Estimating near-roadway air pollution from multi-frequency noise measurements.

Air pollution is a major environmental problem and its monitoring is essential for regulatory purposes, policy making, and protecting public health. However, dense networks of air quality monitoring equipment are prohibitively expensive due to equipment costs, labor requirements, and infrastructure needs. As a result, alternative lower-cost methods that reliably determine air quality levels near potent pollution sources such as freeways are desirable. We present an approach that couples noise frequency measurements with machine learning to estimate near-roadway particulate matter (PM2.5), nitrogen dioxide (NO2), and black carbon (BC) at 1-min temporal resolution. The models were based on data collected by co-located noise and air quality instruments near a busy freeway in Long Beach, California. Model performance was excellent for all three pollutants, e.g., NO2 predictions yielded Pearson's R = 0.87 with a root mean square error of 7.2 ppb; this error represents about 10 % of total morning rush hour concentrations. Among the best air pollutant predictors were noise frequencies at 40 Hz, 500 Hz, and 800 Hz, and meteorology, particularly wind direction. Overall, our method potentially provides a cost-effective and efficient approach to estimating and/or supplementing near-road air pollutant concentrations in urban areas at high temporal resolution.

Disparities in greenspace associated with sleep duration among adolescent children in Southern California.

More than half of adolescent children do not get the recommended 8 hours of sleep necessary for optimal growth and development. In adults, several studies have evaluated effects of urban stressors including lack of greenspace, air pollution, noise, nighttime light, and psychosocial stress on sleep duration. Little is known about these effects in adolescents, however, it is known that these exposures vary by socioeconomic status (SES). We evaluated the association between several environmental exposures and sleep in adolescent children in Southern California. Methods: In 2010, a total of 1476 Southern California Children's Health Study (CHS) participants in grades 9 and 10 (mean age, 13.4 years; SD, 0.6) completed a questionnaire including topics on sleep and psychosocial stress. Exposures to greenspace, artificial light at night (ALAN), nighttime noise, and air pollution were estimated at each child's residential address, and SES was characterized by maternal education. Odds ratios and 95% confidence intervals (95% CIs) for sleep outcomes were estimated by environmental exposure, adjusting for age, sex, race/ethnicity, home secondhand smoke, and SES. Results: An interquartile range (IQR) increase in greenspace decreased the odds of not sleeping at least 8 hours (odds ratio [OR], 0.86 [95% CI, 0.71, 1.05]). This association was significantly protective in low SES participants (OR, 0.77 [95% CI, 0.60, 0.98]) but not for those with high SES (OR, 1.16 [95%CI, 0.80, 1.70]), interaction P = 0.03. Stress mediated 18.4% of the association among low SES participants. Conclusions: Residing in urban neighborhoods of greater greenness was associated with improved sleep duration among children of low SES but not higher SES. These findings support the importance of widely reported disparities in exposure and access to greenspace in socioeconomically disadvantaged populations.

Estimating traffic noise over a large urban area: An evaluation of methods.

Unlike air pollution, traffic-related noise remains unregulated and has been under-studied despite evidence of its deleterious health impacts. To characterize population exposure to traffic noise, both acoustic-based numerical models and data-driven statistical approaches can generate estimates over large urban areas. The aim of this work is to formally compare the performances of the most common traffic noise models by evaluating their estimates for different categories of roads and validating them against a unique dataset of measured noise in Long Beach, California. Specifically, a statistical land use regression model, an extreme gradient boosting machine learning model (XGB), and three numerical/acoustic traffic noise models: the US Noise Model (FHWA-TNM2.5), a commercial noise model (CadnaA), and an open-source European model (Harmonoise) were optimized and compared. The results demonstrate that XGB and CadnaA were the most effective models for estimating traffic noise, and they are particularly adept at differentiating noise levels on different categories of road.

Exposure models for particulate matter elemental concentrations in Southern California.

