Modeling temporal and spatial variability of traffic-related air pollution: Hourly land use regression models for black carbon

Atmospheric Environment (Impact Factor: 3.28). 08/2013; 74. DOI: 10.1016/j.atmosenv.2013.03.050


Land use regression (LUR) modeling is a statistical technique used to determine exposure to air pollutants in epidemiological studies. Time-activity diaries can be combined with LUR models, enabling detailed exposure estimation and limiting exposure misclassification, both in shorter and longer time lags.

In this study, the traffic related air pollutant black carbon was measured with μ-aethalometers on a 5-min time base at 63 locations in Flanders, Belgium. The measurements show that hourly concentrations vary between different locations, but also over the day. Furthermore the diurnal pattern is different for street and background locations. This suggests that annual LUR models are not sufficient to capture all the variation. Hourly LUR models for black carbon are developed using different strategies: by means of dummy variables, with dynamic dependent variables and/or with dynamic and static independent variables.

The LUR model with 48 dummies (weekday hours and weekend hours) performs not as good as the annual model (explained variance of 0.44 compared to 0.77 in the annual model). The dataset with hourly concentrations of black carbon can be used to recalibrate the annual model, resulting in many of the original explaining variables losing their statistical significance, and certain variables having the wrong direction of effect. Building new independent hourly models, with static or dynamic covariates, is proposed as the best solution to solve these issues. R2 values for hourly LUR models are mostly smaller than the R2 of the annual model, ranging from 0.07 to 0.8. Between 6 a.m. and 10 p.m. on weekdays the R2 approximates the annual model R2. Even though models of consecutive hours are developed independently, similar variables turn out to be significant. Using dynamic covariates instead of static covariates, i.e. hourly traffic intensities and hourly population densities, did not significantly improve the models' performance.

Download full-text


Available from: Evi Dons,
190 Reads
  • Source
    • "However, temporally-refined concentration estimates may be needed for birth cohort studies and acute exposure analyses. The few LUR models that incorporate detailed temporal resolution include Su et al. (2008), who use hourly NO 2 concentrations data to assess seasonal variations in Vancouver, Johnson et al. (2013), who predict daily concentrations of NO 2 and PM 2.5 in Windsor, Novotny et al. (2011), who, in addition to annual NO 2 models, estimate seasonal, day-type (weekends, weekdays) and hourly models, and Dons et al. (2013), who estimate hourly black carbon concentrations models. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Transportation and land-use activities are major air pollution contributors. Since their shares of emissions vary across space and time, so do air pollution concentrations. Despite these variations, panel data have rarely been used in land-use regression (LUR) modeling of air pollution. In addition, the complex interactions between traffic flows, land uses, and meteorological variables, have not been satisfactorily investigated in LUR models. The purpose of this research is to develop and estimate nitrogen dioxide (NO2) panel models based on the LUR framework with data for Seoul, Korea, accounting for the impacts of these variables, and their interactions with spatial and temporal dummy variables. The panel data vary over several scales: daily (24 h), seasonally (4), and spatially (34 intra-urban measurement locations). To enhance model explanatory power, wind direction and distance decay effects are accounted for. The results show that vehicle-kilometers-traveled (VKT) and solar radiation have statistically strong positive and negative impacts on NO2 concentrations across the four seasonal models. In addition, there are significant interactions with the dummy variables, pointing to VKT and solar radiation effects on NO2 concentrations that vary with time and intra-urban location. The results also show that residential, commercial, and industrial land uses, and wind speed, temperature, and humidity, all impact NO2 concentrations. The R2 vary between 0.95 and 0.98.
    Atmospheric Environment 04/2015; 107:364-373. DOI:10.1016/j.atmosenv.2015.02.053 · 3.28 Impact Factor
  • Source
    • "Because the spatial concentration pattern varies during a day, hourly LUR models were developed (24 models for weekday-hours, and 24 models for weekend-hours) [22]. Weekday hourly models performed well during the day and on traffic peak hours, explaining 60 to 80% of variability [22]. At night and in the weekend, concentrations were lower and more homogeneous resulting in less predictive models when considering R², on the other hand the mean squared error was also low. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The AB2C model (Activity-Based modeling framework for Black Carbon exposure assessment) was developed to assess personal exposure to air pollution, more specifically black carbon. Currently the model calculates exposure in Flanders, an urbanized region in Western Europe. This model is characterized by the use of time-activity patterns, and air pollution concentrations with a high spatial and temporal resolution, including indoors and in the transport microenvironment. This model can be used for disaggregated exposure assessment or the evaluation of policy scenarios. In this paper, exposure of people from a lower socioeconomic class (SEC) is compared to the exposure of people from a higher SEC. In most North American studies, it is reported that poorer people are exposed to higher concentrations and suffer more from health effects associated with elevated exposure to air pollution. In Europe, fewer studies exist in this field, and results are not always conclusive. In this study, people from a lower SEC were found to be exposed to higher concentrations at home, but ‘richer’ people travel more, especially in traffic peak hours. This results in an average exposure that is higher for members of a lower SEC, but inhaled doses are similar in both groups. This analysis suggests that differences in health impact between the groups are almost completely explainable by increased susceptibility to air pollution health effects, and not by increased air pollutant intake.
    Procedia Computer Science 11/2014; 32:269–276. DOI:10.1016/j.procs.2014.05.424
  • Source
    • "Weekday hourly models performed well during the day and on traffic peak hours, explaining 60 to 80% of variability using mainly traffic variables (Dons et al., 2013b). At night and in the weekend the models were less predictive using only 1 or 2 predictors (mainly land use variables in larger buffers), but RMSE values were low indicative of homogeneous BC concentrations (Dons et al., 2013b). Traffic and population variables from the activity-based model were only sporadically included, e.g. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Because people tend to move from one place to another during the day, their exposure to air pollution will be determined by the concentration at each location combined with the exposure encountered in transport. In order to estimate the exposure of individuals in a population more accurately, the activity-based modeling framework for Black Carbon exposure assessment, AB(2)C, was developed. An activity-based traffic model was applied to model the whereabouts of individual agents. Exposure to black carbon (BC) in different microenvironments is assessed with a land use regression model, combined with a fixed indoor/outdoor factor for exposure in indoor environments. To estimate exposure in transport, a separate model was used taking into account transport mode, timing of the trip and degree of urbanization. The modeling framework is validated using weeklong time-activity diaries and BC exposure as revealed from a personal monitoring campaign with 62 participants. For each participant in the monitoring campaign, a synthetic population of 100 model-agents per day was made up with all agents meeting similar preconditions as each real-life agent. When these model-agents pass through every stage of the modeling framework, it results in a distribution of potential exposures for each individual. The AB(2)C model estimates average personal exposure slightly more accurately compared to ambient concentrations as predicted for the home subzone; however the added value of a dynamic model lies in the potential for detecting short term peak exposures rather than modeling average exposures. The latter may bring new opportunities to epidemiologists: studying the effect of frequently repeated but short exposure peaks on long term exposure and health.
    Environment international 10/2013; 62C:64-71. DOI:10.1016/j.envint.2013.10.003 · 5.56 Impact Factor
Show more