Final Report: Testing of a Model to Predict Human Exposures to Aldehydes Arising from Mobile and Point Sources

EPA Grant Number: R826787
Title: Testing of a Model to Predict Human Exposures to Aldehydes Arising from Mobile and Point Sources
Investigators: Raymer, James H. , Akland, Gerald G. , Clayton, C. Andrew , Johnson, Ted , Michael, L. C. , Pellizzari, Edo D.
Institution: Desert Research Institute , TRJ Environmental Inc.
EPA Project Officer: Chung, Serena
Project Period: October 1, 1998 through September 30, 2002 (Extended to September 30, 2003)
Project Amount: $629,841
RFA: Urban Air Toxics (1998) RFA Text |  Recipients Lists
Research Category: Air Quality and Air Toxics , Air

Objective:

The overall objective of this research project was to estimate human exposure to target aldehydes (formaldehyde, acetaldehyde, acrolein, propionaldehyde, butyraldehyde, crotonaldehyde, glyoxal, methylglyoxal) by means of microenvironmental and personal exposure monitoring for two urban areas. The main hypothesis tested was that a mathematical model can be used to predict personal exposure distribution to aldehydes. Additional hypotheses tested were that: (1) personal exposure levels of aldehydes exceed outdoor concentrations; (2) indoor aldehyde concentrations exceed outdoor concentrations; and (3) the composition of oxygenated fuel results in significant differences in population exposures to aldehydes.

Summary/Accomplishments (Outputs/Outcomes):

This research project consisted of two types of monitoring efforts in both Sacramento, CA, and Milwaukee, WI. The two locations were chosen to capture aldehyde concentrations where either methyl-t-butyl ether ([MTBE], Sacramento) or ethanol (Milwaukee) was used as an oxygenated additive to gasoline. The two types of studies were: (1) a scripted activity study in which 1-hour and 8-hour personal exposures to eight aldehydes (formaldehyde, acetaldehyde, acrolein, propionaldehyde, crotonaldehyde, n-butyraldehyde, glyoxal, and methyl glyoxal) and seven volatile organic compounds (VOCs) (benzene, ethylbenzene, toluene, m,p-xylene, o-xylene, ethanol, and MTBE) were measured by technicians according to scripts that specified a microenvironmental location and activity for each time period; and (2) a personal monitoring effort in which 24-hour personal exposures to eight aldehydes were measured by volunteers as they engaged in typical daily activities. Supplemental data were obtained from indoor and outdoor pollutant monitors at the residences of the volunteers, from ambient pollutant and meteorological monitors at fixed-site locations in each city, from real-time diaries completed by the technicians and volunteers, and from questionnaires completed by the volunteers.

Scripted Study

In the Scripted Activity study, researchers prepared a set of 30 activity scripts (15 for Sacramento and 15 for Milwaukee), each of which listed a sequence of planned exposure events spanning a 24-hour period from 7:00 p.m. to 7:00 p.m. Each exposure event corresponded to a defined 1-hour or 12-hour sampling period. Each script specified start time, duration, district, microenvironment number (ME No), window status, and description of location and activity for each event. The districts were defined as areas surrounding operating fixed-site monitors. The ME No refers to a set of five general microenvironment categories that historically have been of interest to exposure models (indoors-residence, indoors-other, outdoors-near road, outdoors-other, and in vehicle). The scripts were constructed to provide a variety of specific microenvironments belonging to each of these general categories.

Guided by the script, a technician used an instrument cart to measure concentrations of 15 pollutant species during each exposure event. The VOC species were measured using an active pump and a multisorbent trap. The aldehydes were measured using an active pump and 2,4-dinitrophenylhydrazine silica cartridges. One-hour concentration data were collected between 7:00 a.m. and 7:00 p.m. to provide higher time resolution during periods when people tend to change activities more frequently; 12-hour concentration data were collected between 7:00 p.m. and 7:00 a.m. Sampling began and ended on clock hours to facilitate comparisons with hourly data collected at various fixed-site monitors in Sacramento and Milwaukee.

The technician recorded special conditions associated with each activity in a real-time diary. Supplemental data on ambient pollutant levels (measured by fixed-site monitors), meteorological parameters, and other potential explanatory factors also were collected.

Personal Monitoring Study

Each of the volunteer subjects carried a personal exposure monitor that collected a 24-hour air sample while the subject engaged in his or her typical activities. Concurrent 24-hour samples were collected inside and immediately outside of each subject's residence. These samples were analyzed for eight chemicals. Supplemental data were obtained from ambient pollutant and meteorological monitors at fixed-site locations in each city, from real-time diaries completed by each subject, and from two questionnaires completed by each subject.

