Evaluation of artificial neural networks for fine particulate pollution (PM10 and PM2.5) forecasting

Multi-layer perceptron (MLP) artificial neural network (ANN) models are compared with traditional multiple regression (MLR) models for daily maximum and average O3 and particulate matter (PM10 and PM2.5) forecasting. MLP particulate forecasting models show little if any improvement over MLR models and exhibit less skill than do O3 forecasting models. Meteorological variables (precipitation, wind, and temperature), persistence, and co-pollutant data are shown to be useful PM predictors. If MLP approaches are adopted for PM forecasting, training methods that improve extreme value prediction are recommended.     

Prediction of daily averaged PM10 concentrations by statistical time-varying model

In this study, a time-varying statistical model, TVAREX, was proposed for daily averaged PM₁₀ concentrations forecasting of coastal cities. It is a Kalman filter based autoregressive model with exogenous inputs depending on selected meteorological properties on the day of prediction. The TVAREX model was evaluated and compared to an ANN model, trained with the Levenberg–Marquardt backpropagation algorithm subjected to the same set of inputs. It was found that the error statistics of the TVAREX model in general were comparable to those of the ANN model, but the TVAREX model was more efficient in capturing the PM₁₀ pollution episodes due to its online nature, therefore having an appealing advantage for implementation.     

Source identification of personal exposure to fine particulate matter using organic tracers

Personal exposure to fine particulate matter (PM2.5) is due to both indoor and outdoor sources. Contributions of sources to personal exposure can be quite different from those observed at ambient sampling locations. The primary goal of this study was to investigate the effectiveness of using trace organic speciation data to help identify sources influencing PM2.5 exposure concentrations. Sixty-four 24-h PM2.5 samples were obtained on seven different subjects in and around Boulder, CO. The exposure samples were analyzed for PM2.5 mass, elemental and organic carbon, organic tracer compounds, water-soluble metals, ammonia, and nitrate. This study is the first to measure a broad distribution of organic tracer compounds in PM2.5 personal samples. PM2.5 mass exposure concentrations averaged 8.4 μg m^?3. Organic carbon was the dominant constituent of the PM2.5 mass. Forty-four organic species and 19 water-soluble metals were quantifiable in more than half of the samples. Fifty-four organic species and 16 water-soluble metals had measurement signal-to-noise ratios larger than two after blank subtraction.The dataset was analyzed by Principal Component Analysis (PCA) to determine the factors that account for the greatest variance. Eight significant factors were identified; each factor was matched to its likely source based primarily on the marker species that loaded the factor. The results were consistent with the expectation that multiple marker species for the same source loaded the same factor. Meat cooking was an important source of variability. The factor that represents meat cooking was highly correlated with organic carbon concentrations (r = 0.84). The correlation between ambient PM2.5 and PM2.5 exposure was relatively weak (r = 0.15). Time participants spent performing various activities was generally not well correlated with PCA factor scores, likely because activity duration does not measure emissions intensity. The PCA results demonstrate that organic tracers can aid in identifying factors that influence personal exposures to PM2.5.     

Predicting Particulate (PM10) Personal Exposure Distributions Using a Random Component Superposition Statistical Model

ABSTRACT

This paper presents a new statistical model designed to extend our understanding from prior personal exposure field measurements of urban populations to other cities where ambient monitoring data, but no personal exposure measurements, exist. The model partitions personal exposure into two distinct components: ambient concentration and nonambient concentration. It is assumed the ambient and nonambient concentration components are uncorrelated and add together; therefore, the model is called a random component superposition (RCS) model. The 24-hr ambient outdoor concentration is multiplied by a dimensionless “attenuation factor” between 0 and 1 to account for deposition of particles as the ambient air infiltrates indoors. The RCS model is applied to field PM₁₀ measurement data from three large-scale personal exposure field studies: THEES (Total Human Environmental Exposure Study) in Phillipsburg, NJ; PTEAM (Particle Total Exposure Assessment Methodology) in Riverside, CA; and the Ethyl Corporation study in Toronto, Canada. Because indoor sources and activities (smoking, cooking, cleaning, the personal cloud, etc.) may be similar in similar populations, it was hypothesized that the statistical distribution of nonambient personal exposure is invariant across cities.     

Demystifying uncertainty in PM10 susceptibility mapping using variable drop-off in extreme-gradient boosting (XGB) and random forest (RF) algorithms

Environmental Science and Pollution Research - This study investigates uncertainty in machine learning that can occur when there is significant variance in the prediction importance level of the...     

