A summary

Nashville Air Pollution and Health Study

P. A. KENLINE, M.S., and C. J. CONLEE, B.S.

EXTENSIVE air pollution and health in relation to quantity, composition, and the assess ANu study was conducted in Nashville, Tenn., factors affecting it; and (b) how to the in 1958 and 1959 by the Division of Air Pollu¬ health effects that might result from this con- tion, Public Health Service, and Vanderbilt tamination as they vary in geographic and tem- University School of Medicine, together with poral distribution and with the recognition that cooperating State, county, and local agencies. many other factors also might affect them. The The purpose was to investigate possible rela- large-scale investigation was undertaken pri- tions between air pollution and health and to marily to ascertain whether proper measure- gather information on a number of medical, ments and experimental procedures could be engineering, and meteorological phases of air designed to meet these objectives. Such a was a pollution. study prompted by and Conduct of need to assess the chronic or long-term health Objectives Study effects that may occur as the result of daily, and The study (1) was divided into two concur- often lifetime, exposure to air pollution. Nash¬ rent phases: medical and engineering. ville was selected because it was believed to Medical phase. Four separate but related have a chronic but relatively modest, air pollu¬ medical studies were conducted: a morbidity tion problem. In large measure, the pollu¬ survey, a mortality survey, a cardiorespiratory tion was caused by the combustion of coal. disease study, and an anthracosis study. Numerous scientific articles have been written To determine the possible relations between on various aspects and phases of this study. air pollution and health, the following objec¬ In this paper we have presented a bibliography tives were outlined. of all the published articles, together with a 1. Collect general and specific morbidity data brief outline of the study and a summary of on residents of Nashville and determine if there the important findings. is an association between morbid conditions and The multiple objectives of the Nashville study exposure to air pollutants. stemmed from two primary problems: (a) how 2. Determine mortality rates for specific to assess the atmosphericpollution of a commun- causes by census tracts and study relation be¬ ity most accurately, with reference to both tem- tween these death rates and demographic, social, poral variations and long-term averages, and and economic data. 3. Determine whether deaths from specific Mr. Kenline was chief and Mr. Conlee was sanitary causes vary significantly with residence in areas engineer, Aerometry Section, Division of Air Pollu¬ of relatively high or low atmospheric pollution tion, Public Health Service, Robert A. Taft Sanitary that have similar demographic, social, and eco¬ Engineering Center, Cincinnati, Ohio. Mr. Kenline nomic characteristics. is currently with the Career Development Program 4. Determine whether persons known to have of the Division of Air Pollution, and Mr. Conlee is specificcardiorespiratory disease show variation with the University of Washington, Seattle. in frequency or severity of attacks, depending

Vol. 82, No. 1, January 1967 17 on diurnal or seasonal variation in exposure to 6. Determine the relations between air pollu¬ environmental air pollutants. tion concentrations at a selected station and 5. Determine whether the course of disease in meteorological parameters measured at the same persons known to have a specific cardiorespira¬ station and at different stations. tory disease differs according to their residence Sampling stations were established at 123 sites in areas of relatively high or low air pollution. in the Nashville area. Of these, 119 were within 6. Determine whether autopsies on residents an area about 9 miles in diameter, centered at a of Nashville showed a greater degree of anthra- downtown site. The sites, located on an equi- cosis than autopsies on nonresidents. lateraJ triangle grid pattern, were about 0.87 7. Determine whether the degree of anthra- mile apart. The remaining four control stations cosis in Nashville residents varies according to were located in the four cardinal compass direc- where they reside in the city. tions about 8 miles from the central station. 8. Determine whether anthracosis is com- The sampling network consisted of three gen- monly associated with any other pulmonary or eral types of stations (determinedby the amount cardiac pathology and whether anthracosis is and kind of aerometric and meteorological identified with any specific symptomatology. equipment) and several stations for specific A questionnaire designed to produce gen- measurements. The urban area was covered by eral and specific morbidity information was the three general types of stations in a reason- distributed to residents in approximately 5 ably uniform manner. A summarization of the percent of the occupied dwellings. A portion major aerometric measurements taken at the of the test group was examined medically to various stations follows: a check on of information. provide validity Number of stations Aerometric measurements Mortality data were obtained from Davidson death certificates from 1949 to 1956. ( Monthly dustfall. County 123_\ Monthly sulfur dioxide (lead peroxide Patients suffering from cardiorespiratory dis- \ candle method). eases were examinations given complete physical 24-hour daily soiling spot samples. after a detailed was obtained. The history pa¬ 36- 24-hour daily volumetric sulfur dioxide tients mailed report cards with appropriate in¬ (West-Gaeke method). formation at specified intervals after receiving 24-hour daily total wind movement. their examinations. 12 2-hour (high-pollution seasons) or The results of autopsies performed at Vander- 6 4-hour (low-pollution seasons) daily bilt University Hospital between 1953 and 1956 11- soiling spot samples, American Iron data for the anthracosis and Steel Institute (A.I.S.I.) method. provided study. 24-hour daily suspended particulate Engineering phase. A number of related ob¬ (high-volume samples). were interwoven into jectives the investigations __Continuous ambient temperature, rela- with the mission of providing air-quality data tive humidity, and wind speed and for use in medical studies. The following objec¬ direction. tives were outlined. __12 2-hour daily volumetric sulfur di¬ 1. Determine the representativeness of air- oxide (West-Gaeke method). quality data with respect to number and loca- ( Continuous sulfur dioxide (Thomas tion of sampling stations, sampling frequency, autometer). and time. Continuous nitrogen dioxide and oxi- sampling .{ dant (Kruger model 73 atmosphere 2. Evaluate analytical methods, equipment, analyzer). and procedures. Continuous carbon monoxide (Luft-type 3. Study and evaluate dispersion character¬ infrared analyzer). istics of pollutants. Special studies consisted of a sulfur dioxide 4. Determine public opinion on air pollution (S02) emission inventory, a study of dustfall as related to pollution levels and population measurement techniques, a public opinion sur¬ characteristics. vey, pollen measurements, pollution level fore- 5. Attempt to classify atmospheric stability casts, and automobile traffic counts. by use of radiosonde data. The numerous technical and scientific papers