Due to a scarcity of routine monitoring of speciated particulate matter (PM), there has been limited capability to develop exposure models that robustly estimate component-specific concentrations. This paper presents the largest such study conducted in a single urban area. Using samples that were collected at 220 locations over two seasons, quasi-ultrafine (PM0.2), accumulation mode fine (PM0.2-2.5), and coarse (PM2.5-10) particulate matter concentrations were used to develop spatiotemporal regression, machine learning models that enabled predictions of 24 elemental components in eight Southern California communities. We used supervised variable selection of over 150 variables, largely from publicly available sources, including meteorological, roadway and traffic characteristics, land use, and dispersion model estimates of traffic emissions. PM components that have high oxidative potential (and potentially large health effects) or are otherwise important markers for major PM sources were the primary focus. We present results for copper, iron, and zinc (as non-tailpipe vehicle emissions); elemental carbon (diesel emissions); vanadium (ship emissions); calcium (soil dust); and sodium (sea salt). Spatiotemporal linear regression models with 17 to 36 predictor variables including meteorology; distance to different classifications of roads; intersections and off ramps within a given buffer distance; truck and vehicle traffic volumes; and near-roadway dispersion model estimates produced superior predictions over the machine learning approaches (cross validation R-squares ranged from 0.76 to 0.92). Our models are easily interpretable and appear to have more effectively captured spatial gradients in the metallic portion of PM than other comparably large studies, particularly near roadways for the non-tailpipe emissions. Furthermore, we demonstrated the importance of including spatiotemporally resolved meteorology in our models as it helped to provide key insights into spatial patterns and allowed us to make temporal predictions.

Manuscript title: Long─term residential exposure to environmental/transportation noise and the incidence of myocardial infarction.

BACKGROUND: Cardiovascular effects of environmental noise are a growing concern. However, the evidence remains largely limited to the association between road traffic noise and hypertension and coronary heart diseases. OBJECTIVES: To investigate the association between long-term residential exposure to environmental/transportation noise and the incidence of myocardial infarction (MI) in the adult population living in Montreal. METHODS: An open cohort of adults aged 45 years old and over, living on the island of Montreal and free of MI before entering the cohort was created for the years 2000-2014 with the Quebec Integrated Chronic Disease Surveillance System; a systematic surveillance system from the Canadian province of Quebec starting in 1996. Residential noise exposure was calculated in three ways: 1) total ambient noise levels estimated by Land use regression (LUR) models; 2) road traffic noise estimated by a noise propagation model CadnaA and 3) distances to transportation sources (roads, airport, railways). Incident MI was based on diagnostic codes in hospital admission records. Cox models with time-varying exposures (age as the time axis) were used to estimate the associations with various adjustments (material deprivation indicator, calendar year, nitrogen dioxide, stratification for sex). Indirect adjustment based on ancillary data for smoking was performed. RESULTS: 1,065,414 individuals were followed (total of 9,000,443 person-years) and 40,718 (3.8%) developed MI. We found positive associations between total environmental noise, estimated by LUR models and the incidence of MI. Total noise LUR levels ranged from ~44 to ~79 dBA and varied slightly with the metric used. The adjusted hazard ratios (HRs) (also adjusted for smoking) were 1.12 (95% Confidence Intervals [CI]: 1.08-1.15), 1.11 (95%CI: 1.07-1.14) and 1.10 (95%CI: 1.06-1.14) per 10 dBA noise levels increase respectively in Level Accoustic equivalent 24 h (LAeq24 h), Level day-evening-night (Lden) and night level (Lnight). We found a borderline negative association between road noise levels estimated with CadnaA and MI (HR: 0.99 per 10 dBA; 95%CI: 0.98-1.00). Distances to major roads and highways were not associated with MI while the proximity to railways was positively associated with MI (HR for ≤100 vs > 1000 m: 1.07; 95%CI: 1.01-1.14). A negative association was found with the proximity to the airport noise exposure forecast (NEF25); HR (<1 vs >1000 m) = 0.88 (95%CI: 0.81-0.96). CONCLUSIONS: These associations suggest that exposure to total environmental noise at current urban levels may be related to the incidence of MI. Additional studies with more accurate road noise estimates are needed to explain the counterintuitive associations with road noise and specific transportation sources.