南宁市大气颗粒物TSP、PM10、PM2.5污染水平研究

2002年在南宁市的5个典型城市功能区内,共采集了125个大气样品(按季节分别采集),初步调查了大气中颗粒物TSP、PM10、PM2.5的污染状况。结果表明,南宁市TSP、PM10、PM2.5的污染很严重,超标率分别为67．5％、82．5％、92．5％,对人体健康危害更大的PM2．5占到了PM10的63．5％左右。重污染区PM2.5的浓度超过轻污染区近一倍。     

Determinants of exposure to fine particulate matter (PM2.5) for waiting passengers at bus stops

This research evaluates commuter exposure to particulate matter during pre-journey commute segments for passengers waiting at bus stops by investigating 840 min of simultaneous exposure levels, both inside and outside seven bus shelters in Buffalo, New York. A multivariate regression model is used to estimate the relation between exposure to particulate matter (PM_2.5 measured in μg m^?3) and three vectors of determinants: time and location, physical setting and placement, and environmental factors. Four determinants have a statistically significant effect on particulate matter: time of day, passengers’ waiting location, land use near the bus shelter, and the presence of cigarette smoking at the bus shelter. Model results suggest that exposure to PM_2.5 inside a bus shelter is 2.63 μg m^?3 (or 18 percent) higher than exposure outside a bus shelter, perhaps due in part to the presence of cigarette smoking. Morning exposure levels are 6.51 μg m^?3 (or 52 percent) higher than afternoon levels. Placement of bus stops can affect exposure to particulate matter for those waiting inside and outside of shelters: air samples at bus shelters located in building canyons have higher particulate matter than bus shelters located near open space.     

Predicting personal exposure to airborne carbonyls using residential measurements and time/activity data

As a part of the Relationships of Indoor, Outdoor, and Personal Air (RIOPA) study, 48 h integrated residential indoor, outdoor, and personal exposure concentrations of 10 carbonyls were simultaneously measured in 234 homes selected from three US cities using the Passive Aldehydes and Ketones Samplers (PAKS). In this paper, we examine the feasibility of using residential indoor concentrations to predict personal exposures to carbonyls. Based on paired t-tests, the means of indoor concentrations were not different from those of personal exposure concentrations for eight out of the 10 measured carbonyls, indicating indoor carbonyls concentrations, in general, well predicted the central tendency of personal exposure concentrations. In a linear regression model, indoor concentrations explained 47%, 55%, and 65% of personal exposure variance for formaldehyde, acetaldehyde, and hexaldehyde, respectively. The predictability of indoor concentrations on cross-individual variability in personal exposure for the other carbonyls was poorer, explaining<20% of variance for acetone, acrolein, crotonaldehyde, and glyoxal. A factor analysis, coupled with multiple linear regression analyses, was also performed to examine the impact of human activities on personal exposure concentrations. It was found that activities related to driving a vehicle and performing yard work had significant impacts on personal exposures to a few carbonyls.     

Arsenic speciation of atmospheric particulate matter (PM10) in an industrialised urban site in southwestern Spain

An arsenic speciation study has been performed in PM10 samples collected on a fortnight basis in the city of Huelva (SW Spain) during 2001 and 2002. The arsenic species were extracted from the PM10 filters using a NH2OH x HCl solution and sonication, and determined by HPLC-HG-AFS. The mean bulk As concentration of the samples analyzed during 2001 and 2002 slightly exceed the mean annual 6 ng m(-3) target value proposed by the European Commission for 2013, arsenate [As(V)] being responsible for the high level of arsenic. The speciation analyses showed that As(V) was the main arsenic species found, followed by arsenite [As(III)] (mean 6.5 and 7.8 ng m(-3) for As(V), mean 1.2 and 2.1 ng m(-3) for As(III), in 2001 and 2002, respectively). The high levels of arsenic species found in PM10 in Huelva have a predominant industrial origin, such as the one from a nearby copper smelter, and do not present a seasonal pattern. The highest daily levels of arsenic species correspond to synoptic conditions in which the winds with S and SW components transport the contaminants from the main emission source. The frequent African dust outbreaks over Huelva may result in an increment of mass levels of PM10, but do not represent a significant input of arsenic in comparison to the anthropogenic source. The rural background levels of arsenic around Huelva are rather high, in comparison to other rural or urban areas in Spain, showing a relatively high atmosphere residence time of arsenic. This work shows the importance of arsenic speciation in studies of aerosol chemistry, due to the presence of arsenic species [As(III) and As(V)] with distinct toxicity.     