18 Public Health Reports relating to air pollution in Nashville produced HE. "Morbidity in Kelation to Air Pollution" many important medical, aerometric, meteoro¬ (4). A method for studying the association of logical, and misceUaneous findings. air pollution and morbidity in an urban popula¬ tion and its limitations were discussed. Direct Medical Findings correlations of total morbidity and levels of Several relationships between air pollution pollution as measured by the soiling index and and health have been determined from the 24-hour S02 were observed among persons 55 studies of Zeidberg and associates. Their find¬ years old and older of the middle socioeconomic ings are presented in the following series of six class. Direct correlations for the same aero¬ papers on "The Nashville Air Pollution Study." metric parameters were also noted for cardio- I. "Sulfur Dioxide and Bronchial Asthma; vascular diseases but not for any other specific A Preliminary Report" (2). A group of 84 group of diseases. patients (49 adults and 35 children) with bron¬ Refinement of analyses to differentiate home chial asthma reported 3,647 asthmatic attacks and occupational environment influences on during 27,440 person-days of observation, or a morbidity revealed that white women keeping total attack rate of 0.133 per person-day. In house manifested direct correlations of pollu¬ adults, the asthmatic attack rate varied directly tion and total morbidity for all aerometric with the level of sulfation in their residential parameters, while white women who worked environment. Attack rates on days with the showed none. Nonwhite women tended to show highest and lowest S02 values were significantly direct correlations in both housekeeping and different. A 1-day lag showed even more sig- woridng groups. The effect of pollution on nificant differences. specific diseases of the respiratory, cardio- The influence of temperature, humidity, and vascular, or gastrointestinal systems could not barometric pressure on the asthmatic attack rate be demonstrated because such breakdowns pro¬ could not be demonstrated, but wind velocity duced numbers too small for valid analysis. showed an inverse relation. Pulmonary func- IV. "Sulfur Dioxide and Bronchial tion tests indicated that persons with a 1-second- Asthma; A Multivariate Analysis" (5). This timed vital capacity of less than 50 percent had progress report demonstrates the association of a significantly higher attack rate than patients a variable (day from beginning of study) with with more than 75 percent function. asthmatic attacks. The next most important U. "Pulmonary Anthracosis as an Index of variable in the multiple correlation and regres- Air Pollution" (3). In 641 consecutive autop¬ sion analysis was sulfur dioxide. Much ad- sies, excluding those on subjects under 5 years ditional work was thought necessary to in- of age, performed at Vanderbilt University vestigate the possibility that susceptible people Hospital from 1953 to 1956, microscopic evi- have attacks on or aboutthe same day. Further dence of pulmonary anthracosis was sought. study also was recommended on the environ¬ The deposition of anthracotic pigment in- mental factors characterizing the days on which creased with age; it was more severe in men large numbers of patients report attacks. It than in women and in Nashville residents than was also suggested that the linear analysis made out-of-city dwellers. Among Nashville resi- thus far should be supplemented by nonlinear dents, it increased with length of residence in methods. the city and was more severe, at least in women, V. "Mortality From Diseases of the Respira- in those who had lived in the more polluted tory System in Relation to Air Pollution" (6*). areas of the city. The influence of occupation This study of respiratory-disease mortality in could not be demonstratedbecause of insufficient relation to air pollution was designed to control occupational history. socioeconomic factors as well as degree of ex¬ No association was found between anthracosis posure to air pollutants. and specific pulmonary or cardiac symptoms or Except for lung and bronchial cancer, mor¬ pathology. Pulmonary anthracosis appeared tality from respiratory diseases varied inversely to reflect a person's exposure to environmental with the socioeconomic class when the degree of air polluted with coal smoke. exposure to air pollutants was kept constant.

VoL 82, No. 1, Jamuuy 1967 19 With the socioeconomic factor controlled, res¬ Except for arteriosclerotic heart disease, cardio¬ piratory disease mortality was directly related vascular mortality rates were higher among to the degree of exposure to sulfation and soil¬ nonwhites than whites of the same socioeco¬ ing, except for lung and bronchial cancer, nomic class. This was generally true at all bronchitis, and emphysema. Age-specific res¬ pollution levels. piratory disease mortality rates up to 65 years Mortality rates for cardiovascular disease in- of age were directly related to the degree of creased regularly with age, except for rheumatic exposure measured by sulfation. After 65 heart disease, which peaked at 65 to 74 years. years, the highest rates were observed in the When analysis was made on the basis of air low pollution group. pollution exposure, the greatest proportional At all levels of exposure to sulfation, respira¬ difference was found in those under age 45, and tory disease mortality rates were higher for the smallest difference in those age 75 and over. men than for women. The difference was espe- This analysis pointed to the need for addi- cially marked for mortality from lung and tional studies designed to control such factors bronchial cancer. Except for lung and bron¬ as stress, smoking, diet, and occupation. chial cancer, respiratory-disease mortality rates for nonwhites were higher than for whites Aerometric Findings at high and moderate levels of exposure to sul¬ fation. The deviation of mortality by lung Stalker and co-workers outlined their work and bronchial and on "Sampling Station and Time Requirements cancer, bronchitis, emphy¬ for Urban Air Pollution in a sema from the shown other Surveys" group pattern by respira¬ of four papers. Their follow. tory diseases was discussed. findings VI. "Cardiovascular Disease Mortality in I. "Lead Peroxide Candles and Dustfall Col- Relation to Air Pollution" Socioeconomic lectors" (8). Seven uniformly spaced lead per¬ (7). oxide air stations were selected to factors as well as of to air sampler degree exposure estimate the mean seasonal and annual sulfation were controlled in this of pollution study rates over the the relation of cardiovascular disease to air community. The estimates pollution. agreed with the results obtained at the 119 sta¬ When the level of to tions actually used, but with a considerably less exposure air pollutants statistical of was a degree confidence. kept constant, generally regular pattern use of two mile or 119 of an inverse relation between socioeconomic By samplers per square the true mean class and rates for cardiovascular dis¬ sampling stations, annual sulfa¬ mortality tion rates and dustfall levels were estimated ease was observed for the Nashville metropol- within 10 with a 95 of itan area for the 1949 percent percent degree 12-year period through confidence. This of stations was re- 1960. An was arteriosclerotic heart density exception to describe the disease, which showed a direct relation. Of the quired adequately geographic four studied for their association distribution for the purpose of supporting the pollutants health-effects of this In the selec- with cardiovascular the most con- phase study. mortality, tion of stations for the determination sistent was noted for sampling pattern suspended particu- of sulfation rates or late matter as measured the index. frequency distributions, by soiling of to be the most The authors had no explanation for this uniformity spacing appeared observation. important criterion; this became more appar- With the socioeconomic factor controlled by ent as the number of sampling stations was the to the middle-class reduced. limiting analyses popula¬ In stations for the deter¬ tion, men generally were subject to higher selecting sampling mortality rates from cardiovascular diseases mination of dustfall levels or frequency distri¬ than women. This was particularly true for butions, the random selection of stations on a arteriosclerotic heart disease. When compari- uniform network grid appeared to be the most sons of mortality were made by air pollution important criterion. Again this became more exposure for each sex, a regular pattern of asso¬ apparent as the number of sampling stations ciation appeared for women but not for men. was reduced.