Modeling/Statistical Analysis

Modeling. Using the data from the scripted study, a model was created that utilized knowledge of both the duration of participants' activities and the microenvironments in which they conducted those activities. This model was used to calculate the average integrated air concentration of aldehydes that would be anticipated to be measured in personal air.

These values were calculated by the following equation:

in which Ejk is the estimated average exposure of subject k to pollutant j, Ti is the number of minutes spent in microenvironment i, and Mij is the average concentration of pollutant j estimated to occur in microenvironment i. The summations are performed over all microenvironments visited by a subject.

The model ignores the variability of microenvironmetal concentration with time and omits most of the predictor variables identified by the statistical analyses discussed in this report. However, the model does account for the total time spent by the subject in each microenvironment and provides estimates of the microenvironmental concentrations based on pollutant measurements made at the subject’s residence and at nonresidential locations within the applicable city.

To evaluate the performance of the model, we performed a series of linear regression analyses in which the dependent variable was the measured personal exposure for a particular subject and the independent variables included associated modeled exposure, indoor residential concentration, and outdoor residential concentration. Formaldehyde and acetaldehyde, the two most frequently detected aldehydes, were evaluated. In general, the largest R2 values were obtained when the independent variable was modeled exposure, and the smallest R2 values were obtained when the independent variable was outdoor residential concentration. Overall, the results suggest that the model provides a better estimate of exposure than does the measured indoor and outdoor concentrations at the subject's residence.

Statistical Analyses

A series of statistical analyses were performed with the principal goals of: (1) identifying significant predictors of the aldehyde, glyoxal, and VOC concentrations measured during the scripted studies; and (2) identifying significant predictors of the aldehyde and glyoxal concentrations measured during the personal monitoring studies. Depending on the study and location, the candidate predictor variables included other pollutant concentrations measured by personal and fixed-site monitors, meteorological parameters, geographic location, microenvironment, activity, special conditions reported in the diary, and special conditions reported in the monitoring period questionnaire. Because of the large number of candidate variables to be considered in these analyses, we elected to use stepwise linear regression (SLR) as our principal tool for identifying significant predictors.

Microenvironmental Results

To compare microenvironments across the two study locations, the sum of all of the aldehydes and of the VOCs in each microenvironmental sample was calculated to represent a "total" aldehyde or VOC amount. These total values were analyzed to permit the evaluation of those microenvironments that contribute the greatest to the total exposures. Total aldehyde concentrations as a function of microenvironment for Sacramento and Milwaukee are shown in Figures 1 and 2, respectively. In each figure, the number at the top of each bar represents the number of measurable values across the microenvironmental setting. The locations with the highest median total aldehyde concentrations in Sacramento were indoors at a residence (median ~ 98 µg/m3), indoors at a restaurant (median ~ 60 µg/m3), and indoors at a grocery store (median ~ 50 µg/m3). In Milwaukee, the highest median concentrations were measured indoors at a grocery store (~ 43 µg/m3), indoors in a public building (~ 36 µg/m3), and indoors at a restaurant (~ 25 µg/m3).

Figure 1. Total Aldehyde Concentrations as a Function of Microenvironment. Numbers above the bars indicate the number of measurable values across all microenvironments of that type.

Figure 2. Total Aldehyde Concentrations as a Function of Microenvironment. Numbers above the bars indicate the number of measurable values across all microenvironments of that type.

A comparison of analytes for both cites in a high-exposure microenvironment (restaurant) and a much lower exposure microenvironment (outdoors within 10 yards of a street) showed similar aldehydes in both cites, yet much higher concentrations of acetaldehyde measured in Sacramento restaurants as compared to those in Milwaukee; acetaldehyde and formaldehyde are responsible for the majority of the total concentrations. Acetaldehyde is the initial metabolite of ethanol and likely is exhaled by individuals drinking alcohol. It also is known that various aldehydes are produced as a result of deep-frying foods or cooking foods in hot oil, as in stir fry. The data also showed that aldehydes were higher in Sacramento in outdoor locations near a street than in the same microenvironments in Milwaukee.

The VOC concentrations were much higher in Sacramento than in Milwaukee. In Sacramento, the highest median values were measured indoors at a restaurant (~ 580 ng/L), outdoors at a service station (~ 760 ng/L), and indoors at a residence (~ 440 ng/L). Maximum values of 9,100 ng/L and 2,900 ng/L were measured indoors at a restaurant and outdoors within 10 yards of a street, respectively. In Milwaukee, the highest median concentrations were measured indoors at a public building (~ 130 ng/L), although there were only two of these microenvironments, indoors at a grocery store (~ 95 ng/L) and indoors at a restaurant (~ 40 ng/L).

