Source apportionment of fine particulate matter by clustering single-particle data: tests of receptor model accuracy

Single particle mass spectral signatures from on-road and non-road vehicle exhaust particles and their application in refined source apportionment using deep learning

With advances in vehicle emission control technology, updating source profiles to meet the current requirements of source apportionment has become increasingly crucial. In this study, on-road and non-road vehicle particles were collected, and then the chemical compositions of individual particles were analyzed using single particle aerosol mass spectrometry. The data were grouped using an adaptive resonance theory neural network to identify signatures and establish a mass spectral database of mobile sources. In addition, a deep learning-based model (DeepAerosolClassifier) for classifying aerosol particles was established. The objective of this model was to accomplish source apportionment. During the training process, the model achieved an accuracy of 98.49 % for the validation set and an accuracy of 93.36 % for the testing set. Regarding the model interpretation, ideal spectra were generated using the model, verifying its accurate recognition of the characteristic patterns in the mass spectra. In a practical application, the model performed hourly source apportionment at three specific field monitoring sites. The effectiveness of the model in field measurement was validated by combining traffic flow and spatial information with the model results. Compared with other machine learning methods, our model achieved highly automated source apportionment while eliminating the need for feature selection, and it enables end-to-end operation. Thus, in the future, it can be applied in refined and online source apportionment of particulate matter.

Self-feedback LSTM regression model for real-time particle source apportionment

Atmospheric particulate matter pollution has attracted much wider attention globally. In recent years, the development of atmospheric particle collection techniques has put forwards new demands on the real-time source apportionments techniques. Such demands are summarized, in this paper, as how to set up new restraints in apportionment and how to develop a non-linear regression model to process complicated circumstances, such as the existence of secondary source and similar source. In this study, we firstly analyze the possible and potential restraints in single particle source apportionment, then propose a novel three-step self-feedback long short-term memory (SF-LSTM) network for approximating the source contribution. The proposed deep learning neural network includes three modules, as generation, scoring and refining, and regeneration modules. Benefited from the scoring modules, SF-LSTM implants four loss functions representing four restraints to be followed in the apportionment, meanwhile, the regeneration module calculates the source contribution in a non-linear way. The results show that the model outperforms the conventional regression methods in the overall performance of the four evaluation indicators (residual sum of squares, stability, sparsity, negativity) for the restraints. Additionally, in short time-resolution analyzing, SF-LSTM provides better results under the restraint of stability.

Single particle mass spectral signatures from vehicle exhaust particles and the source apportionment of on-line PM2.5 by single particle aerosol mass spectrometry

In order to accurately apportion the many distinct types of individual particles observed, it is necessary to characterize fingerprints of individual particles emitted directly from known sources. In this study, single particle mass spectral signatures from vehicle exhaust particles in a tunnel were performed. These data were used to evaluate particle signatures in a real-world PM_2.5 apportionment study. The dominant chemical type originating from average positive and negative mass spectra for vehicle exhaust particles are EC species. Four distinct particle types describe the majority of particles emitted by vehicle exhaust particles in this tunnel. Each particle class is labeled according to the most significant chemical features in both average positive and negative mass spectral signatures, including ECOC, NaK, Metal and PAHs species. A single particle aerosol mass spectrometry (SPAMS) was also employed during the winter of 2013 in Guangzhou to determine both the size and chemical composition of individual atmospheric particles, with vacuum aerodynamic diameter (d_va) in the size range of 0.2-2μm. A total of 487,570 particles were chemically analyzed with positive and negative ion mass spectra and a large set of single particle mass spectra was collected and analyzed in order to identify the speciation. According to the typical tracer ions from different source types and classification by the ART-2a algorithm which uses source fingerprints for apportioning ambient particles, the major sources of single particles were simulated. Coal combustion, vehicle exhaust, and secondary ion were the most abundant particle sources, contributing 28.5%, 17.8%, and 18.2%, respectively. The fraction with vehicle exhaust species particles decreased slightly with particle size in the condensation mode particles.