20 Public Health Reports When circumstances required a sampling pro- from the respective sample means obtained at gram of limited duration it was possible, at the seven-station urban network, by means of least in communities resembling the area investi- which 2,098 samples were collected during the gated in this study, to obtain reliable estimates study. Confidence limks for the smaller net¬ of annual means and frequency distributions work mean were very good (±19 percent), con- for both sulfation and dustfall bysamplingonly sidering the small number of samples collected. during the fall or spring season. Cumulative frequency distributions of meas¬ The authors devised a statistical tool that will urements in the reduced sampling programs assist any investigator in designing and evaluat- agreed well with that of the seven-station urban ing his sampling program. network, and geographic distribution of sus¬ IL "Suspended Particulate Matter and Soil¬ pended particulate matter could be fairly well ing Index" (9). At least 60 sampling stations described by several of the smaller sampling (about 1 station per square mile) for this com- programs. None of the suspended particulate munity would be required to estimate the daily networks used in this study, however, would mean level of suspended particulate matter with supply the geographic detail desirable for the 95 percent assurance of being within 20 percent best support of a communitywide health-effects of the daily mean value. At the seven urban study. stations used in the study, the true daily mean For 95 percent assurance of estimating the (±20 percent) was estimated with good assur¬ daily mean community-soiling index (24-hour ance on only 8 percent of the 26 days of study. samples) within 20 percent of the true mean, However, monthly, seasonal, and annnal means about six uniformly placed stations per square were probably estimated within 35 percent of mile would be required. The stations would the respective true mean values. About one be spaced slightly less than one-half mile apart. station per 4 square miles would be required to By means of the 32-station network (one station estimate monthly, seasonal, and annual means per 2 square miles) used in the study, the true with ±20 percent accuracy. daily mean (±20 percent) was estimated with By means of a network consisting of only good assurance only on 6.3 percent of the 23 days two sampling stations (one high-pollution of the study. However, true monthly and sea¬ urban and one suburban), the annual mean sonal means most likely were estimated within particulate matter over the community was esti¬ 30 percent, and the true annual mean within mated within 2 percent of the sample mean 20 percent. Networks with less than one station measured by a network of the seven urban per 2 square miles were not thought to be con- stations. No significant difference occurred be¬ sistently reliable for estimating monthly, sea¬ tween the annual, seasonal, and most monthly sonal, or annual mean 24-hour soiling indexes, means measured by the two networks, and their particularly to support epidemiologic studies. 95 percent confidence intervals were in close The seasonal cumulative frequency distribu¬ agreement. Even the daily mean estimates of tions of soiling-index measurements closely the two networks were reasonably close. resembled those of monthly sulfation measure¬ Annual, seasonal, and most monthly mean ments in that both fall and spring distributions levels of particulate matter were estimated al- corresponded with the annual distribution. If most as well by the use of several highly re¬ it were necessary to describe the annual mean duced, sampling-frequency schedules. The soiling index and annual frequency distribu¬ most extreme reduction in number of stations tions by limited sampling, fall or preferably and sampling frequencies that continued to give spring would be the time to conduct such a good estimates of annual and seasonal means program. was the combination of an elevated central Reducing either the number of stations or the urban station and one suburban station oper- frequency of sampling resulted in only fair ating only 22 days a year. The annual mean estimates of monthly, seasonal, and annual measurement obtained with this reduced sam¬ 24-hour soiling-index means. pling program differed by only 8 percent and m. "Two- and Four-Hour Soiling Index" the seasonal mean by a maximum of 14 percent (10). At least two automatic spot sampling VoL 82, No. 1, Janaary 1967 21 stations per square mile (about 0.87 mile be¬ From October through April significant dif- tween stations) would be required to estimate ferences in most monthly mean 4-hour soiling- the 24-hour mean soiling index in this commu- index values were noted; however, from May nity with good assurance of ±20 percent accu- through September, insignificant differences racy. For the seven urban stations used in the were found between the diurnal means. Di¬ study, the true daily mean (±30 percent) was urnal sampling during this 5-month period was estimated with good assurance on about 18 per¬ therefore usually unjustified insofar as soiling- cent of the days. index measurements were concerned in the com¬ From four to six stations per square mile munity studied. would be required to estimate mean urban soil¬ IV. "2- and 24-Hour Sulfur Dioxide and ing for any single 2- or 4-hour period with good Summary of Other Pollutants" (11). At least assurance of ±20 percent accuracy. At least 245 stations or 4 stations per square mile one station per square mile (1.22 miles between (spaced about 0.62 mile apart) would be re¬ stations) would be needed to determine the quired in this community to estimate most daily monthly mean of any single 2- or 4-hour period mean levels of sulfur dioxide with 95 percent with ±20 percent accuracy. One station per 3 assurance of ±20 percent accuracy. With the square miles (2.1 miles between stations) would 32-station network (1 station per 2 square be required for diurnal seasonal means, and miles) used in this study, daily means were one station per 6 square miles (3.0 miles between estimated with ±20 percent accuracy on 51 per¬ stations) for diurnal annual means. cent of the days of the 1-year study, but only on The true 2-hour mean (±20 percent accu¬ 2 percent of the days during the winter season. racy) was estimated for the seven-station urban There was, however, assurance of ±50 percent network used in this study only 6.2 percent of accuracy for the 32-station network on 85 per¬ the time. Estimated diurnal (2-hour) monthly cent of the days throughout the year. About means usually had accuracy limits ranging from 60 stations (1 per square mile) were required 20 to 50 percent; diurnal seasonal means, from for estimating the winter seasonal and monthly 20 to 40 percent; and the diurnal annual mean, means during the heating season (October ±22 percent. through March), but only 1 or 2 stations were Close estimates of mean monthly and seasonal required for the entire community during the soiling indexes and