Predicting Fine Spatial Scale Traffic Noise Using Mobile Measurements and Machine Learning.

Environmental noise has been associated with a variety of health endpoints including cardiovascular disease, sleep disturbance, depression, and psychosocial stress. Most population noise exposure comes from vehicular traffic, which produces fine-scale spatial variability that is difficult to characterize using traditional fixed-site measurement techniques. To address this challenge, we collected A-weighted, equivalent noise (LAeq in decibels, dB) data on hour-long foot journeys around 16 locations throughout Long Beach, California and trained four machine learning models, linear regression, random forest, extreme gradient boosting, and a neural network, to predict noise with 20 m resolution. Input variables to the models included traffic metrics, road network features, meteorological conditions, and land use type. Among all machine learning models, extreme gradient boosting had the best results in validation tests (leave-one-route-out R2 = 0.71, root mean square error (RMSE) of 4.54 dB; 5-fold R2 = 0.96, RMSE of 1.8 dB). Local traffic volume was the most important predictor of noise; road features, land use, and meteorology including humidity, temperature, and wind speed also contributed. We show that a novel, on-foot mobile noise measurement method coupled with machine learning approaches enables highly accurate prediction of small-scale spatial patterns in traffic-related noise over a mixed-use urban area.

Capturing the spatial variability of noise levels based on a short-term monitoring campaign and comparing noise surfaces against personal exposures collected through a panel study.

Environmental noise can cause important cardiovascular effects, stress and sleep disturbance. The development of appropriate methods to estimate noise exposure within a single urban area remains a challenging task, due to the presence of various transportation noise sources (road, rail, and aircraft). In this study, we developed a land-use regression (LUR) approach using a Generalized Additive Model (GAM) for LAeq (equivalent noise level) to capture the spatial variability of noise levels in Toronto, Canada. Four different model formulations were proposed based on continuous 20-min noise measurements at 92 sites and a leave one out cross-validation (LOOCV). Models where coefficients for variables considered as noise sources were forced to be positive, led to the development of more realistic exposure surfaces. Three different measures were used to assess the models; adjusted R2 (0.44-0.64), deviance (51-72%) and Akaike information criterion (AIC) (469.2-434.6). When comparing exposures derived from the four approaches to personal exposures from a panel study, we observed that all approaches performed very similarly, with values for the Fractional mean bias (FB), normalized mean square error (NMSE), and normalized absolute difference (NAD) very close to 0. Finally, we compared the noise surfaces with data collected from a previous campaign consisting of 1-week measurements at 200 fixed sites in Toronto and observed that the strongest correlations occurred between our predictions and measured noise levels along major roads and highway collectors. Our validation against long-term measurements and panel data demonstrates that manual modifications brought to the models were able to reduce bias in model predictions and achieve a wider range of exposures, comparable with measurement data.

Evaluating the Impact of Neighborhood Characteristics on Differences between Residential and Mobility-Based Exposures to Outdoor Air Pollution.

Epidemiological studies often assign outdoor air pollution concentrations to residential locations without accounting for mobility patterns. In this study, we examined how neighborhood characteristics may influence differences in exposure assessments between outdoor residential concentrations and mobility-based exposures. To do this, we linked residential location and mobility data to exposure surfaces for NO2, PM2.5, and ultrafine particles in Montreal, Canada for 5452 people in 2016. Mobility data were collected using the MTL Trajet smartphone application (mean: 16 days/subject). Generalized additive models were used to identify important neighborhood predictors of differences between residential and mobility-based exposures and included residential distances to highways, traffic counts within 500 m of the residence, neighborhood walkability, median income, and unemployment rate. Final models including these parameters provided unbiased estimates of differences between residential and mobility-based exposures with small root-mean-square error values in 10-fold cross validation samples. In general, our findings suggest that differences between residential and mobility-based exposures are not evenly distributed across cities and are greater for pollutants with higher spatial variability like NO2. It may be possible to use neighborhood characteristics to predict the magnitude and direction of this error to better understand its likely impact on risk estimates in epidemiological analyses.