Trace element concentration in fine particulate matter (PM2.5) and their bioavailability in different microenvironments in Agra, India: a case study

Exposure to airborne particulate matter results in the deposition of millions of particle in the lung; consequently, there is need for monitoring them particularly in indoor environments. Case study was conducted in three different microenvironments, i.e., urban, rural and roadside to examine the elemental bioavailability in fine particulate matter and its potential health risk. The samples were collected on polytetrafluoroethylene filter paper with the help of fine particulate sampler during August-September, 2012. The average mass concentration of PM2.5 was 71.23 µg m(-3) (rural), 45.33 µg m(-3) (urban) and 36.71 µg m(-3) (roadside). Elements in PM2.5 were analyzed by inductively coupled plasma atomic emission spectroscopy. Percentage bioavailability was determined to know the amount of soluble fraction that is actually taken across the cell membrane through inhalation pathway. Cadmium and lead were found to have cancer risk in a risk evaluation using an Integrated Risk Information system.

Characterization of aerosol optical properties, chemical composition and mixing states in the winter season in Shanghai, China

Physical and chemical properties of ambient aerosols at the single particle level were studied in Shanghai from December 22 to 28, 2009. A Cavity-Ring-Down Aerosol Extinction Spectrometer (CRD-AES) and a nephelometer were deployed to measure aerosol light extinction and scattering properties, respectively. An Aerosol Time-of-Flight Mass Spectrometer (ATOFMS) was used to detect single particle sizes and chemical composition. Seven particle types were detected. Air parcels arrived at the sampling site from the vicinity of Shanghai until mid-day of December 25, when they started to originate from North China. The aerosol extinction, scattering, and absorption coefficients all dropped sharply when this cold, clean air arrived. Aerosol particles changed from a highly aged type before this meteorological shift to a relatively fresh type afterwards. The aerosol optical properties were dependent on the wind direction. Aerosols with high extinction coefficient and scattering Ångström exponent (SAE) were observed when the wind blew from the west and northwest, indicating that they were predominantly fine particles. Nitrate and ammonium correlated most strongly with the change in aerosol optical properties. In the elemental carbon/organic carbon (ECOC) particle type, the diurnal trends of single scattering albedo (SSA) and elemental carbon (EC) signal intensity had a negative correlation. We also found a negative correlation (r=-0.87) between high mass-OC particle number fraction and the SSA in a relatively clean period, suggesting that particulate aromatic components might play an important role in light absorption in urban areas.

Source apportionment for sediment PAHs using hybrid genetic pattern search treatment of a chemical mass balance receptor model: application to the Pearl River Delta region, China

In order to solve the collinear problem and improve the estimation accuracy of the chemical mass balance (CMB) model which can be essentially regarded as a constrained optimization process, in this study, a hybrid genetic pattern search algorithm (HGPS) was proposed and applied to apportion the source contributions for sediment polycyclic aromatic hydrocarbons (PAHs) in the Pearl River Delta (PRD) region, China. Simulation results with developed synthetic datasets indicated that the estimated source contributions by HGPS were more close to the true values than CMB8.2. Utilizing the HGPS-CMB, residential coal and traffic tunnel were apportioned as the major sources of sediment PAHs in the PRD region. For freshwater surface sediments, the average contribution from residential coal ranged from 32 to 55%, and traffic tunnel ranged from 13 to 33%, while the major sources for marine sediments were traffic tunnel (10 ~ 56%). These results provide information for developing better PAH pollution control strategies for the PRD.

Analysis of atmospheric aerosols

Aerosols represent an important component of the Earth's atmosphere. Because aerosols are composed of solid and liquid particles of varying chemical complexity, size, and phase, large challenges exist in understanding how they impact climate, health, and the chemistry of the atmosphere. Only through the integration of field, laboratory, and modeling analysis can we begin to unravel the roles atmospheric aerosols play in these global processes. In this article, we provide a brief review of the current state of the science in the analysis of atmospheric aerosols and some important challenges that need to be overcome before they can become fully integrated. It is clear that only when these areas are effectively bridged can we fully understand the impact that atmospheric aerosols have on our environment and the Earth's system at the level of scientific certainty necessary to design and implement sound environmental policies.