cumulative frequency dis¬ nonheating period (May through September). tributions were obtained by combining the About 24 stations (1 per 3 square miles) were sampling data obtained from one central urban required for estimating spring and fall sea¬ and one suburban (south control) station. sonal means. Most 2-hour seasonal mean soiling values ob¬ With a network of 7 uniformly spaced sta¬ tained by this small network were not signifi¬ tions (1 per 9 square miles), estimates of cantly different from those of the full seven- monthly and seasonal sulfur dioxide means in station urban network. most cases were not significantly different from At least 20 percent of the geographic detail in the means obtained by the larger 32-station net¬ 2-hour seasonal mean soiling patterns over the work. With the 7-station network, daily mean community was probably lost as a result of sulfur dioxide estimates usually agreed closely using a 7-station network instead of a 22-sta- with the 32-station network, but with a wider tion network (1 station per 3 square miles). confidence range than the larger network. For Such detail may be significant in various types less precise general monitoring purposes, esti¬ of epidemiologic studies. mates of monthly and seasonal means that were Sampling on a reduced-frequency schedule, usually within 40 percent of the larger network such as every other day, every fourth day, or means were provided by two stations (one weekdays only, gave acceptably close estimates urban station in the high-pollution area and of diurnal monthly, seasonal, and annual one suburban station). means, cumulative frequency distributions, and Annual, seasonal, and most monthly mean geographic distributions for use in most general sulfur dioxide levels were estimated almost as monitoring programs. well by several reduced-frequency sampling 22 Public Health Reports schedules (for example, every second day and at the urban center and in the suburban area weekdays only) as by everyday sampling. only, seasonal mean concentrations at several The seasonal cumulative frequency distribu¬ distances from the center could be approxi¬ tions of daily sulfur dioxide measurements mated by use of the modified Gaussian equation. were similar to the distributions of monthly Larsen and associates found that both sulfur sulfation and daily soiling measurements. The dioxide emission data and sulfation data ap¬ annual frequency distribution was similar to peared to fit a Gaussian or normal curve with fall and spring distributions and could be ap- distance from center of city. The relations are proximated by sampling in fall or spring only. given in "The Radial Distribution of Sulfur The frequency distributions of 24-hour con- Dioxide Source Strength and Concentration in centrations could be closely approximated by Nashville" (12). The variation in emission samplings from smaller networks, such as seven may be expressed as follows: uniformly spaced stations or even by a two- station network. The seasonal distribution E=Ece 2O also be slopes of everyday sampling could where E is emission of on reduced strength (in pounds closely approximated by sampling S02 square mile per day), Ee is emission schedules such as every other day or per frequency strength at the center of Nashville, e is the weekdaysonly. transcendental number 2.718, r is the radial Geographic distributions of seasonal mean sulfur dioxide based distance in miles from the center of Nashville, concentrations, upon daily a new termed "standard from 32 could be and sr is parameter, samples stations, closely ap¬ radial deviation," which is analogous to stand¬ proximated by everyday measurements from 16 In stations station 4 For ard deviation in the "normal" equation. (1 per square miles). pro- was 1.2 or about one-fourth data to health-effect Nashville sr miles, viding support studies, season and for sulfur dioxide at the the radius of Nashville, for each everyday sampling the total least number of stations possible was the most year. economical for four stations A similar equation may be used to express approach; example, sulfation in Nashville: per square mile for daily measurements and at least one station per 2 square miles for deter- S=S>+Sce .2^(-Y mination of monthly and seasonal mean values (preferably one station per square mile during where S is sulfation, S* is the background sulfa¬ heating season). tion 7.5 miles from the center of town, and Se The number of stations for measur- is the sulfation level in the center of town required from ing mean 2-hour sulfur dioxide concentrations (.Sj>). (Se is the measured sulfation with ±20 percent accuracy would be the same which is substracted the background value £&.) as for 24-hour sulfur dioxide or for 2-hour-soil- For sulfation, sr varied from 0.2 of the radius of ingquality determinations; that is, four stations Nashville in the summer to 0.4 of the radius of per square mile for daily 2-hour means and at Nashville in the winter. The fall, spring, and least one station per 2 or 3 square miles for annual values were 0.3 of the radius of monthly and seasonal 2-hour means. Cumula¬ Nashville. tive frequency and geographic distributions of The fall, spring, and annual sulfation values 2-hour seasonal sulfur dioxide means could be were of comparable magnitude. Winter sulfa¬ approximated by samples taken from a small tion at the center of Nashville was three times number of stations, on a reduced frequency as great as the annual value; summer sulfation schedule, or in the same manner as with 24-hour was one-fourth the annual value. For each sea¬ sulfur dioxide sampling. son and the total year, sulfation measurements The seasonal mean, 24-hour sulfur dioxide at the center of Nashville were approximately concentrations measured at the urban center seven times the nonurban values. A computer- and at several distances from the center were solved mathematical model relating source found to fit the normal distribution curve fairly strength to sulfation indicated a "fairly good well. Thus from sulfur dioxide measurements fit" to the study findings when sulfation was

