1972) applied to a dry atmosphere is (visible range)-’
= constant x (smoke density).
It should be noted, though, that the fitting of a linear regression to this form is unsatisfactory for two reasons. The variance is not evenly distributed along the range of observations, being least for the lowest values. Furthermore, the density of observations is generally heavily biased toward the low values of both variables, the end of the range with the lower variance. The form used in this study, assuming a linear relationship between smoke density and log (visibility), is far from ideal and yet it accounts for almost two thirds of the variation in log (visibility) when applied to rudimentary data. Had the water vapour and the smoke been transported with the air from a common and distant source region a high degree of correlation might have been found between the two visibility-affecting species. In such a situation a failure to deal with all significant variables in a single analysis would produce misleading results and perhaps attribute the major effect to the wrong species. To use visibility data to compute the air’s particle load with scant control on humidity conditions, as do RODHE et al. (1972), may involve a systematic bias due to a correlation between particulates and relative humidity. 6. INTERPRETATION (i) Absolute concentrations of smoke. The high degree of coincidence of the regressions of log (visibility) on humidity in the first and second periods suggests that the mean smoke concentrations in the layer of air in which visibility is observed were similar in the first and second periods. The geometric mean visibilities differed by over 20 per cent, in keeping with the lower mean relative humidity of the second period. Between the second and third periods however, the geometric mean visibility failed to show an increase, despite a further decrease in relative humidity. It is estimated that in the absence of any increase in smoke the mean visibility would have increased by a further 20 per cent. We therefore attribute a 20 per cent reduction in visibility to the increase in smoke between the second and third periods, and from the regression computed in Section 5 we estimate this increase to be about O-5 COH units to the final value of 1.6 COH units. These values all refer to 9 a.m. conditions in May. The beneficial effects on visibility of implementing the Clean Air Act are yet to be observed or demonstrated. (ii) Redistribution in the week. Two phenomena concerning distribution of the variables by day of the week are to be distinguished. The change in Saturday 9 a.m.
Visibility,Humidity and Smoke in Sydney
289
visibility from weekend character to weekday has already been mentioned in Section 4. With the limited data of this study its cause must remain a matter of conjecture. An industrial change from a 5-day to a 5$-day week could produce the observed change. The growing significance of the motor car as a pollution source and its heavy use through Saturday may equally well explain the change. A third possibility is that the enlarged Sydney of today stores polluted air from Friday in the basin to the west and this is still draining past Observatory Hill at the time of the Saturday morning observation. The third suggestion may seem to be at odds with the second phenomenon to be discussed in this section. This is the nature of the Monday morning onset of pollution. It has been observed in Section 4 that very low visibility has been recorded in each period on Monday mornings. This is not immediately consistent with the storage mechanism suggested above or with the observations of smoke at George Street North in the third period which show surface smoke concentrations on Monday to be rather lower than concentrations on all other weekdays. The explanation may be that the visible range is determined in a layer of air somewhat above the surface while the smoke is sampled at ground level. The residual effect of Sunday’s clean air can still be measured at ground level while the air at higher levels has been substantially polluted by the emissions attending the start-up of heating and industrial plant for the week. (iii) Visual evaluation of the pollution. People who have observed Sydney’s atmosphere over many years consider that the winter morning pollution layer “looks worse” now than it did some years ago. Such evaluations are made not on the visible range through the atmosphere, but on the apparent murkiness of the air. It is conceivable that in the absence of any change in the concentration of pollutants a change in visible range might lead to a changing impression of murkiness. At lower relative humidity the white veil of the limit of vision caused.by the water droplet haze recedes from the viewer. This leaves a greater depth of the coloured pollutants along the line of sight in the intervening air. The impression is of an atmosphere with more brown or black matter, that is, one which looks worse. The effect will be enhanced if instead of remaining constant the pollutant concentration increases as the relative humidity decreases. On the figures obtained in this study, if the geometric mean weekday visibility has increased from 4 to 5 miles between the first and third periods, while smoke concentration has increased from 1.1 to 1.6 COH units per thousand linear feet, then the total content of pollutants in the visible column of air has risen from 23 COH units to 42 COH units. The impression of a substantial increase in pollution is to be expected, though the geometric mean visible range has increased. 7. CONCLUSIONS It has proven impossible in this simple study to analyze air pollution trends without taking account also of climatic variations. Visible range, though a poor indicator of air pollution, is the only one in Sydney for which more than forty years of records are available. It is determined predominantly by relative humidity but there is sufficient residual dependence on smoke concentration to allow the effects of the two variables to be separated. Many other pollution measurements can be affected to some extent A.%1/3-D
290
M.P. PATERSON
by climatic factors. Failure to account for these could invalidate deductions about the effectiveness of air pollution control measures. This study has yielded no evidence of the effectiveness of the NSW Clean Air Act. If visible range and murkiness of the air are mainly affected by pollution from car exhausts then control of industrial emissions through the Act may give no detectable improvement in visual air quality. Climatic variations, as well as causing variations in perceived pollution, can affect the measured levels of pollution. SCHMIDTand VELDS(1969) have reported a major decrease in Rotterdam’s measured SO2 pollution which may have resulted in large part from a trend to higher wind speed and rainfall. It is not necessarily possible to statistically separate the effects on visible range of natural factors, such as humidity, and man-made factors such as smoke and other visibly obscuring particles. Had the water vapour and smoke been transported from a common distant source region a high degree of correlation might have been found between the visibility-affecting variables. In such a situation a failure to deal with all significant variables in a single analysis would produce misleading results and perhaps attribute the major effect to the wrong species. Acknowledgements-This study has been made possible by the cooperation of the Commonwealth Bureau of-Meteorology, Sydney Regional Office, in furnishing the meteorological records, and of Dr. N. B. JONESand Mr. S. LANNIGAN-O’KEEFE of the NSW Treasury ADP Section who collaborated on the computations. The initial work on this problem was performed by Mr. D. JOHNSTONof the Air Pollution Control Branch, to whom the author acknowledges his indebtedness. Miss Gaye Ferguson gave invaluable assistance in the data handling. The paper is published with the authorization of the Director-General of Health in N.S.W.
REFERENCES COMMONWEALTH BUREAUOF METEOROLOGY (1965) The Climate of Sydney and its Environs. Pamphlet, 40 PP. DAS S. C. (1956) Statistical analysis of Australian pressure data. Aust. J. Phys. 9, 394-399. DEACONE. L. (1953) Climatic change in Australia since 1880. Aust. J. Phys. 6,209-218. EDGAR C. and LINACREE. T. Private communication. HOEL P. G. (1954) Introduction to Mathematical Statistics. Wiley, New York. HORVATHH. (1971) The brown colour of atmospheric haze. Atmospheric Environment 5, 333-344. MIDDLETONW. E. K. (1952) Vision Through the Atmosphere. University Toronto Press, Canada. PENDER E. Private communication. RODHE H., PJXR~~ON C. and AKESSON0. (1972) An investigation into regional transport of soot and sulfate aerosol. Atmospheric Environment, 6, 675-694. SETTREEA. Private communication. SCHMIDTF. H. and VELDSC. A. (1969) Changing meteorological circumstances and the decrease in SO, concentration around Rotterdam. Atmospheric Environment 3,455-W.