Instrumentation, data evaluation and quantification in on-line aerosol mass spectrometry

On-line micro- and nanoparticle mass spectrometry has evolved into a prominent analytical method for the characterization of airborne particles, particle populations and aerosols over the recent years, driven by essential developments in instrumentation, data evaluation and validation. In this tutorial, the fundamental aspects of the technology and methodology for qualitative and quantitative on-line aerosol particle analysis are discussed. Specific properties of the on-line mass spectrometric instrumentation for particle analysis are described, combined with a discussion of basic differences of the instruments and demands for future improvements of instruments and data analysis techniques. Optimized technology and methodology in particle analysis is expected to lead to essential growth of the knowledge and to quality improvement of the description of atmospheric processes and health effects in the future.

Seasonal trends in coarse and fine particle sources in Delhi by the chemical mass balance receptor model

A study of the source contribution of atmospheric particulate matter and associated heavy metal concentrations using chemical mass balance model Version 8 (CMB8) in coarse and fine size mode has been carried out for the city of Delhi. Urban particles were collected using a five-stage impactor at six sites in three different seasons, viz. winter, summer and monsoon in the year 2001. Five samples from each site in each season were collected. The results obtained indicate the dominance of vehicular pollutants in fine size mode, whilst the contribution in coarse mode to some extent is site specific but largely due to vehicular pollution and, soil and crustal dust. Seasons also play an important role but in coarse size fraction only.

Chemical Fractionation of Ellements in Airborne Particulate Matter: Primary Results on PM10 and PM2.5 Samples in the Lazio Region (Central Italy)

This paper describes how a two-step chemical fractionation method that allows the determination of 17 elements in airborne particulate matter, has been applied to a monitoring campaign of PM10 and PM2.5 in the Lazio region (Italy). This method involved an extraction in a pH buffered aqueous solution followed by a microwave-assisted acid digestion of the residue. With respect to the determination of the total elemental contents, the evaluation of a soluble fraction provides more reliable information on the presence and of the destiny of heavy metals in the environment. Furthermore, the pH buffered extraction conditions chosen, rendered the results independent of the intrinsic acidity of the collected samples and, although the chemical fractionation has a purely operational function, it facilitates the study of the relationship between the distribution of solubility and the different emission sources. Results are discussed in relation to the different concentration and the different degrees of solubility of the elements observed in two sampling sites; one in an urban and one in a rural environment. Since in-parallel sampling of PM2.5 and PM10 were performed in both sites, the influence of particle size is also discussed. Behaviour of some tracers deriving from both vehicular traffic, with particular attention to re-suspended road dusts, and naturally generated particulate matter, such as marine aerosol and Saharian dust, are discussed.

Determination of single particle mass spectral signatures from light-duty vehicle emissions