Comparing emission rates derived from a model with those estimated using a plume-based approach and quantifying the contribution of vehicle classes to on-road emissions and air quality.

This study presents a comparison of fleet average emission factor (s) derived from a traffic emission model with EFs estimated using plume-based measurements, including an investigation of the contribution of vehicle classes to carbon monoxide (CO), nitrogen oxides (NOx), and elemental carbon (EC) along an urban corridor. To this end, a field campaign was conducted over one week in June 2016 on an arterial road in Toronto, Canada. Traffic data were collected using a traffic camera and a radar, whereas air quality was characterized using two monitoring stations: one located at ground level and another at the rooftop of a four-story building. A traffic simulation model was calibrated and validated, and second-by-second speed profiles for all vehicle trajectories were extracted to model emissions. In addition, dispersion modeling was conducted to identify the extent to which differences in emissions translate to differences in near-road concentrations. The results indicate that modeled EFs for CO and NOx are twice as high as plume-based EFs. Besides, modeled results indicate that transit bus emissions accounted for 60% and 70% of the total emissions of NOx and EC, respectively. Transit bus emission rates in g/passenger·km for NOx and EC were up to 8 and 22 times, respectively, the emission rates of passenger cars. In contrast, the Toronto streetcars, which are electrically fueled, were found to improve near-road air quality despite their negative impact on traffic speeds. Finally, we observe that the difference in estimated concentrations derived from the two methods is not as large as the difference in estimated emissions due to the influence of meteorology and of the urban background given that the study network is located in a busy downtown area. Implications: This study presents a comparison of fleet average emission factor (s) derived from a traffic emission model with EFs estimated using plume-based measurements, including an investigation of the contribution of vehicle classes to various pollutants. Besides, dispersion modeling was conducted to identify the extent to which differences in emissions translate to differences in near-road concentrations. It was observed that the difference in estimated concentrations derived from the two methods is not as large as the difference in estimated emissions due to the influence of meteorology and of the urban background, as the study network is located in a busy downtown area.

Road traffic impact on urban water quality: a step towards integrated traffic, air and stormwater modelling.

Methods for simulating air pollution due to road traffic and the associated effects on stormwater runoff quality in an urban environment are examined with particular emphasis on the integration of the various simulation models into a consistent modelling chain. To that end, the models for traffic, pollutant emissions, atmospheric dispersion and deposition, and stormwater contamination are reviewed. The present study focuses on the implementation of a modelling chain for an actual urban case study, which is the contamination of water runoff by cadmium (Cd), lead (Pb), and zinc (Zn) in the Grigny urban catchment near Paris, France. First, traffic emissions are calculated with traffic inputs using the COPERT4 methodology. Next, the atmospheric dispersion of pollutants is simulated with the Polyphemus line source model and pollutant deposition fluxes in different subcatchment areas are calculated. Finally, the SWMM water quantity and quality model is used to estimate the concentrations of pollutants in stormwater runoff. The simulation results are compared to mass flow rates and concentrations of Cd, Pb and Zn measured at the catchment outlet. The contribution of local traffic to stormwater contamination is estimated to be significant for Pb and, to a lesser extent, for Zn and Cd; however, Pb is most likely overestimated due to outdated emissions factors. The results demonstrate the importance of treating distributed traffic emissions from major roadways explicitly since the impact of these sources on concentrations in the catchment outlet is underestimated when those traffic emissions are spatially averaged over the catchment area.