VoL 82, No. 1, January 1967 23 considered as being inversely proportional to predictable rate during the sampling cycle. the square of distance between source and The tape used in this study did not show a sig¬ receptor. nificant variation in the flow rate of areas within McCormick and Xintaras observed diurnal the roll. variations of carbon monoxide concentrations An A.I.S.I. sampler calibration procedure and traflic density at "curbside" studies in Nash¬ was developed that was both simpler and more ville. The following findings were presented accurate than previous procedures. in "Variation of Carbon Monoxide Concentra¬ In "Atmospheric Sulfur Dioxide and Partic¬ tions As Related to Sampling Interval, Traflic, ulate Matter: A Comparison of Methods of and Meteorological Factors" (13). Measurements" (16), Stalker and co-workers This investigation showed that in urban areas described the variations of pollution concentra¬ some order could be found in carbon monoxide tions with different methods of measurement. (CO) fluctuation with time, but that a complete Interrelations between sulfur dioxide and par¬ interpretation of such "curbside" data in terms ticulate matter were investigated. Their find¬ of source and meteorological factors was very ings follow: complex. The hourly variation of CO concen- The 2-hour, automatic smoke sampler was a tration with traflic density confirmed a moder- more reliable and much less variable method of ate degree of correlation between these two measuringsoiling quality than the 24-hour, slow factors; however, this limited correlation flow-rate method. For determining ambient emphasized also the importance of local condi¬ S02, the 24-hour standard bubbler method was tions of transport and dispersion on this varia¬ more efficient than the 2-hour sequential sam¬ tion. Marked positive anomalies of the normal pling method used in the Nashville study. CO trends could be associated with wind speed From measurements of certain types of par¬ decay, and therefore a technique was suggested ticulate matter in a community, other types of for evaluating the effects of traflic changes on airborne particulate matter could be reasonably CO levels, with wind speed incorporated as a well estimated. For example, total dustfall contributory element. could bo estimated from measurements of water- The results of single-source field experiments insoluble dustfall, suspended particulate mat¬ validated to some extent the interpretation of ter from either water-insoluble dustfall or total data collected in Nashville. Reasonable first- dustfall, organic or sulfate particulates from approximation estimates of the higher short- total suspended particulate matter, and soiling period concentations during 1-hour sampling quality from suspended particulate matter. In periods are now possible. Further information the Nashville study, these estimates were always on the diurnal variation of the ratios is required, more reliable during the winter than during however, before the general effect of time of other seasons, but were often acceptable in all day on the estimates can be determined. seasons for general air pollution monitoring Park and associates evaluated the method purposes. for measuring soiling index and suggested cali- The relations between results obtained with bration procedures in "Developments in the different methods of measuring S02 were best Use of the A.I.S.I. Automatic Smoke Sampler" during the fall and winter when atmospheric (lli). They found that 11 automatic A.I.S.I. concentrations were highest. During the win¬ smoke samplers were in continuous and simul- ter season of high S02 levels, 1 mg. of S03 per taneous operation as part of a year-long Public 100 square centimeters per day (sulfation rate), Health Service field study in Nashville, Tenn., measured by the peroxide candle method, was and that the heavy load of reading so many equivalent to 0.042 part per million S02 meas¬ tapes prompted the development of an auto¬ ured by the sodium tetrachloromercurate meth¬ matic spot evaluator that could read approxi- od (TCM). During the fall season of moderate mately 12 samples per minute. S02 levels, the 1-mg. equivalent was 0.035 part Tests revealed that the flow rate varied sig¬ per million S02, and during the spring season, nificantly between samplers and between filter the 1-mg. equivalent decreased to only 0.015 tape rolls, and that the flow rate decreased at a part per million S02. On the average, the 24 Public Health Reports Thomas autometer measurements of S02 were Reference 17 is not outlined in this paper. about twice as high as simultaneous 24-hour Sampling, analytical, and standardization TCM measurements of SO* procedures used to measure sulfur dioxide were Atmospheric concentrations of suspended or- presented in "Developments in the Measurement ganic particulate matter over urban and sub¬ of Atmospheric SulfurDioxide" (18),by Welch urban areas apparently could be estimated with and Terry. reasonable confidence from the measurements The TCM method for collection and analysis of volumetric S02 during the fall, winter, and of low concentrations of sulfur dioxide was spring seasons in coal-burning communities. found to be accurate, highly sensitive, flexible, The seasonal regression slopes or equivalents and practical in this study. The adaptation of differed, particularly the slope for the fall sea¬ this method for use with an automatic analyt¬ son when the measured SOz was equivalent to ical instrument, and methods of rapid midget about twice as much organic particulate mat¬ bubbler handling, washing, and filling were ter as during winter or spring. Suspended sul- described. fate particulate matter was reasonably well as¬ The relation between this modified TCM sociated with volumetric S02 during the winter method and two other methods was investigated season only (r=0.70). by using synthetic mixtures of sulfur dioxide in Volumetric S02 and atmospheric soilingqual¬ air. A dynamic calibration procedure showed ity were significantly related during the winter that the Thomas autometer and TCM results season only (r=0.70); the relation, however, were identical. The relation between thehydro- varied geographically (from station to station) gen peroxide and TCM methods was found to and temporally (from month to month). be (H202 results in ppm) = 0.72 X (TCM re¬ Many S02 and particulate matter relations sults ppm) +0.04. might be reliable enough to use for general air The lead peroxide candle method used in pollution monitoring. Before the measure- Nashville for the estimation of sulfur dioxide ment of one pollutant could be reliably used for also was described. predicting the concentration of another type of pollutant, however, these relations would have Meteorologic Findings to be investigated in several different commu¬ The data in Baulch's, "Relation of Gustiness nities. to Sulfur Dioxide Concentration" (19) showed A detailed method for inventorying sulfur significant relations between the two most com¬ dioxide emissions by square-mile areas was mon gustiness classes (Brookhaven types B and described by Stalker and co-workers in "Nash¬ D) and S02 concentration. Although the ville Sulfur Dioxide Emission Inventory and simplicity of this method may be criticized from the Relationship of Emission to Measured Sul¬ a theoretical point of view, the Brookhaven fur Dioxide" (16). The emission inventory classification of gustiness provides a useful supplied considerable useful information for index of turbulence for this type of analysis. (a) control purposes, (b) development of a Further application and refinement of the mathematical diffusion-ambient sulfur dioxide method will be necessary, however, before it will prediction equation, and (c) demonstration of be satisfactory for use in determining atanos- a simple regression relation between average pheric diffusion on a routine basis. emission and average ambient concentratioiis of Effective application of this method of analy¬ sulfur dioxide. sis could provide a useful tool for the develop¬ In Nashville, a square-mile area with an aver¬ ment of a diffusion climatology for communities age winter-season emission of 1 ton of sulfur of all sizes. The economic advantages of utiliz- dioxide per day showed an atmospheric sulfur ing gustiness data in addition to standard dioxide concentration of 0.022 ppm, an area in meteorological observations would make such which 5 tons per day were emitted showed a a method especially valuable to smaller com¬ concentration of 0.042 ppm, and one in which munities for which the cost of more extensive 10 tons of sulfur dioxide per day were emitted diffusion research has been prohibitive. showed a concentration of 0.067 ppm. A method of forecasting air pollution levels