In this study, 28 light-duty gasoline vehicles (LDV) were operated on a chassis dynamometer at the California Air Resources Board Haagen-Smit Facility in El Monte, CA. The mass spectra of individual particles emitted from these vehicles were measured using aerosol time-of-flight mass spectrometry (ATOFMS). A primary goal of this study involves determining representative size-resolved single particle mass spectral signatures that can be used in future ambient particulate matter source apportionment studies. Different cycles were used to simulate urban driving conditions including the federal testing procedure (FTP), unified cycle (UC), and the correction cycle (CC). The vehicles were selected to span a range of catalytic converter (three-way, oxidation, and no catalysts) and engine technologies (vehicles models from 1953 to 2003). Exhaust particles were sampled directly from a dilution and residence chamber system using particle sizing instruments and an ATOFMS equipped with an aerodynamic lens (UF-ATOFMS) analyzing particles between 50 and 300 nm. On the basis of chemical composition, 10 unique chemical types describe the majority of the particles with distinct size and temporal characteristics. In the ultrafine size range (between 50 and 100 nm), three elemental carbon (EC) particle types dominated, all showing distinct EC signatures combined with Ca, phosphate, sulfate, and a lower abundance of organic carbon (OC). The relative fraction of EC particle types decreased as particle size increased with OC particles becoming more prevalent above 100 nm. Depending on the vehicle and cycle, several distinct OC particle types produced distinct ion patterns, including substituted aromatic compounds and polycyclic aromatic hydrocarbons (PAH), coupled with other chemical species including ammonium, EC, nitrate, sulfate, phosphate, V, and Ca. The most likely source of the Ca and phosphate in the particles is attributed to the lubricating oil. Significant variability was observed in the chemical composition of particles emitted within the different car categories as well as for the same car operating under different driving conditions. Two-minute temporal resolution measurements provide information on the chemical classes as they evolved during the FTP cycle. The first two minutes of the cold start produced more than 5 times the number of particles than any other portion of the cycle, with one class of ultrafine particles (EC coupled with Ca, OC, and phosphate) preferentially produced. By number, the three EC with Ca classes (which also contained OC, phosphate, and sulfate) were the most abundant classes produced by the nonsmoking vehicles. The smoker category produced the highest number of particles, with the dominant classes being OC comprised of substituted monoaromatic compounds and PAHs, coupled with Ca and phosphate, thus suggesting used lubricating oil was associated with many of these particles. These studies show, by number, EC particles dominate gasoline emissions in the ultrafine size range particularlyforthe lowest emitting newer vehicles, suggesting the EC signature alone cannot be used as a unique tracer for diesels. This represents the first report of high time- and size-resolved chemical composition data showing the mixing state of nonrefractory elements in particles such as EC for vehicle emissions during dynamometer source testing.

Source identification of fine particles in Washington, DC, by expanded factor analysis modeling

An expanded factor analysis model (ME-2) that is capable of taking into account the influence of independent variables such as wind speed, wind direction, time of year and other variables of the measured fine particle matter (PM2.5) concentration data was utilized for identifying sources of airborne pollutants and providing quantitative estimations of the contribution of each source. The chemical composition data used in this study were obtained from PM2.5 samples collected using the Interagency Monitoring of Protected Visual Environments samplers from August 1999 to December 2001 at an urban monitoring site in Washington, DC. The expanded model has been applied to two different data sets based on the particulate carbon variables. Such an approach had been successfully applied previously and provided improved source resolution in simulated and ambient concentration data. Initially, total OC and EC were used in the expanded model and were compared to the results using conventional positive matrix factorization that had been done previously using the individual carbon fractions data. In the other expanded model analysis, the eight carbon fractions were used during the modeling in order to ascertain if additional source information could be extracted from the data. In both cases, it was possible to separate diesel from spark-ignition vehicles. The use of the individual carbon fractions in the model provides information on what appears to be secondary organic aerosol formation.

Evaluation of elemental carbon as a marker for diesel particulate matter

Elemental carbon (EC) in atmospheric particulate matter originates from a broad range of sources in many urban locations. As health and air quality studies are using elemental carbon measurements to better understand the impact of diesel engines and other combustion sources, there is a great need to clearly understand the relative source contributions to EC concentrations in the atmosphere. However, the different analytical techniques currently used to measure EC do not show good agreement for many particulate matter samples. To this end, studies that use EC as a tracer and integrate different analytical techniques for EC can significantly bias estimates of source contributions to atmospheric particulate matter. In addition, source attribution studies that do not properly address all sources of EC in the atmosphere can also lead to inaccuracies and biases. To better understand the use of EC as a tracer, a review of the distribution of EC in the primary particulate matter emissions from air pollution sources using different analytical methods is discussed. A review of previous apportionment studies of particulate matter is presented to elucidate the fraction of EC that results from emissions from diesel engines in urban locations. These results demonstrate that EC is not a unique tracer for diesel exhaust and efforts to utilize EC as an indicator of diesel exhaust must properly address other sources of EC as well as utilize a consistent measurement technique for EC when comparing source and ambient EC measurements to avoid significant biases.

