Trends in Meteorologically Adjusted Ozone Concentrations in Six Kentucky Metro Areas, 1998–2002

Abstract

Meteorologically adjusted ozone (O₃) concentrations during five recent O₃ seasons (1998-2002) were computed for six Kentucky metro areas using a nonlinear regression model originally developed for forecasting ground-level O₃ concentrations. The meteorological adjustment procedure was based on modifying actual measured O₃ concentrations according to model-predicted responses to climate departures with respect to a reference year. For all six Kentucky metro areas, meteorologically adjusted O₃ concentrations declined over the five-year period. The linear best-fit rate of decline in mean adjusted O₃ concentrations ranged from 0.9 to 2.6 ppb/yr for these metro areas; the average rate of decline was 1.6 ppb/yr. The rates of decline in meteorologically adjusted extreme value (e.g., 95th percentile) concentrations were approximately the same, but there is greater statistical uncertainty concerning the extreme value trends.     

Ambient ozone and crop loss: establishing a cause-effect relationship

This paper provides the results of a retrospective mathematical analysis of the US NCLAN (National Crop Loss Assessment Network) open-top chamber data. Some 77% of the 73 crop harvests examined, showed no statistically significant yield differences between NF (non-filtered open-top chamber) and AA (chamberless, ambient air) treatments (no easily discernable chamber effects on yield). However, among these cases only seven acceptable examples showed statistically significant yield reductions in NF compared to the CF (charcoal filtered open-top chamber) treatment. An examination of the combined or cumulative hourly ambient O3 frequency distribution for cases with yield loss in NF compared to a similar match of cases without yield loss showed that the mean, median and the various percentiles were all higher (>/= 3 X) in the former in contrast to the latter scenario. The combined frequency distribution of hourly O3 concentrations for the cases with yield loss in NF were clearly separated from the corresponding distribution with no yield loss, at O3 concentrations > 49 ppb. Univariate linear regressions between various O3 exposure parameters and per cent yield losses in NF showed that the cumulative frequency of occurrence of O3 concentrations between 50 and 87 ppb was the best predictor (adjusted R2 = 0.712 and p = 0.011). This analysis also showed that the frequency distribution of hourly concentrations up to 87 ppb O3 represented a critical point, since the addition of the frequency distributions of > 87 ppb O3 did not improve the R2 values. In fact as the frequency of hourly O3 concentrations included in the regression approached 50-100 ppb, the R2 value decreased substantially and the p value increased inversely. Further, univariate linear regressions between the frequencies of occurrence of various O3 concentrations between 50 and 90 ppb and: (a) cases with no yield difference in NF and (b) cases with yield increase in NF compared to the CF treatment (positive effect) provided no meaningful statistical relationship (adjusted R2 = 0.000) in either category. These results support the basis that additional evaluation of the frequency of occurrence of hourly O3] concentrations between 50 and 87 ppb for cases with the yield reductions could provide a meaningful ambient O3 standard, objective or guideline for vegetation.     

Accuracy and reliability of an automated air quality forecast system for ozone in seven Kentucky metropolitan areas

An automated forecast system for ozone in seven Kentucky metropolitan areas has been operational since 2004. The forecast system automatically downloads the required input data twice each day, produces next-day forecasts of metro area peak 8-h average ozone concentration using a computer coded hybrid nonlinear regression (NLR) model, and posts the results on a website. The automated models were similar to previous NLR models, first applied to forecasting ozone in the Louisville metro area. The forecast system operated reliably during the 2004 and 2005 O₃ seasons, producing at least one forecast per day better than 99% of the time. The forecast accuracy of the automated system was good. For all 2004 and 2005 forecasts, the mean absolute error was equal to 8.7 ppb, or 15.6% of the overall mean concentration. The overall detection rate of air quality standard exceedences was 56%, and the overall false alarm rate was 42%. In Louisville, the performance of the automated system was comparable to that of expert forecasters using the NLR model as a forecast tool.     