VoL 82, No. 1, January 1967 25 was discussed by Boettger and Smith in "The height, and distances between trees. These find¬ Nashville Daily Air Pollution Forecast" (20). ings were presented in "A Study of the Effect of Their findings and recommendations follow. Tree Leaves on Wind Movement" (21). The tested meteorological criteria were shown In his paper "On the Representativeness of to be good indicators of three air pollution Surface Wind Observations Using Data from levels during the winter season, but could dis- Nashville, Tennessee" (22), Frederick discussed criminate between only two levels during the the mean wind speed as applied to air pollution spring and summer. Of three methods of meas¬ sampling. He found that carefully designed uring air quality, only the total particulate minimum-sized wind networks yielded reason- levels were high enough to be considered as a ably reliable data for studies of urban airflow year-round air pollution indicator. Sulfur and ventilation. On the scale of the "Nashville dioxide and soiling-index measurements were Community Air Pollution Study," the correla¬ very low during the summer season and, when tion of wind measurements was independent of considered alone, implied the absence of com¬ any parameter readily available or obtainable munity air pollution problems during that for other cities. It appeared to be dependent season. upon micro-environmental exposure. While the forecasting results were not out- The Weather Bureau Airport Station's hourly standing, the skill could be improved with wind observations at Nashville were highly cor- further experience. More elaborate meteorolog¬ related with the average airflow over the city. ical criteria could be used that might give better Measurements of the urban area wind at Nash¬ discrimination among the levels, but such a ville averaged 60 to 70 percent of the suburban method would lack the forthrightness of the wind as measured at the airport. one used. Other straightforward criteria could Some generalized statements were possible be devised and tested with the same air quality about the reliability of wind direction observa¬ data and associated meteorological conditions. tions at the Weather Bureau Airport Station. Various weather parameters that could be in¬ Pooler used an empirical diffusion equation, vestigated were average wind speed and direc- with published summaries of wind direction and tion and the presence or absence of an upper speed frequencies obtained at the Nashville inversion. An additional influential parameter Weather Bureau Station, to compute patterns could be the duration of low wind speeds or of mean monthly relative concentrations. He interruption of these low speeds by periods of discussed the following findings in "A Predic- significant wind speed or both. tion Model of Mean Urban Pollution For Use This program demonstrated that objective With Standard Wind Roses" (23): forecast methods could be used by local Weather Hillside locations facing the sources showed Bureau oflices to forecast daily air pollution generally higher than predicted values, and levels during the seasons of the year when air valley locations sheltered from the sources by in- pollution concentrations are high and trouble- tervening terrain showed lower values. Higher some. The forecast method was general enough wind speeds appeared to increase the sulfation that it could be adapted to other cities after rate. consideration of climatology, topography, in- Relative monthly emission rates, determined dustrial locations, and the individual needs of from a comparison of predicted and observed the organizations that would use the forecasts. values for the area, showed a close linear rela¬ Frederick found that, in an area of numerous tion to monthly degree-day totals for the 5 but well-spaced, mature deciduous trees or months considered. A dropoff from the linear clumps of trees, after defoliation the wind in- trend could be noted for nonheating and tran- creased by a factor nearly 25 percent over "aver¬ sitional months. age exposure" and as much as 40 percent over In "A Diffusion Model For An Urban Area" exposures in areas nearly devoid of trees, and (24), Turner presented a working model for the slowly decreased by the same amount as the diffusion of from sources in an leaves gases multiple emerged and matured. Values varied urban area. He found that the model, while somewhat according to number of trees, tree giving encouraging results, generally tended to 26 Public Health Reports overcalculate, especially downwind of major knowledge of conversion rates of sulfur dioxide sources. Calculated gaseous concentrations ex- to sulfur trioxide, sulfuric acid droplets, and ceeded observed concentrations more frequently sulfates was limited. He thought improvement (28 percent) than observed concentrations ex- could possibly be achieved by more direct meas¬ ceeded calculated concentrations (14 percent) urements ofstability than were possible in Nash¬ for 32 stations for 35 periods. ville. Increasing the complexity of either the Undercalculation upwind of themajor sources diffusion model or the source inventory did not also was noted. It probably resulted from the appear to be warranted for this time-and-space fact that pollutant concentrations at the center scale (24 hours with several kilometers between of a square-mile area received no contributions sampling stations) until air pollutant field from sources within that area but only from measurements with lowerthresholds and greater sources in other areas. accuracy could be routinely obtained. Despite the overcalculation, 58 percent of all Another study by Turner on the "Relation- calculated concentrations were within ± 1 part ships Between 24-Hour Mean Air Quality per hundred million (pphm) of the observed Measurements and Meteorological Factors in concentrations for the thirty-five 24-hour Nashville, Tennes9ee" (25) showed that meteor¬ periods randomly chosen from the winter half ological variables of temperature, wind speed, of the year. When zero values of both calcu¬ and stability were well related to 24-hour city- lated and observed concentrations were ex- wide sulfur dioxide concentrations and soiling cluded, 70 percent of the calculated values were indexes. Sulfur dioxide concentrations and within