Particulate Matter Forecasting Research Articles

Particulate matter forecast and prediction in Curitiba using machine learning.

Air quality is directly affected by pollutant emission from vehicles, especially in large cities and metropolitan areas or when there is no compliance check for vehicle emission standards. Particulate Matter (PM) is one of the pollutants emitted from fuel burning in internal combustion engines and remains suspended in the atmosphere, causing respiratory and cardiovascular health problems to the population. In this study, we analyzed the interaction between vehicular emissions, meteorological variables, and particulate matter concentrations in the lower atmosphere, presenting methods for predicting and forecasting PM2.5. Meteorological and vehicle flow data from the city of Curitiba, Brazil, and particulate matter concentration data from optical sensors installed in the city between 2020 and 2022 were organized in hourly and daily averages. Prediction and forecasting were based on two machine learning models: Random Forest (RF) and Long Short-Term Memory (LSTM) neural network. The baseline model for prediction was chosen as the Multiple Linear Regression (MLR) model, and for forecast, we used the naive estimation as baseline. RF showed that on hourly and daily prediction scales, the planetary boundary layer height was the most important variable, followed by wind gust and wind velocity in hourly or daily cases, respectively. The highest PM prediction accuracy (99.37%) was found using the RF model on a daily scale. For forecasting, the highest accuracy was 99.71% using the LSTM model for 1-h forecast horizon with 5 h of previous data used as input variables. The RF and LSTM models were able to improve prediction and forecasting compared with MLR and Naive, respectively. The LSTM was trained with data corresponding to the period of the COVID-19 pandemic (2020 and 2021) and was able to forecast the concentration of PM2.5 in 2022, in which the data show that there was greater circulation of vehicles and higher peaks in the concentration of PM2.5. Our results can help the physical understanding of factors influencing pollutant dispersion from vehicle emissions at the lower atmosphere in urban environment. This study supports the formulation of new government policies to mitigate the impact of vehicle emissions in large cities.

Concentration Separation Prediction Model to Enhance Prediction Accuracy of Particulate Matter

Demand for more accurate particulate matter forecasts is accumulating owing to the increased interest and issues regarding particulate matter. Incredibly low concentration particulate matter, which accounts for most of the overall particulate matter, is often underestimated when a particulate matter prediction model based on machine learning is used. This study proposed a concentration-specific separation prediction model to overcome this shortcoming. Three prediction models based on Deep Neural Network (DNN), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM), commonly used for performance evaluation of the proposed prediction model, were used as comparative models. Root mean squared error (RMSE), mean absolute percentage error (MAPE), and accuracy were utilized for performance evaluation. The results showed that the prediction accuracy for all Air Quality Index (AQI) segments was more than 80 percent in the entire concentration spectrum. In addition, the study confirmed that the over-prediction phenomenon of single neural network models concentrated in the ‘normal’ AQI region was alleviated.

Seminal Stacked Long Short-Term Memory (SS-LSTM) Model for Forecasting Particulate Matter (PM2.5 and PM10)

With increased industrialization and urbanization, sustainable smart environments are becoming more concerned with particulate matter (PM) forecasts that are based on artificial intelligence (AI) techniques. The intercorrelation between multiple pollutant components and the extremely volatile PM pattern changes are the key impediments to effective prediction. For accurate PM forecasting with the benefit of federated learning, a new architecture incorporating seminal stacked long short-term memory networks (SS-LSTM) is presented in this research. The historical data are analyzed using SS-LSTM to reveal the location-aware behavior of PM, and a new prediction model is generated that takes into account the most prevalent pollutants and weather conditions. The stacking of LSTM units adds hierarchical levels of knowledge that help to tune the forecast model with the most appropriate weighting to the external features that contribute toward PM. The suggested SS-LSTM model is compared with traditional machine learning approaches and deep learning models to see how well it performs in predicting PM2.5 and PM10. The suggested strategy outperforms all other models tested in experiments carried out for the data collected from Delhi in India.