Meteorologically adjusted trends of daily maximum ozone concentrations in Taipei,Taiwan

The meteorological conditions exert large impacts on ozone concentrations, and may mask the long-term trends in ozone concentrations resulting from precursor emissions. Estimation of long-term trends of ozone concentrations due to the changes in precursor emissions is important for corresponding control strategy. Multiple linear regression (method I), multilayer perceptron (MLP) neural network (method II) and Komogorov-Zurbenko (KZ) filter method plus MLP methodology (method III), are used to estimate the meteorologically adjusted long-term trends of daily maximum ozone concentrations by removing the masking effects of meteorological conditions in this study. The daily maximum ozone concentrations and relative meteorological variables were extracted from six air-monitoring stations in Taipei area from 1994 to 2001. The data collected during 1994–2000 period were used as modeling set and utilized to estimate the meteorologically adjusted trends, and the data of 2001 were used as the validation data. The meteorologically adjusted trends of ozone for these three methods were calculated and compared. The results show that both MLP and KZ filter +MLP models are more suitable than multiple linear regression for estimating the long-term trends of ozone in Taipei, Taiwan. The long-term linear trends of meteorologically adjusted ozone concentrations due to the precursor emissions show an increase trend at all stations, and the percent changes per year range from 1.0% to 2.25% during the modeling period in Taipei area.     

Ozone distribution and phytotoxic potential in mixed conifer forests of the San Bernardino Mountains, southern California

In the San Bernardino Mountains of southern California, ozone (O(3)) concentrations have been elevated since the 1950s with peaks reaching 600ppb and summer seasonal averages >100ppb in the 1970s. During that period increased mortality of ponderosa and Jeffrey pines occurred. Between the late 1970s and late1990s, O(3) concentrations decreased with peaks approximately 180ppb and approximately 60ppb seasonal averages. However, since the late 1990s concentrations have not changed. Monitoring during summers of 2002-2006 showed that O(3) concentrations (2-week averages) for individual years were much higher in western sites (58-69ppb) than eastern sites (44-50ppb). Potential O(3) phytotoxicity measured as various exposure indices was very high, reaching SUM00 - 173.5ppmh, SUM60 - 112.7ppmh, W126 - 98.3ppmh, and AOT40 - 75ppmh, representing the highest values reported for mountain areas in North America and Europe.     

Temporal variation in daily concentrations of ozone and acid-related substances at Saturna Island, British Columbia

A multiple linear regression model was used to investigate seasonal and long-term trends in concentrations of ozone (O3) and acid-related substances at the Saturna Island monitoring station in southwestern British Columbia from 1991 to 2000. Statistically significant primary (dominant) cycles with a period of 1 yr were found for O3, sulfur dioxide (SO2), nitric acid (HNO3), and aerosol concentrations of sulfate (SO4(2-)), calcium (Ca2+) and chloride (Cl-). Of these, peak median concentrations occurred during the spring for O3 and Ca2+, during the warmer, drier months (April-September) for SO4(2-) and HNO3, and during the cooler, wetter months (October-March) for SO2 and Cl-. Statistically significant secondary cycles of 6 months duration were seen for concentrations of O3, SO4(2-), HNO3, Ca2+, and Cl-. Daily maximum O3 concentrations exhibited a statistically significant increase over the period of record of 0.33 +/- 0.26 ppb/yr. Statistically significant declines were found for concentrations of SO2, SO4(2-), HNO3, Ca2+, and potassium, ranging from 20 to 36% from levels at the start of the sampling period. Declines in ambient concentrations of SO2, SO4(2-), and HNO3 reflect local declines in anthropogenic emissions of the primary precursors SO2 and NOx over the past decade. Trends in Ca2+ and potassium ion concentrations are in line with a broader North American declining trend in acid-neutralizing cations.     

Long-term meteorologically independent trend analysis of ozone air quality at an urban site in the greater Houston area

The Houston-Galveston-Brazoria (HGB) area of Texas has a history of ozone exceedances and is currently classified under moderate nonattainment status for the 2008 8-hr ozone standard of 75 ppb. The HGB area is characterized by intense solar radiation, high temperature, and high humidity, which influence day-to-day variations in ozone concentrations. Long-term air quality trends independent of meteorological influence need to be constructed for ascertaining the effectiveness of air quality management in this area. The Kolmogorov-Zurbenko (KZ) filter technique, used to separate different scales of motion in a time series, is applied in the current study for maximum daily 8-hr (MDA8) ozone concentrations at an urban site (U.S. Environmental Protection Agency [EPA] Air Quality System [AQS] Site ID: 48-201-0024, Aldine) in the HGB area. This site, located within 10 miles of downtown Houston and the George Bush Intercontinental Airport, was selected for developing long-term meteorologically independent MDA8 ozone trends for the years 1990–2016. Results from this study indicate a consistent decrease in meteorologically independent MDA8 ozone between 2000 and 2016. This pattern could be partially attributed to a reduction in underlying nitrogen oxide (NO_x) emissions, particularly lowering nitrogen dioxide (NO₂) levels, and a decrease in the release of highly reactive volatile organic compounds (HRVOCs). Results also suggest solar radiation to be most strongly correlated to ozone, with temperature being the secondary meteorological control variable. Relative humidity and wind speed have tertiary influence at this site. This study observed that meteorological variability accounts for a high of 61% variability in baseline ozone (low-frequency component, sum of long-term and seasonal components), whereas 64% of the change in long-term MDA8 ozone post 2000 could be attributed to NO_x emission reduction. Long-term MDA8 ozone trend component was estimated to be decreasing at a linear rate of 0.412 ± 0.007 ppb/yr for the years 2000–2016 and 0.155 ± 0.005 ppb/yr for the overall period of 1990–2016.