a factor of two of the observed values. soiling indexes varied directly with degree-day The model results were better than constant- value and 24-hour mean stability, and inversely concentration estimates in four of the five statis¬ with 24-hour mean wind speed. This study tics used for comparison. The model had a illustrated the magnitude of the variability of skill score of 0.38 (significant at the 0.1 percent daily citywide air-quality levels, which might level) in estimating three classes of concentra¬ be explained by meteorological parameters that tion: threshold, 0 to 1 pphm; low, 2 to 10 were routinely measured or obtained objectively pphm; and high, more than 10 pphm. These from routine measurements. results were especially encouraging in view of the following limitations of the model. Miscellaneous Hndings 1. Area sources were represented as Gaussian- Morbidity and mortality data were collected distributed crosswind line sources. by personal interview from a sample of 3,000 2. Aconstanteffectiveheightof 20meterswas households in Nashville and certain surround- assumed. ing areas. Finkner and co-workers described 3. Topographic variations were ignored. the design, conduct, and results of the Nashville 4. Meteorological elements were given no spa- morbidity survey in their paper, "Nashville tial variation. Morbidity Survey: Design of a Household Sur¬ 5. The temporal variation of meteorological vey For Air Pollution Research"(26). elements was by 2-hour periods only. Kftilin used data from the morbidity survey 6. Emissions occurring during a 2-hour for his "A Use of the Information Statistic as period reached receptorsduringthe same 2-hour a Measure of Conformity in Comparing Two period regardless of travel time. Sets of Responses" (27). He found that the Although the model indicated the levels of use of the basic information theory provided concentrations andthe general area extent ofthe an easy and interpretational method for analyz- pollution, certain sources of error were pointed ing two sets of responses to the same questions. out. From this experience, improvement was In addition it had the advantage of being di¬ expected in future attempts at devising models. rectly applicable to useful tests of significance. The author planned to test a modified version Smith and associates presented the following of this diffusion model in another urban area conclusions in "Public Reaction to Air Pollu¬ and if possible with other pollutants, since tion in Nashville, Tennessee" (28).

VoL 82, No. 1, January 1967 27 Responses to questions put to people in Nash- provided air quality data for use in the medical ville indicated an increasing awareness of and studies as well as in determining the representa- concern aibout air pollution as pollution levels tiveness of air quality data with respect to num- in the respondents' neighborhoods increased. ber and location of sampling stations, sampling Air pollution levels were not the only factor frequency, and sampling time, and evaluating influencing people's awareness of and concern analytical methods, equipment, and procedures. about air pollution. Women were more con- These findings, plus an emission inventory and cerned than men, and socioeconomic factors in- a discussion of the diurnal variation of carbon fluenced awareness and concern. monoxide concentrations, have been outlined. Up to 3.8 percent of the respondents in the Meteorological factors were studied in rela- sample population were sufficiently aware of tion to air quality. A method of forecasting and concerned about air pollution as a health air pollution levels and a discussion of a mathe- problem that they so expressed themselves of matical diffusion model have been presented their own volition. An extension of the sample along with the meteorological findings. population to the total population of the Nash- The findings from miscellaneous studies on ville area in 1958 indicated that somewhat less design of morbidity surveys and opinions of than 10,000 people were concerned about air pol- people indicating their awareness and concern lution as a health problem. about air pollution have been presented. Twenty-three percent of the respondents A bibliography of papers on the, Nashville when asked the direct question said they were study was included. bothered by smog in one way or another. An extension of the sample to the total population REFERENCES indicated that about 50,000 people in the Nash- (1) Zeidberg, L. D., et al.: Air pollution and health: ville area were bothered smog. In to General description of a study in Nashville, by reply Tennessee. J Air Pollut Contr Assoc 11: 289- direct questions, 18 to 51 percent of the 294, June 1961. respondents said they were bothered by some (2) Zeidberg, L. D., Prindle, R. A., and Landau, E.: specific nonhealth effect of smog. Again, an The Nashville air pollution study. I. Sulfur extension of the data indicated that 40,000 to dioxide and bronchial asthma; a preliminary more in report. Amer Rev Resp Dis 84: 489-503, Oc- than 100,000 people the survey area tober 1961. were bothered by the soiling of surfaces and (3) Zeidberg, L. D., and Prindle, R. A.: The Nashville objec.ts, decreased visibility, odors, and damage air pollution study. II. Pulmonary anthra- to property because of air pollution. cosis as an index of air pollution. Amer J The data indicated that the people's aware- Public Health 53: 185-199, February 1963. ness of and concern about air pollution were (4) Zeidberg, L. D., Prindle, R. A., and Laudau, E.: The Nashville air pollution study. III. Mor- more influenced by the frequency of days of bidity in relation to air pollution. Amer J unusually high pollution than by the high Public Health 54: 85-97, January 1964. monthly, seasonal, or annual average levels of (5) Landau, E., Prindle, R. A., and Zeidberg, L. D.: pollution. The Nashville air pollution study. IV. Sulfur dioxide and bronchial asthma; a multivariate Summary analysis. Presented at the 90th annual meet- ing, American Public Health Association, Miami An extensive air pollution study was con- Beach, Fla., Oct. 17, 1962. ducted in Nashville, Tenn., to investigate pos- (6) Zeidberg, L. D., Horton, R. J. M., and Landau, E.: sible relations between air pollution and health The Nashville air pollution study. V. Mortal- and to gather information on a number of medi- ity from diseases of the respiratory system in cal, engineering, meteorological and other relation to air pollution. Presented at the annual meeting, National Tuberculosis Associa- phases of air pollution. tion, Denver, Colo., May 1963. Bronchial asthma, anthracosis, morbidity, (7) Zeidberg, L. D., Horton, R. J. M., and Landau, E.: respiratory disease mortality, and cardiovascu- The Nashville air pollution study. VI. Cardio- lar disease mortality were studied in relation to vascular disease mortality in relation to air air pollution. Presented at the annual meeting, pollution. California Air Pollution Medical Research More than 200,000 aerometric observations Conference, Los Angeles, February 1964.