Effectiveness of Particulate Matter Forecasting and Warning Systems within Urban Areas

The close relation between atmospheric pollution and human health has been well documented. Accordingly, various policies have been enacted worldwide to reduce and regulate air pollution, with most countries having established correlated monitoring systems. Notably in South Korea, increasing concerns about particulate matter (PM) concentrations led to the establishment of a nationwide forecasting and warning system in 2014. In this study, the PM trends in South Korea over the past decade were examined, and the correlated social issues were analyzed. In addition, the relationships between PM concentration, the forecasting–warning system, and people’s urban park use were analyzed to assess the efficacy of policy introduction. The results indicated that PM concentrations were an obstacle to outdoor activities, and the PM forecasting–warning system affected urban park use. Whereas the effects of PM forecasting and warning systems have not been sufficiently explored in practical terms in the literature, this study could be significant in proving the validity of environmental policies through the evidence including urban park visitors. This study also suggests future directions for developing PM forecasting and warning systems.

Influence of the transported Canadian wildfire smoke on the ozone and particle pollution over the Mid-Atlantic United States

The integration of observations and models can improve air quality forecasts (in particular ozone (O3) and particulate matter (PM)) for extreme events (e.g., wildfires). We present our work on the Canadian wildfire event on 6–12 June 2015 that impacted the air quality in the Mid-Atlantic region in the U.S. We use the Weather Research and Forecasting model coupled with Chemistry package (WRF-Chem), and various measurements from both ground-based and spaceborne observations, including the U.S. Environmental Protection Agency (EPA) AirNow data, the National Aeronautics and Space Administration (NASA) operated TROPospheric OZone lidar (TROPOZ), wind radar profiler, ceilometer, Moderate Resolution Imaging Spectroradiometer (MODIS), Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP). The objective is to understand the physics of the Planetary Boundary Layer (PBL) and its role on the O3 and PM forecast. The findings show that the model captured the O3 diurnal variation and PM spatial distribution both horizontally and vertically by comparing with EPA AirNow and MODIS/CALIOP observations, respectively. Wildfire smoke was transported from central Canada through Lake Michigan, passing the Ohio River Valley and down to the Baltimore-Washington D.C. metropolis. The night-time O3 mixing ratio reached 30 ppbv, while the daytime O3 mixing ratio approached larger than 100 ppbv near AirNow stations in Maryland, due to the mixing of the transported smoke into the PBL. The novel NASA TROPOZ lidar at Beltsville resolved the O3 vertical profile and the ceilometer identified the smoke intrusion at altitudes above 3.5 km, but later mixed down into the PBL and surface which was also resolved by the model. Thus, integrating both model and observations from different platforms confirms the Canadian wildfire source and transport pathway and improves the understanding of the air quality forecast during the extreme wildfire event.

Analog ensemble technique to post-process WRF-CAMx ozone and particulate matter forecasts

Post-processing techniques can provide significant improvement in the forecast skill of air quality models. In this study, the implementation of an analog-based technique to Comprehensive Air Quality Model with Extensions (CAMx) coupled with Weather Research and Forecasting (WRF) model results is examined. WRF-CAMx runs with a 2-km horizontal grid increment over Greece for one month of every season of the year 2012 (i.e., January, April, July and October). The analog ensemble (AnEn) technique attempts to improve the accuracy of ozone and particulate matter forecasts by using a method that searches for analogs in past forecasts. An optimization process that minimizes Root Mean Square Error (RMSE) metric has been used to find the best AnEn configuration. The corrected forecasts are computed with two approaches, i.e., AnEn ‘mean’ and AnEn ‘bias correction’ (AnEn-bias) approach. The methods are tested with observations from 23 surface stations for ozone, 16 stations for PM10 and 3 stations for PM2.5 for an 11-day period for each month. The results which are very similar for both techniques show an improvement of the forecast skill of all pollutants. The corrected forecasts have smaller RMSE and higher Correlation Coefficient (R). A reduction of 40 and 70% for AnEn RMSE values is found for ozone and particulate matter, respectively. For AnEn R, an improvement of 11% for ozone, 46% for PM10 and 26% for PM2.5 is estimated. These techniques are also successful in drastically reducing the mean bias of raw forecasts to close to zero.