Implications: The effectiveness of air emission controls can be evaluated by developing long-term air quality trends independent of meteorological influences. The KZ filter technique is a well-established method to separate an air quality time series into short-term, seasonal, and long-term components. This paper applies the KZ filter technique to MDA8 ozone data between 1990 and 2016 at an urban site in the greater Houston area and estimates the variance accounted for by the primary meteorological control variables. Estimates for linear trends of MDA8 ozone are calculated and underlying causes are investigated to provide a guidance for further investigation into air quality management of the greater Houston area.     

An enhanced ozone forecasting model using air mass trajectory analysis

An enhanced ozone forecasting model using nonlinear regression and an air mass trajectory parameter has been developed and field tested. The model performed significantly better in predicting daily maximum 1-h ozone concentrations during a five-year model calibration period (1993–1997) than did a previously reported regression model. This was particularly true on the 28 “high ozone” days ([O₃]>120 ppb) during the period, for which the mean absolute error (MAE) improved from 21.7 to 12.1 ppb. On the 77 days meteorologically conducive to high ozone, the MAE improved from 12.2 to 9.1 ppb, and for all 580 calibration days the MAE improved from 9.5 to 8.35 ppb. The model was field-tested during the 1998 ozone season, and performed about as expected. Using actual meteorological data as input for the ozone predictions, the MAE for the season was 11.0 ppb. For the daily ozone forecasts, which used meteorological forecast data as input, the MAE was 13.4 ppb. The high ozone days were all anticipated by the ozone forecasters when the model was used for next day forecasts.     

Ambient ozone in forests of the Central and Eastern European mountains

Ambient ozone (O(3)) concentrations in the forested areas of the Central and Eastern European (CEE) mountains measured on passive sampler networks and in several locations equipped with active monitors are reviewed. Some areas of the Carpathian Mountains, especially in Romania and parts of Poland, as well as the Sumava and Brdy Mountains in the Czech Republic are characterized by low European background concentrations of the pollutant (summer season means approximately 30 ppb). Other parts of the Carpathians, especially the western part of the range (Slovakia, the Czech Republic and Poland), some of the Eastern (Ukraine) and Southern (Romania) Carpathians and the Jizerske Mountains have high O(3) levels with peak values >100 ppb and seasonal means approximately 50 ppb. Large portions of the CEE mountain forests experience O(3) exposures that are above levels recommended for protection of forest and natural vegetation. Continuation of monitoring efforts with a combination of active monitors and passive samplers is needed for developing risk assessment scenarios for forests and other natural areas of the CEE Region.     

Detecting ozone and demonstrating its phytotoxicity in forested areas of Poland: a pilot study

Ambient concentrations of ozone (O(3)) were measured and O(3) phytotoxicity to tobacco (Nicotiana tabacum L.) was demonstrated in several forest locations in Poland during a pilot study from July-October, 1991. At southern and central locations in Poland, the 24-hour average O(3) concentrations measured with a UV absorption photometer were in the range of 32-55 ppb, and the corresponding 1-hour maxima in the range of 39-83 ppb. At these locations longer period (four to fifteen days) average concentrations were determined using O(3) passive samplers (DGA, Inc.) and were reaching 60 ppb, while at Bialowieza in eastern Poland O(3) concentrations averaged less than 40 ppb. In Szarow, near the Niepolomice Forest in southern Poland, 1-hour O(3) maxima estimated from the data obtained using passive samplers were about 105 ppb in early September. At several locations in southern and central Poland, extensive O(3) injury was determined on O(3)-sensitive Bel W-3 tobacco plants; such injury did not occur in the Bialowieza Forest of eastern Poland. The results of this pilot study indicate that O(3) is present at phytotoxic levels in southern and central Poland.     