28 Public Health Reports (8) Keagy, D. IL, et aL: Sampng station and time stract.] Int J Air Water Po4lut 7: 989-990 requirements for urban air pollution sureys. (1&). L Iead peroxide candles and dstfall colletors (18) Welch, A. ., and Terry, J. P.: Developments in J Air Pollut Contr Am 11: Z70-28 Jume 19. the measurement of atmospheric sulfur dioxide. (9) Slker, W. W, and Dickersn, R. C.: mpling Amer Industr Hyg Assoc J 21: 316-21, Au- station and time ru e for urban air gust L9W. pollution surveys, IL Suspeded particulate (19) Baulch, D. M.: Relation of gustiness to sulfur di- matter and solung index J Air Pollut Contr oxide concentration. J Air Pollut Contr Assoc Asso 12: 111-128, March 1962. 12: 539-52, November L9L (10) Stalker, W. W, and Dickerson, B. C.: Smving (20) Boettger, C. M., and Smith, ]L J.: The Nashville station and tim uirments for urban air pol- daily air pollution forecast Monthly Weather lution surveys. IIL Two- and four-hour soiling Rev 89: 479-481, November 1961 index. J Air Pollut Contr Assoe 12: 170-178, (21) Frederick, EL H.: A study of the effects of tree Apri 1962. leves on wind movement. Monthly Weather (11) Stalker, W. W., Dickerson, R. C., and Kramer, Rev 89: 39-44, February 196L G. D.: mpling station and time rquirents (22) Frederick, R.]L: On the enativeness of for urban air pollution surveys. IV. 2- and 24- surfae wind observations using data from hour sulfur dioxide and summary of other pol- Nashville, Tennessee. Int J Air Water Pollut lnts. J Air Pollut Contr Asoc 12: 361-375, 8: 11-19, January 1964. Augut1962 (23) Pooler, F., Jr.: A prediction model of mean urban (12) Larsen, R. L, Stalker, W. W., and Claydon, C. R.: pollution for use with nda wind roses. The radial distribuion of sulfur dioxide source Iht J Air Water Pollut 4: 199-21 September srength and concentration in Nashville. J Air 196L Pollut Contr Asoc 11: 52-534, November 1961. (24) Turner, D. B.: A diffusion model for an urban (13) McCormick, B. A., and Xintaras, C. J.: Varition areaL J Appl Meteorology 3: 83-91, February of carbon monoxide concentrations as reiated to 1964. sampln interval, traffic, and meteorological (25) Turner, D. B.: Relationships between 24-hour factors. J Appl Meteorology 1: 237-243, June mean air quality m urements and meteoro- logical factors in Nashville, Tennessee J Air (14) Parrk, J. C., Keagy, D. ML, and Stalker, W. W.: Pollut Contr Assoe 12: 539-542, November 1962 Developments in the use of the A.LS.L auto- (26) Flnkner, A. I, Monroe, J., and Fleischer, J.: matic smoke sampler. J Air Pollut Contr Assoc Nashville morbidity survey: Design of a house- 10: ilS-W6, August 19M. hold survey for air pollution research. Insti- (15) Salker, W.. W, Dickerson, R. C., and mer, tute of Statistics Memo Series No. 252. Re- G. D.: Atmospheric sulfur dioxide and particu- eh Triale I t, Durham, N.C., April late matter: A comparison of methods of mea- 1960. surem Amer Indstr Ebyg Assoc J 24: 68- (27) Keilin, J. E.: A use of the information satistic 79, January-February 196. as a measure of conformity in comparing two (16) Stalkr, W. W, et aL: Nashville sulfur dioxide sets of responses. Presnted at the annual emission inventory and the relationship of emis- meeting, American Statiical Society, Stanford, sion to measured sulfur dioxide. J Air Pollut Calit, 1960. Contr Assoc 14: 467-474, November 1964. (28) Smith, W. S., Schunan, J. J., and Zeidberg, (1.) S3talker, W. W., and Dickerson, B. C.: Sampling L. D.: Publi reaction to air pollution in Nash- station and sampling frequency requirements ville Tens J Air Pollut Contr Assc 14: for commmunity air pollution studies. [Ab- 418-42, October 1964.

VL U N. L Januay 1967 29