Description and Evaluation of the Fine Particulate Matter Forecasts in the NCAR Regional Air Quality Forecasting System

This paper describes a quasi-operational regional air quality forecasting system for the contiguous United States (CONUS) developed at the National Center for Atmospheric Research (NCAR) to support air quality decision-making, field campaign planning, early identification of model errors and biases, and support the atmospheric science community in their research. This system aims to complement the operational air quality forecasts produced by the National Oceanic and Atmospheric Administration (NOAA), not to replace them. A publicly available information dissemination system has been established that displays various air quality products, including a near-real-time evaluation of the model forecasts. Here, we report the performance of our air quality forecasting system in simulating meteorology and fine particulate matter (PM2.5) for the first year after our system started, i.e., 1 June 2019 to 31 May 2020. Our system shows excellent skill in capturing hourly to daily variations in temperature, surface pressure, relative humidity, water vapor mixing ratios, and wind direction but shows relatively larger errors in wind speed. The model also captures the seasonal cycle of surface PM2.5 very well in different regions and for different types of sites (urban, suburban, and rural) in the CONUS with a mean bias smaller than 1 µg m−3. The skill of the air quality forecasts remains fairly stable between the first and second days of the forecasts. Our air quality forecast products are publicly available at a NCAR webpage. We invite the community to use our forecasting products for their research, as input for urban scale (<4 km), air quality forecasts, or the co-development of customized products, just to name a few applications.

Attention-based parallel networks (APNet) for PM2.5 spatiotemporal prediction

Urban particulate matter forecast is an important part of air pollution early warning and control management, especially the forecast of fine particulate matter (PM2.5). However, the existing PM2.5 concentration prediction methods cannot effectively capture the complex nonlinearity of PM2.5 concentration, and most of them cannot accurately simulate the temporal and spatial dependence of PM2.5 concentration at the same time. In this paper, we propose an attention-based parallel network (APNet), which can extract short-term and long-term temporal features simultaneously based on the attention-based CNN-LSTM multilayer structure to predict PM2.5 concentration in the next 72 h. Firstly, the Maximum Information Coefficient (MIC) is designed for spatiotemporal correlation analysis, fully considering the linearity, non-linearity and non-functionality between the data of each monitoring station. The potential inherent features of the input data are effectively extracted through the convolutional neural network (CNN). Then, an optimized long short-term memroy (LSTM) network captures the short-term mutations of the time series. An attention mechanism is further designed for the proposed model, which automatically assigns different weights to different feature states at different time stages to distinguish their importance, and can achieve precise temporal and spatial interpretability. In order to further explore the long-term time features, we propose a Bi-LSTM parallel module to extract the periodic characteristics of PM2.5 concentration from both previous and posterior directions. Experimental results based on a real-world dataset indicates that the proposed model outperforms other existing state-of-the-art methods. Moreover, evaluations of recall (0.790), precision (0.848) (threshold: 151 μg/m3) for 72 h prediction also verify the feasibility of our proposed model. The methodology can be used for predicting other multivariate time series data in the future.