Preliminary background ozone concentrations in the mountain and coastal areas of Bulgaria

Urban and non-urban rural ozone (O3) concentrations are high in Bulgaria and often exceed the European AOT40 ecosystem as well as the AOT60 human health standards. This paper presents preliminary estimates to establish background, non-urban O3 concentrations for the southern region of Bulgaria. Ozone concentrations from three distinctly different sites are presented: a mountain site influenced by mountain-valley wind flow; a coastal site influenced by sea-breeze wind flow; and a 1700-m mountain peak site without 'local' wind flow characteristics. The latter offers the best estimate of 46-50 ppb for a background O3 level. The highest non-urban hourly value, 118 ppb, was measured at the mountain-valley site.     

Ambient ozone (O3) in three Class I wilderness areas in the northeastern USA: measurements with Ogawa passive samplers

Ambient ozone (O(3)) was measured in two Class I wilderness areas, one within the White Mountain National Forest in north central New Hampshire (NH) and one within the Green Mountain National Forest in southwestern Vermont (VT), for six weeks between 5 July and 16 August 1994, using Ogawa passive nitrite-coated filter samplers and a one week exposure period each time. Results for 7-day mean ambient O(3) concentrations from the passive samplers were compared to results from a co-located continuous ultraviolet photometric O(3) analyzer both in NH and VT. Although the size of the data set was small, agreement through simple linear regression between mean 7-day O(3) concentrations determined by the continuous monitors, and those obtained from the passive samplers was generally very good (adjusted R(2) = 0.759; p = 0.0003). Overall, excluding the one outlier value, any observed differences in the results obtained by the two methods when comparing co-located passive samplers to the continuous monitor within a given study location, appeared to be solely due to experimental error.     

Application of artificial neural networks to modeling and prediction of ambient ozone concentrations

The deterministic modeling of ambient O3 concentrations is difficult because of the complexity of the atmospheric system in terms of the number of chemical species; the availability of accurate, time-resolved emissions data; and the required rate constants. However, other complex systems have been successfully approximated using artificial neural networks (ANNs). In this paper, ANNs are used to model and predict ambient O3 concentrations based on a limited number of measured hydrocarbon species, NOx compounds, temperature, and radiant energy. In order to examine the utility of these approaches, data from the Coastal Oxidant Assessment for Southeast Texas (COAST) program in Houston, TX, have been used. In this study, 53 hydrocarbon compounds, along with O3, nitrogen oxides, and meteorological data were continuously measured during summer 1993. Steady-state ANN models were developed to examine the ability of these models to predict current O3 concentrations from measured VOC and NOx concentrations. To predict the future concentrations of O3, dynamic models were also explored and were used for extraction of chemical information such as reactivity estimations for the VOC species. The steady-state model produced an approximation of O3 data and demonstrated the functional relationship between O3 and VOC-NOx concentrations. The dynamic models were able to the adequately predict the O3 concentration and behavior of VOC-NOx-O3 system a number of hourly intervals into the future. For 3 hr into the future, O3 concentration could be predicted with a root-mean squared error (RMSE) of 8.21 ppb. Extending the models further in time led to an RMSE of 11.46 ppb for 5-hr-ahead values. This prediction capability could be useful in determining when control actions are needed to maintain measured concentrations within acceptable value ranges.     

Characteristics of urban ground-level ozone in Korea

This study considers the characteristics of ground-level ozone (O3) in five Korean cities over a time period of 6-8 years. The focus of this study is daily maximum 1-hr and 8-hr concentrations. For all the study cities in the period examined, the mean and most of the percentiles (5, 10, 25, 50, 75, 90, and 95) for the daily maximum 1-hr and 8-hr concentrations showed increasing trends, although not all trends were statistically significant. The daily maximum 1-hr and 8-hr concentrations slowly increased during late winter, and peaks were attained during the summer season (from May to September). All the selected cities exhibited a high degree of correlation between their daily maximum 8-hr and 1-hr concentrations. The daily maximum 8-hr concentrations, which were climatologically equivalent to the Korean 1 hr/100 parts per billion (ppb) standard, were higher than the current 8 hr/60 ppb by a difference of 8-16 ppb. Compared with other cities in Korea, Seoul recorded a substantially higher frequency of days and hours with concentrations above 1 hr/100 ppb, and a higher frequency of days with concentrations above 8 hr/60 ppb and 8 hr/80 ppb. Seoul also recorded a substantially higher frequency of hours with concentrations above 1 hr/100 ppb than days with concentrations above 1 hr/100 ppb, implying that on some days severe exceedances persisted for more than one hour per day. During multiple-day episodes a North Pacific High dominated Korea, which is quite typical in Korea during the summer season.     