Year-long continuous personal exposure to PM2.5 recorded by a fast responding portable nephelometer

Personal exposure to particulate matter of aerodynamic diameter under 2.5 μm (PM_2.5) was monitored using a DustTrak nephelometer. The battery-operated unit, worn by an adult individual for a period of approximately one year, logged integrated average PM_2.5 concentrations over 5 min intervals. A detailed time-activity diary was used to record the experimental subject’s movement and the microenvironments visited. Altogether 239 days covering all the months (except April) were available for the analysis. In total, 60 463 acceptable 5-min averages were obtained. The dataset was divided into 7 indoor and 4 outdoor microenvironments. Of the total time, 84% was spent indoors, 10.9% outdoors and 5.1% in transport. The indoor 5-min PM_2.5 average was higher (55.7 μg m^?3) than the outdoor value (49.8 μg m^?3). The highest 5-min PM_2.5 average concentration was detected in restaurant microenvironments (1103 μg m^?3), the second highest 5-min average concentration was recorded in indoor spaces heated by stoves burning solid fuels (420 μg m^?3). The lowest 5-min mean aerosol concentrations were detected outdoors in rural/natural environments (25 μg m^?3) and indoors at the monitored person’s home (36 μg m^?3). Outdoor and indoor concentrations of PM_2.5 measured by the nephelometer at home and during movement in the vicinity of the experimental subject’s home were compared with those of the nearest fixed-site monitor of the national air quality monitoring network. The high correlation coefficient (0.78) between the personal and fixed-site monitor aerosol concentrations suggested that fixed-site monitor data can be used as proxies for personal exposure in residential and some other microenvironments. Collocated measurements with a reference method (β-attenuation) showed a non-linear systematic bias of the light-scattering method, limiting the use of direct concentration readings for exact exposure analysis.     

Toxic metal content of particulate matter (PM10), within the Greater Area of Athens

This study presents results from a yearlong particulate matter measurement campaign, conducted across the Greater Athens Area, at four locations, between 1st June 2001 and 31st May 2002. The collected PM(10) 24-h samples were analyzed for nine toxic metals and metalloids (Pb, As, Cd, Ni, Cr, Mn, V, Cu, Hg). Concerning the five elements regulated by the European Union, annual average concentrations of Pb were found below the limit values at all sites, Cd and Ni concentrations were lower than the prospective assessment thresholds at all sites, concentrations of As exceeded the assessment threshold at two sites, while concentrations of Hg were found below detection limits in all samples. Concentration levels of Mn and V were in compliance with the values proposed by the World Health Organization. The seasonal and spatial variability of metal concentrations was examined and site-specific correlation analysis was conducted for the identification of metals with similar origin. The association between trace metals and NO(x) concentrations was explored to account for the impact of automotive sources, at two traffic-impacted sites. Cu was the metal most closely linked with the road transport sector. The relation of concentration levels with the prevalence of winds from different sectors was studied in an effort to investigate the transport of metal particles from various zones of the city. Finally, factor analysis was carried out to extract the main components responsible for the variance of the dataset and to attribute them to specific source categories, with vehicle-related sources being important in all cases.     

An hourly PM10 diagnosis model for the Bilbao metropolitan area using a linear regression methodology

There is extensive evidence of the negative impacts on health linked to the rise of the regional background of particulate matter (PM) 10 levels. These levels are often increased over urban areas becoming one of the main air pollution concerns. This is the case on the Bilbao metropolitan area, Spain. This study describes a data-driven model to diagnose PM10 levels in Bilbao at hourly intervals. The model is built with a training period of 7-year historical data covering different urban environments (inland, city centre and coastal sites). The explanatory variables are quantitative—log [NO₂], temperature, short-wave incoming radiation, wind speed and direction, specific humidity, hour and vehicle intensity—and qualitative—working days/weekends, season (winter/summer), the hour (from 00 to 23 UTC) and precipitation/no precipitation. Three different linear regression models are compared: simple linear regression; linear regression with interaction terms (INT); and linear regression with interaction terms following the Sawa’s Bayesian Information Criteria (INT-BIC). Each type of model is calculated selecting two different periods: the training (it consists of 6 years) and the testing dataset (it consists of 1 year). The results of each type of model show that the INT-BIC-based model (R ²?=?0.42) is the best. Results were R of 0.65, 0.63 and 0.60 for the city centre, inland and coastal sites, respectively, a level of confidence similar to the state-of-the art methodology. The related error calculated for longer time intervals (monthly or seasonal means) diminished significantly (R of 0.75–0.80 for monthly means and R of 0.80 to 0.98 at seasonally means) with respect to shorter periods.     