Neural-Based Ensembles for Particulate Matter Forecasting

The air pollution caused by particulate matter (PM) has become a public health issue due to the risks to human life and the environment. The PM concentration in the air causes haze and affects the lungs and the heart, leading to reduced visibility, allergic reactions, pneumonia, asthma, cardiopulmonary diseases, lung cancer, and even death. In this context, the development of systems for monitoring, forecasting, and controlling emissions plays an important role. The literature about forecasting systems based on Artificial Neural Networks (ANNs) ensembles has been highlighted regarding statistical accuracy and efficiency. In this article, trainable and non-trainable combination methods are used for PM10 and PM2.5 (particles with an aerodynamic diameter less than 10 and 2.5 micrometers, respectively) time series forecasting for eight different locations, in Finland and Brazil, for different periods. Trainable ensembles based on ANNs, linear regression, and Copulas are compared with non-trainable combinations (mean and median), single ANNs, and linear statistical approaches. Different models are considered so far, including Autoregressive model (AR), Autoregressive and Moving Average Model (ARMA), Infinite Impulse Response Filters (IIR), Multilayer Perceptron (MLP), Radial Basis Function Networks (RBF), Extreme Learning Machines (ELM), Echo State Networks (ESN), and Adaptive Network Fuzzy Inference System (ANFIS). The use of ANNs ensembles, mainly combined with MLP, leads to a better one step ahead forecasting performance. The use of robust air pollution forecasting tools is prime to assist governments in managing air pollution issues like hospital collapse during adverse air quality situations. In this sense, our study is indirectly related to the following United Nations sustainable development goals: SDG 3 - good health and well-being and SDG 11 - sustainable cities and communities.

Demand Response Program Expansion in Korea through Particulate Matter Forecasting Based on Deep Learning and Fuzzy Inference

The increase in ambient particulate matter (PM) is affecting not only our daily life but also various industries. To cope with the issue of PM, which has been detrimental to the population of megacities, an advanced demand response (DR) program is established by Korea Power Exchange (KPX) to supplement existing policies in Korea. Ironically, however, DR programs have been launched hurriedly, creating problems for several stakeholders such as local governments, market operators, and DR customers. As an alternative, a method for predicting and categorizing the PM through deep learning and fuzzy inference is suggested in this study. The simulation results based on Seoul data show that the proposed model can overcome the problems related to current DR programs and policy loopholes and can provide improvements for some stakeholders. However, the proposed model also has some limitations, which require an in-depth policy consideration or an incentive system for power generation companies.

Aspiration ratio of a double-shrouded probe under low pressure conditions in troposphere

Particulate matter forecasting is based on data measured at several monitoring stations fixed on the ground. The most effective way to improve forecast accuracy is to obtain measurements over a wid...

Urban PM2.5 Concentration Prediction via Attention-Based CNN–LSTM

Urban particulate matter forecasting is regarded as an essential issue for early warning and control management of air pollution, especially fine particulate matter (PM2.5). However, existing methods for PM2.5 concentration prediction neglect the effects of featured states at different times in the past on future PM2.5 concentration, and most fail to effectively simulate the temporal and spatial dependencies of PM2.5 concentration at the same time. With this consideration, we propose a deep learning-based method, AC-LSTM, which comprises a one-dimensional convolutional neural network (CNN), long short-term memory (LSTM) network, and attention-based network, for urban PM2.5 concentration prediction. Instead of only using air pollutant concentrations, we also add meteorological data and the PM2.5 concentrations of adjacent air quality monitoring stations as the input to our AC-LSTM. Hence, the spatiotemporal correlation and interdependence of multivariate air quality-related time-series data are learned by the CNN–LSTM network in AC-LSTM. The attention mechanism is applied to capture the importance degrees of the effects of featured states at different times in the past on future PM2.5 concentration. The attention-based layer can automatically weigh the past feature states to improve prediction accuracy. In addition, we predict the PM2.5 concentrations over the next 24 h by using air quality data in Taiyuan city, China, and compare it with six baseline methods. To compare the overall performance of each method, the mean absolute error (MAE), root-mean-square error (RMSE), and coefficient of determination (R2) are applied to the experiments in this paper. The experimental results indicate that our method is capable of dealing with PM2.5 concentration prediction with the highest performance.