A breakdown of the VOCs measured in each city for the restaurant microenvironment and the outdoor microenvironment within 10 yards of a street showed that in the restaurant, ethanol was the predominant VOC and was probably the result of the consumption of alcoholic beverages. MTBE concentrations were much higher in Sacramento and are consistent with the use of gasoline containing MTBE, which was in use at the time of this study. In the outdoor microenvironment near a street, ethanol and MTBE were the main VOCs from the Sacramento study. Although MTBE would be expected, the reason for the large amount of ethanol is unclear. Except for the one measurement of MTBE in Milwaukee, all of the other VOCs are unremarkable.

Scripted Study Sequential Linear Regression Results

Correlations Among the Pollutants Measured. The data indicated that aldehydes are more likely to cluster together in Milwaukee, where ethanol is used, relative to Sacramento, where MTBE was in use. In Sacramento, the data suggested that atmospheric transformation of benzene and toluene can lead to the formation of the glyoxals, as indicated in the literature. SLR results for the Milwaukee aldehyde/glyoxal species revealed no predictor variables identified for acrolein, glyoxal, and methyl glyoxal.

Relationships to Meteorological Factors. Parameters relating to meteorology included in the master database developed for the scripted studies were evaluated. Parameters include variables relating to temperature, relative humidity, wind speed, wind direction, cloud cover, haze, and precipitation. For Sacramento, the best performing predictors included the reciprocal of wind speed (IWNDSP) for formaldehyde, crotonaldehyde, n-butyraldehde, benzene, ethylbenzene, toluene, m,p-xylene, and MTBE; wind from the southwest (WINDSW) for propionaldehyde; wind from the northwest (WINDNW) for glyoxal; and windspeed (WNDSP) for ethanol. This suggests that low winds favor formation (or lack of dispersion) of these aldehydes. In addition, glyoxal in Sacramento could be influenced by VOCs flowing from San Francisco. The best performing predictors for Milwaukee included WINDNW for formaldehyde, crotonaldehyde, and toluene; wind from the southeast (WINDSE) for acrolein; and wind from the northeast (WINDNE) for methyl glyoxal and MTBE. These results are not consistent with air movement from Chicago (SW of Milwaukee).

Microenvironment. The SLR model was applied to specific microenvironments to evaluate those microenvironments that were the largest contributors to human exposure. In most cases, these variables (microenvironments) were based on the entries made under LOCATION in the diary carried by the technician. Microenvironmental location tended to be a better predictor of pollutant concentration in Sacramento than meteorology; 6 of the 15 R2 values exceeded 0.5. For Milwaukee, 7 of the 15 R2 values exceeded 0.5, with n-butyraldehyde yielding the largest R2 value of 0.9205.

Personal Monitoring Studies

Modeling. In the personal studies, VOCs were not monitored. To evaluate the performance of the model for formaldehyde and acetaldehyde, a series of linear regression analyses were performed in which the dependent variable was the measured personal exposure for a particular subject and the independent variables included associated modeled exposure, indoor residential concentration, and outdoor residential concentration. In general, the largest R2 values were obtained when the independent variable was modeled exposure, and the smallest R2 values were obtained when the independent variable was outdoor residential concentration. Overall, the results suggest that the model provides a better estimate of exposure than the measured indoor and outdoor concentrations at the subject's residence.

In cases in which modeled exposures exceeded measured exposures, we often found that the subject's diary and questionnaire entries indicated extended periods when conditions favored low exposures. Examples include subjects who never came in contact with smokers, subjects who spent extended periods at indoor locations with open windows, a subject who worked in a hospital, and a subject who spent much of the monitoring period on a fishing trip. In some cases, the measured exposure was significantly less than the corresponding concentration measured by the in-residence fixed-site monitor, although the subject's diary indicated he or she spent the majority of the monitoring period inside the residence. Such discrepancies suggest that the fixed-site monitor in the residence was not sited in close proximity to the subject during much of the monitoring period, and that significant concentration variations can occur within the home environment.

Relationships Between Pollutant Concentrations Measured by Personal Exposure Monitors and Diary Entries. Each subject of the Sacramento and Milwaukee personal monitoring studies was asked to complete an activity diary covering the monitoring period. Analysts determined the number of minutes associated with each diary entry and used these data to develop variables indicating the fraction of a subject’s monitoring period associated with a specific microenvironment, activity, or situation. In calculating these values, analysts omitted data from any diary that contained less than 1,075 minutes of activity data (~ 75 percent of the typical 24-hour monitoring period).