A comparison of nonlinear regression and neural network models for ground-level ozone forecasting

A hybrid nonlinear regression (NLR) model and a neural network (NN) model, each designed to forecast next-day maximum 1-hr average ground-level O3 concentrations in Louisville, KY, were compared for two O3 seasons--1998 and 1999. The model predictions were compared for the forecast mode, using forecasted meteorological data as input, and for the hindcast mode, using observed meteorological data as input. The two models performed nearly the same in the forecast mode. For the two seasons combined, the mean absolute forecast error was 12.5 ppb for the NLR model and 12.3 ppb for the NN model. The detection rate of 120 ppb threshold exceedances was 42% for each model in the forecast mode. In the hindcast mode, the NLR model performed marginally better than the NN model. The mean absolute hindcast error was 11.1 ppb for the NLR model and 12.9 ppb for the NN model. The hindcast detection rate was 92% for the NLR model and 75% for the NN model.     

Ambient ozone (O3) and adverse crop response: a unified view of cause and effect

This paper presents a cohesive view of the dynamics of ambient O(3) exposure and adverse crop response relationships, coupling the properties of photochemical O(3) production, flux of O(3) from the atmosphere into crop canopies and the crop response per se. The results from two independent approaches ((a) statistical and (b) micrometeorological) were analyzed for understanding cause-effect relationships of the foliar injury responses of tobacco cv Bel-W3 to the exposure dynamics of ambient O(3) concentrations. Similarly, other results from two independent approaches were analyzed in: (1) establishing a micrometeorological relationship between hourly ambient O(3) concentrations and their vertical flux from the air into a natural grassland canopy; and (2) establishing a statistical relationship between hourly ambient O(3) concentrations in long-term, chronic exposures and crop yield reductions. Independent of the approach used, atmospheric conditions appeared to be most conducive and the crop response appeared to be best explained statistically by the cumulative frequency of hourly ambient O(3) concentrations between 50 ppb and 90 ppb (100 and 180 microg m(-3)). In general, this concentration range represents intermediate or moderately enhanced hourly O(3) values in a polluted environment. Further, the diurnal occurrence of this concentration range (often approximately between 0900 and 1600 h in a polluted, agricultural environment) coincided with the optimal CO(2) flux from the atmosphere into the crop canopy, thus high uptake. The frequency of occurrence of hourly O(3) concentrations > 90 ppb (180 microg m(-3)) appeared to be of little importance and such concentrations in general appeared to occur during atmospheric conditions which did not facilitate optimal vertical flux into the crop canopy, thus low uptake. Alternatively, when > 90 ppb (180 microg m(-3)) O(3) concentrations occurred during the 0900-1600 h window, their frequency of occurrence was low in comparison to the 50-90 ppb (100-180 microg m(-3)) range. Based on the overall results, we conclude that if the cumulative frequency of hourly ambient O(3) concentrations between 50-62 ppb (100-124 microg m(-3)) occurred during 53% of the growing season and the corresponding cumulative frequency of hourly O(3) concentrations between 50-74 ppb (100-148 microg m(-3)) occurred during 71% of the growing season, then yield reductions in sensitive crops could be expected, if other factors supporting growth, such as adequate soil moisture are not limiting.     

Fuzzy system models combined with nonlinear regression for daily ground-level ozone predictions

This study focuses on applying a Takagi–Sugeno fuzzy system and a nonlinear regression (NLR) model for ozone predictions in six Kentucky metropolitan areas. The fuzzy “c-means” clustering technique coupled with an optimal output predefuzzification approach (least square method) was used to train the Takagi–Sugeno fuzzy system. The fuzzy system was tuned by specifying the number of rules and the fuzziness factor. The NLR models were based in part on a previously reported, trajectory-based hybrid NLR model that has been used for years for forecasting ground-level ozone in Louisville, KY. The NLR models were each composed of an interactive nonlinear term and several linear terms. Using a common meteorological parameter set as input variables, the NLR models and the Takagi–Sugeno fuzzy systems model exhibited equivalent forecasting performance on test data from 2004. For all 2004 ozone season forecasts for the six metropolitan areas, the mean absolute error was 8.1 ppb for the NLR model and 8.0 ppb for the Takagi–Sugeno fuzzy model. When a nonlinear term (which was part of the NLR model) was included in the fuzzy model, the combined NLR–fuzzy model had slightly better performance than the original NLR model. For all 2004 metropolitan area forecasts, the mean absolute error of the NLR–fuzzy model forecasts was 7.7 ppb. These small differences may be statistically significant, but for practical purposes the performance of the fuzzy models was equivalent to that of the NLR models.     