Validation of linear, nonlinear, and hybrid models for predicting particulate matter concentration in Tehran, Iran

Information on particulate matter forecast is significant as it allows residents to manage its undesirable effects. For the purpose of predicting PM10 concentration in the air of Tehran, various models were used, including (i) a linear model (multiple liner regression, MLR), (ii) two hybrid models (Adaptive Neuro-Fuzzy Inference System, ANFIS as well as ensemble empirical mode decomposition and general regression neural network, EEMD-GRNN), and (iii) a nonlinear model (multi-layer perceptron, MLP). The output variable in these models was the measure of suspended particles of PM10 while the predictor variables were the information on air quality which consisted of CO, NO2, O3, PM10 of the previous day, PM2.5, and SO2 as well as meteorological data which included average atmospheric pressure (AP), average maximum temperature (Max T), average minimum temperature (Min T), daily relative humidity level of the air (RH), daily total precipitation (TP), and daily wind speed (WS) for the year 2016 in Tehran. Analysis of the data revealed that in comparison with the results of MLR and MLP, ANFIS obtained the most accurate output (R2 = 0.97, root mean square error (RMSE) = 1.0713, and mean absolute error (MAE) = 0.6111) for the training phase and (R2 = 0.89, RMSE = 3.6165, and MAE = 2.8993) the testing phase. However, the hybrid models which were used in the current study had almost similar prediction results. As it can be concluded, in comparison with linear and nonlinear models, hybrid models turn out to have higher accuracy in predicting PM10 concentration.

Dynamic ensemble mechanisms to improve particulate matter forecasting

Respirable solid particles and liquid droplets suspended in the air, known as particulate matter (PM), may have a significant impact on human health, urban infrastructure, and natural and agricultural systems. The adverse effects of PM have raised public concern, especially in heavily polluted areas in the world, making it imperative the development of strategies to keep the concentration levels of these pollutants below harmful thresholds. Traditional machine learning approaches have been used to forecast PM concentrations. However, complex chemical processes may be involved in the composition of PM in the atmosphere and influenced by many meteorological parameters. Thus, underlying data distributions of PM data, uninterruptedly collected, may evolve over time. This phenomenon, known as concept drift, implies an important challenge for traditional machine learning techniques since they do not have mechanisms to handle changes on data distribution at the running time, thus limiting their forecasting capabilities. The overall goal of this work is to evaluate whether the incorporation of mechanisms to deal with concept drift, together with online sequential learning approaches, can improve the accuracy of PM forecasting. To do so, new mechanisms that enable online dynamic ensembles to handle and retain knowledge from different concepts for more time were proposed and adapted to EOS and DOER algorithms, resulting in three approaches: EOS-rank, EOS-D and DOER-rank. These ensemble strategies, which were based on Online Sequential Extreme Learning Machines (OS-ELM), were compared with five algorithms from the literature. To evaluate their performance, real-world and artificial datasets, with known dynamic behaviors, and PM concentration datasets from different cities of the State of São Paulo, Brazil, were used in the experiments. The obtained results showed that the proposed approaches can handle dynamic environments with different rates of drift and that EOS-rank was capable of outperforming most approaches from the literature in scenarios with higher rates of drift. The results also indicate that PM data distributions slowly evolve over time and, consequently, the proposed mechanisms that keep information of past concepts and slowly adapt the ensemble tend to present better results when applied to forecast PM concentration.

Autoencoder-based deep belief regression network for air particulate matter concentration forecasting

Particulate matter (PM) is one of the most significant air pollutants in recent decades that has tremendous negative effects on the ambient air quality and the public health. Accurate PM forecasting provides a possibility for establishing an early warning system. In this paper, a deep feature learn ing architecture, i.e., autoencoder-based deep belief regression network (AE-based DBRN), is introduced and utilized to forecast the daily PM concentrations (PM2.5 and PM10). Prior to establishing this model, Pearson correlation analysis is applied to look for the possible input-output mapping, where the input candidate variables contain seven meteorological parameters and PM concentrations within one-day ahead, and the output variables are the local PM forecasts. The addressed model was evaluated by the dataset in the period of 28/10/2013 to 31/8/2016 in Chongqing municipality of China. Moreover, two shallow models, feed forward neural network and least squares support vector regression, were employed for the comparison. The results indicate that the AE-based DBRN model has remarkable better performances among the comparison models in terms of mean absolute percentage error (PM2.5 21.092%, PM10 19.474%), root mean square error (PM2.5 8.600μg/m3, PM10 11.239μg/m3) and correlation coefficient criteria (PM2.5 0.840, PM10 0.826).