Receptor modeling application framework for particle source apportionment

Receptor models infer contributions from particulate matter (PM) source types using multivariate measurements of particle chemical and physical properties. Receptor models complement source models that estimate concentrations from emissions inventories and transport meteorology. Enrichment factor, chemical mass balance, multiple linear regression, eigenvector. edge detection, neural network, aerosol evolution, and aerosol equilibrium models have all been used to solve particulate air quality problems, and more than 500 citations of their theory and application document these uses. While elements, ions, and carbons were often used to apportion TSP, PM10, and PM2.5 among many source types, many of these components have been reduced in source emissions such that more complex measurements of carbon fractions, specific organic compounds, single particle characteristics, and isotopic abundances now need to be measured in source and receptor samples. Compliance monitoring networks are not usually designed to obtain data for the observables, locations, and time periods that allow receptor models to be applied. Measurements from existing networks can be used to form conceptual models that allow the needed monitoring network to be optimized. The framework for using receptor models to solve air quality problems consists of: (1) formulating a conceptual model; (2) identifying potential sources; (3) characterizing source emissions; (4) obtaining and analyzing ambient PM samples for major components and source markers; (5) confirming source types with multivariate receptor models; (6) quantifying source contributions with the chemical mass balance; (7) estimating profile changes and the limiting precursor gases for secondary aerosols; and (8) reconciling receptor modeling results with source models, emissions inventories, and receptor data analyses.

Chemically-assigned classification of aerosol mass spectra

An Algorithm for Discriminant Analysis of Mass Spectra--ADAMS--was created that classified aerosol mass spectra into dominant chemically-assigned classes, and grouped rare cases in an outlier class. ADAMS was trained with ambient particulate matter (PM) mass spectra, and then validated through classification tests on known spectra with random noise added, various standard chemicals, and salt-spiked polystyrene latex microspheres. The classification results showed that ADAMS gave a reasonable chemical description of the particle populations. In contrast to adaptive resonance theory (ART-2a) classification, ADAMS could be trained to be advantageously sensitive or insensitive to selected chemical markers. Application of ADAMS to Toronto ambient PM and diesel PM (NIST 2975) demonstrated that these samples could be well described, with a low proportion of the cases falling into the outlier class. Such an algorithm may find application for source-receptor modeling of aerosol mass spectra.

Evaluation of an air quality model for the size and composition of source-oriented particle classes

Air quality model predictions of the size and composition of atmospheric particle classes are evaluated by comparison with aerosol time-of-flight mass spectrometry (ATOFMS) measurements of single-particle size and composition at Long Beach and Riverside, CA, during September 1996. The air quality model tracks the physical diameter, chemical composition, and atmospheric concentration of thousands of representative particles from different emissions classes as they are transported from sources to receptors while undergoing atmospheric chemical reactions. In the model, each representative particle interacts with a common gas phase but otherwise evolves separately from all other particles. The model calculations yield an aerosol population, in which particles of a given size may exhibit different chemical compositions. ATOFMS data are adjusted according to the known particle detection efficiencies of the ATOFMS instruments, and model predictions are modified to simulate the chemical sensitivities and compositional detection limits of the ATOFMS instruments. This permits a direct, semiquantitative comparison between the air quality model predictions and the single-particle ATOFMS measurements to be made. The air quality model accurately predicts the fraction of atmospheric particles containing sodium, ammonium, nitrate, carbon, and mineral dust, across all particle sizes measured by ATOFMS at the Long Beach site, and in the coarse particle size range (Da > or = 1.8 microm) atthe Riverside site. Given thatthis model evaluation is very likely the most stringent test of any aerosol air quality model to date, the model predictions show impressive agreement with the single-particle ATOFMS measurements.