Measurements of sulfur dioxide and formaldehyde in Taipei using a differential optical absorption spectrometer

Taipei, the capital city of Taiwan, lies in a basin, and its topography prevents the dispersion of pollutants in the city. As a continuation of our air quality study, from February 1999 through June 1999, we measured the concentrations of SO2 at six different locations and of formaldehyde at five locations using a differential optical absorption spectrometer (DOAS). The average concentration of SO2 varied from 3.5 to 6.6 ppb. The average concentration was highest at Toucheng because of its proximity to point sources. The level in Hsientien was close to that in Toucheng, with Hsinyu showing the lowest concentrations. The DOAS and the Taiwan Air Quality Monitoring Network (TAQMN) measurements for SO2 were highly correlated (r2 > 0.9) for Toucheng, Panchiao, and Hsientien. However, DOAS SO2 concentrations were 2 times higher for Hsientien and slightly lower for Panchiao than the TAQMN concentrations were. The average concentration of formaldehyde varied from 7 to 10 ppb. Diurnal variation of formaldehyde closely followed the variation of ozone, especially when the 1-hr peak ozone concentration was > 60 ppb. Photochemical formation accounted for the ambient levels of formaldehyde in Taipei. Concentration of formaldehyde became significant on days when O3 concentration was high. Our results indicate that DOAS can replace conventional measurement techniques.     

Ozone formation in California's San Joaquin Valley: a critical assessment of modeling and data needs

Data from the 1990 San Joaquin Valley Air Quality Study/Atmospheric Utility Signatures, Predictions, and Experiments (SJVAQS/AUSPEX) field program in California's San Joaquin Valley (SJV) suggest that both urban and rural areas would have difficulty meeting an 8-hr average O3 standard of 80 ppb. A conceptual model of O3 formation and accumulation in the SJV is formulated based on the chemical, meteorological, and tracer data from SJVAQS/AUSPEX. Two major phenomena appear to lead to high O3 concentrations in the SJV: (1) transport of O3 and precursors from upwind areas (primarily the San Francisco Bay Area, but also the Sacramento Valley) into the SJV, affecting the northern part of the valley, and (2) emissions of precursors, mixing, transport (including long-range transport), and atmospheric reactions within the SJV responsible for regional and urban-scale (e.g., down-wind of Fresno and Bakersfield) distributions of O3. Using this conceptual model, we then conduct a critical evaluation of the meteorological model and air quality model. Areas of model improvements and data needed to understand and properly simulate O3 formation in the SJV are highlighted.     

Sources and concentrations of indoor nitrogen dioxide in Barcelona, Spain

Sources and concentrations of indoor nitrogen dioxide (NO2) were examined in Barcelona, Spain, during 1996-1999. A total of 340 dwellings of infants participating in a hospital-based cohort study were selected from different areas of the city. Passive filter badges were used for indoor NO2 measurement over 7-30 days. Dwelling inhabitants completed a questionnaire on housing characteristics and smoking habits. Data on outdoor NO2 concentrations were available for the entire period of the study in the areas of the city where indoor concentrations were determined. Bivariate analysis was performed to investigate relationships between indoor NO2 concentrations on one hand and outdoor NO2 concentrations, housing, and occupant characteristics on the other. Stepwise multiple linear regression was performed with variables that were found to have a significant bivariate relationship. Indoor NO2 mean values ranged between 23.57 ppb in 1996 and 27.02 ppb in 1999, with the highest yearly value of 27.82 ppb in 1997. In the same time period, mean outdoor NO2 concentration ranged between 25.26 and 25.78 ppb with a peak of 30.5 ppb in 1998. Multiple regression analysis showed that principal sources of indoor NO2 concentrations were the use of a gas cooker, the absence of an extractor fan when cooking, and cigarette smoking. The absence of central heating was also associated with higher NO2 concentrations. Finally, each ppb increase in outdoor NO2 was associated with a 1% increase in indoor concentrations.