For Sacramento, R2 values were relatively large, ranging from 0.5735 (personal methyl glyoxal) to 0.9964 (personal glyoxal). Best predictors include grass/wood fire burning for personal formaldehyde, window closed for personal acetaldehyde and personal propionaldehyde, indoors in a store for personal acrolein, indoors at a health care facility for personal crotolaldehyde, special condition (very long drive through) for personal n-butyraldehyde, indoors-other location for personal glyoxal, and special condition: cooking with gas for personal methyl glyoxal. Other significant predictors for Sacramento include indoors-dry cleaners, no one is smoking, errands and shopping, window open, special condition: using epoxy, and window status not applicable.

For Milwaukee, no predictors were selected for personal methyl glyoxal. The regression equations for the other seven pollutants have relatively large R2 values, varying from 0.5235 to 0.9951. Best predictors include windows open for personal formaldehyde, windows closed for personal acetaldehyde, others are smoking for personal acrolein, getting ready for bed for crotonaldehyde, "I am smoking" for personal butyraldehyde, and indoors-bar for personal glyoxal. Other significant predictors for Milwaukee include town: Milwaukee, indoors-not specified, special condition: solvent use, and air conditioning on.

The above predictors can provide clues about the true source of the aldehydes. For example, aldehydes are known to be produced by high temperature cooking with oils, as noted for the linkage of personal glyoxal concentrations and cooking with gas. The predictive ability of "windows open" suggests that the source of formaldehyde is the outdoor air and "windows closed" suggests that acetaldehyde and propionaldehyde arise from an indoor or personal source.

Conclusions:

The various microenvironments that people encounter on a daily basis provide for widely varying exposures to aldehydes and VOCs. The activities that occur in those microenvironments can modulate the aldehyde concentrations dramatically, especially for environments like "indoor at home." By taking personal activity, location (microenvironment), duration in the microenvironment, and a knowledge of the general concentrations of aldehydes in the various micronenvironments into account, a model can do a reasonably good job of predicting the time-averaged personal exposures to aldehydes. Consistent with many earlier studies, personal exposures are difficult to predict using data from regional monitors. The mobility and activities of people have a far greater influence on their personal exposures.

Meteorological conditions help to explain concentrations for some of the aldehydes and VOCs to which people are exposed. There is some evidence that air downwind of an urban VOC source, such as San Francisco for Sacramento and Chicago for Milwaukee, can contribute to aldehyde exposures in the downwind cities. There were insufficient data, however, to demonstrate that conclusively.


Journal Articles on this Report : 1 Displayed | Download in RIS Format

Other project views: All 5 publications 1 publications in selected types All 1 journal articles
Type Citation Project Document Sources
Journal Article Raymer JH, Akland G, Johnson TR, Long T, Michael L, Cauble L, McCombs M. Microenvironmental characteristics important for personal exposures to aldehydes in Sacramento, CA, and Milwaukee, WI. Atmospheric Environment 2009;43(25):3910-3917. R826787 (Final)
  • Full-text: ScienceDirect-Full Text HTML
    Exit
  • Abstract: ScienceDirect-Abstract
    Exit
  • Other: ScienceDirect-Full Text PDF
    Exit
  • Supplemental Keywords:

    exposure, exposure and effects, health effects, human exposure, air, mobile sources, monitoring, aldehydes, volatile organic compound, VOC, modeling, air quality models, predictive model, personal exposure distribution, aldehydes, acetaldehyde, acrolein, methyl-t-butyl ether, MTBE, benzene, formaldehyde, toluene, xylene, urban air pollution, mobile sources, air toxics., Health, Air, Toxics, air toxics, HAPS, VOCs, Risk Assessments, mobile sources, 33/50, health effects, urban air toxics, Methyl tert butyl ether, exposure and effects, air pollutants, aldehydes, Toluene, air quality models, Acetaldehyde, Acrolein, Xylenes, modeling, benzene, human exposure, predictive model, personal exposure distribution, hazardous air pollutants (HAPs), urban air pollution, Volatile Organic Compounds (VOCs), Benzene (including benzene from gasoline), Xylenes (isomers and mixture)

    Relevant Websites:

    http://www.rti.org Exit

    Progress and Final Reports:

    Original Abstract
  • 1999 Progress Report
  • 2000 Progress Report
  • 2001 Progress Report
  • 2002