Characterization of model error in a simulation of fine particulate matter exposure distributions of the working age population in Helsinki, Finland

Exposure models are needed for comparison of scenarios resulting from alternative policy options. The reliability of models used for such purposes should be quantified by comparing model outputs in a real situation with the corresponding observed exposures. Measurement errors affect the observations, but if the distribution of these errors for single observations is known, the bias caused for the population statistics can be corrected. The current paper does this and calculates model errors for a probabilistic simulation of 48-hr fine particulate matter (PM2.5) exposures. Direct and nested microenvironment-based models are compared. The direct model requires knowledge on the distribution of the indoor concentrations, whereas the nested model calculates indoor concentrations from ambient levels, using infiltration factors and indoor sources. The model error in the mean exposure level was <0.5 microg m(-3) for both models. Relative errors in the estimated population mean were +1% and -5% for the direct and nested models, respectively. Relative errors in the estimated SD were -9% and -23%, respectively. The magnitude of these errors and the errors calculated for population percentiles indicate that the model errors would not drive general conclusions derived from these models, supporting the use of the models as a tool for evaluation of potential exposure reductions in alternative policy scenarios.     

Estimating fine particulate matter component concentrations and size distributions using satellite-retrieved fractional aerosol optical depth: part 1--method development

We develop a method that uses both the total column aerosol optical depth (AOD) and the fractional AOD values for different aerosol types, derived from Multiangle Imaging SpectroRadiometer (MISR) aerosol data, to estimate ground-level concentrations of fine particulate matter (PM2.5) mass and its major constituents in eastern and western United States. Compared with previous research on linking column AOD with ground-level PM2.5, this method treats various MISR aerosol components as individual predictor variables. Therefore, the contributions of different particle types to PM2.5 concentrations can be estimated. When AOD is greater than 0.15, MISR is able to distinguish dust from non-dust particles with an uncertainty level of approximately 4%, and light-absorbing from non-light-absorbing particles with an uncertainty level of approximately 20%. Further analysis shows that MISR Version 17 aerosol microphysical properties have good sensitivity and internal consistency among different mixture classes. The retrieval uncertainty of individual fractional AODs ranges between 5 and 11% in the eastern United States, and between 11 and 31% in the west for non-dust aerosol components. These results provide confidence that the fractional AOD models with their inherent flexibility can make more accurate predictions of the concentrations of PM2.5 and its constituents.     

Source apportionment of fine particulate matter (PM2.5) at a rural Ohio River Valley site

Twenty four-hour averaged concentrations of fine particulate matter were collected at Athens, OH between March 2004 and November 2005 in an effort to characterize the nature of PM_2.5 and apportion its sources. PM_2.5 samples were chemically analyzed and positive matrix factorization was applied to this speciation data to identify the probable sources. PMF arrived at a 7-factor model to most accurately apportion sources of the PM_2.5 observed at Athens. Conditional probability function (CPF) and potential source contribution function (PSCF) were applied to the identified sources to investigate the geographical location of these sources. Secondary sulfate source dominated the contributions with a total contribution of 62.6% with the primary and secondary organic source following second with 19.9%. Secondary nitrate contributed a total of 6.5% with the steel production source and Pb- and Zn-source coming in at 3.1% and 2.9%, respectively. Crustal and mobile sources were small contributors (2.5% each) of PM_2.5 to the Athens region. The secondary sulfate, secondary organic and nitrate portrayed a clear seasonal nature with the sulfate and secondary organic peaking in the warm months and the nitrate reaching a high in the cold months. The high percentage of secondary sulfate observed at a rural site like Athens suggests the involvement of regional transport mechanisms.     

Evaluation of human exposure to ambient PM10 in the metropolitan area of Mexico City using a GIS-based methodology.

The main goal of this study was to evaluate the magnitude of outdoor exposure to fine particulate matter (PM10) potentially experienced by the population of metropolitan Mexico City. With the use of a geographic information system (GIS), spatially resolved PM10 distributions were generated and linked to the local population. The PM10 concentration exceeded the 24-hr air quality standard of 150 microg/m3 on 16% of the days, and the annual air quality standard of 50 microg/m3 was exceeded by almost twice its value in some places. The basic methodology described in this paper integrates spatial demographic and air quality databases, allowing the evaluation of various air pollution reduction scenarios. Achieving the annual air quality standard would represent a reduction in the annual arithmetic average concentration of 14 microg/m3 for the typical inhabitant. Human exposure to particulate matter (PM) has been associated with mortality and morbidity in Mexico City; reducing the concentration levels of this pollutant would represent a reduction in mortality and morbidity and the associated cost of such impacts. This methodology is critical to assessing the potential benefits of the current initiative to improve air quality implemented by the Environmental Metropolitan Commission of Mexico City.     

Source apportionment of particulate matter (PM2.5) in an urban area using dispersion,receptor and inverse modelling

Air pollution emission inventories are the basis for air quality assessment and management strategies. The quality of the inventories is of great importance since these data are essential for air pollution impact assessments using dispersion models. In this study, the quality of the emission inventory for fine particulates (PM_2.5) is assessed: first, using the calculated source contributions from a receptor model; second, using source apportionment from a dispersion model; and third, by applying a simple inverse modelling technique which utilises multiple linear regression of the dispersion model source contributions together with the observed PM_2.5 concentrations. For the receptor modelling the chemical composition of PM_2.5 filter samples from a measurement campaign performed between January 2004 and April 2005 are analysed. Positive matrix factorisation is applied as the receptor model to detect and quantify the various source contributions. For the same observational period and site, dispersion model calculations using the Air Quality Management system, AirQUIS, are performed. The results identify significant differences between the dispersion and receptor model source apportionment, particularly for wood burning and traffic induced suspension. For wood burning the receptor model calculations are lower, by a factor of 0.54, but for the traffic induced suspension they are higher, by a factor of 7.1. Inverse modelling, based on regression of the dispersion model source contributions and the PM_2.5 concentrations, indicates similar discrepancies in the emissions inventory. In order to assess if the differences found at the one site are generally applicable throughout Oslo, the individual source category emissions are rescaled according to the receptor modelling results. These adjusted PM_2.5 concentrations are compared with measurements at four independent stations to evaluate the updated inventory. Statistical analysis shows improvement in the estimated concentrations for PM_2.5 at all sites. Similarly, inverse modelling is applied at these independent sites and this confirms the validity of the receptor model results.     

Genotoxicity of organic extracts of urban airborne particulate matter: an assessment within a personal exposure study

Airborne particulate matter, PM(10) and PM(2.5), are associated with a range of health effects including lung cancer. Their complex organic fraction contains genotoxic and carcinogenic compounds such as polycyclic aromatic hydrocarbons (PAHs) and their derivatives. This study evaluates the genotoxicity of the PM(10) and PM(2.5) organic extracts that were sampled in the framework of a personal exposure study in three French metropolitan areas (Paris, Rouen and Strasbourg), using the comet assay, performed on HeLa S3 cells. In each city, 60-90 non-smoking volunteers composed of two groups of equal size (adults and children) carried the personal Harvard Chempass multi-pollutant sampler during 48h along two different seasons ('hot' and 'cold'). Volunteers were selected so as to live (home and work/school) in 3 different urban sectors contrasted in terms of air pollution within each city (one highly exposed to traffic emissions, one influenced by local industrial sources, and a background urban environment). Genotoxic effects are stronger for PM(2.5) extracts than for PM(10), and greater in winter than in summer. Fine particles collected by subjects living within the traffic proximity sector present the strongest genotoxic responses, especially in the Paris metropolitan area. This work confirms the genotoxic potency of particulate matter (PM(10) and PM(2.5)) organic extracts to which urban populations are exposed.     

Elemental characterization of PM10, PM2.5 and PM1 in the town of Genoa (Italy)

The particulate matter (PM) concentration and composition, the PM10, PM2.5, PM1 fractions, were studied in the urban area of Genoa, a coastal town in the northwest of Italy. Two instruments, the continuous monitor TEOM and the sequential sampler PARTISOL, were operated almost continuously on the same site from July 2001 to September 2004. Samples collected by PARTISOL were weighted to obtain PM concentration and then analysed by PIXE (particle induced X-ray emission) and by ED-XRF (energy dispersion X-ray fluorescence), obtaining concentrations for elements from Na to Pb. Some of the filters used in the TEOM microbalance were analysed by ED-XRF to calculate Pb concentration values averaged over 7-30 d periods.     

Assessment of hospitalization and mortality from exposure to PM10 using AirQ modeling in Ilam,Iran

Environmental Science and Pollution Research - The aims of this study were to assess the health impact of PM10 on inhabitants and to investigate the trend of PM10 concentrations in Ilam, Iran, from...