Recursive neural network model for analysis and forecast of PM10 and PM2.5

Atmospheric particulate matter (PM) is one of the pollutant that may have a significant impact on human health. Data collected during three years in an urban area of the Adriatic coast are analysed using three models: a multiple linear regression model, a neural network model with and without recursive architecture. Measured meteorological parameters and PM10 concentration are used as input to forecast the daily averaged concentration of PM10 from one to three days ahead. All simulations show that the neural network with recursive architecture has better performances compared to both the multiple linear regression model and the neural network model without the recursive architecture. Results of PM forecasts are compared with the air quality limits for health protection to test the performance as operational tool. The inclusion of carbon monoxide (CO) concentration as further input parameter in the model, has been evaluated in terms of forecast improvements. Finally, all models are used to forecast the PM2.5 concentration, using as input the meteorological data, the PM10 and CO concentration, to simulate the situation when PM2.5 is not observed. The comparison between observed and forecasted PM2.5 shows that the neural network is able to forecast the PM2.5 concentrations even if PM2.5 is not included among the input parameters.

GIST-PM-Asia v1: development of a numerical system to improve particulate matter forecasts in South Korea using geostationary satellite-retrieved aerosol optical data over Northeast Asia

Abstract. To improve short-term particulate matter (PM) forecasts in South Korea, the initial distribution of PM composition, particularly over the upwind regions, is primarily important. To prepare the initial PM composition, the aerosol optical depth (AOD) data retrieved from a geostationary equatorial orbit (GEO) satellite sensor, GOCI (Geostationary Ocean Color Imager) which covers a part of Northeast Asia (113–146° E; 25–47° N), were used. Although GOCI can provide a higher number of AOD data in a semicontinuous manner than low Earth orbit (LEO) satellite sensors, it still has a serious limitation in that the AOD data are not available at cloud pixels and over high-reflectance areas, such as desert and snow-covered regions. To overcome this limitation, a spatiotemporal-kriging (STK) method was used to better prepare the initial AOD distributions that were converted into the PM composition over Northeast Asia. One of the largest advantages in using the STK method in this study is that more observed AOD data can be used to prepare the best initial AOD fields compared with other methods that use single frame of observation data around the time of initialization. It is demonstrated in this study that the short-term PM forecast system developed with the application of the STK method can greatly improve PM10 predictions in the Seoul metropolitan area (SMA) when evaluated with ground-based observations. For example, errors and biases of PM10 predictions decreased by ∼ 60 and ∼ 70 %, respectively, during the first 6 h of short-term PM forecasting, compared with those without the initial PM composition. In addition, the influences of several factors on the performances of the short-term PM forecast were explored in this study. The influences of the choices of the control variables on the PM chemical composition were also investigated with the composition data measured via PILS-IC (particle-into-liquid sampler coupled with ion chromatography) and low air-volume sample instruments at a site near Seoul. To improve the overall performances of the short-term PM forecast system, several future research directions were also discussed and suggested.

Analysis and forecasting of the particulate matter (PM) concentration levels over four major cities of China using hybrid models

The analysis and forecasting of PM concentrations play a significant role in regulatory planning on the reduction and control of PM emission and precautionary strategies. However, accurate PM forecasting, which is needed to establish an early warning system, is still a huge challenge and a critical issue. Determining how to address the accurate forecasting problem becomes an even more significant and urgent task. Based on gray correlation analysis (GCA), Ensemble Empirical Mode Decomposition (EEMD), Cuckoo search (CS) and Back-propagation artificial neutral networks (BPANN), this paper proposes the CS-EEMD-BPANN model for forecasting PM concentrations. Prior to establishing this model, gray correlation has been uniquely used to search for possible predictors of PM among other air pollutants (CO, NO2, O3 and SO2) and meteorological environments (wind speed, wind direction, temperature, humidity and pressure). The proposed method was investigated in four major cities of China (Beijing, Shanghai, Guangzhou and Lanzhou) with different characteristics of climatic, terrain and emission sources. The results of the gray correlation analysis indicate that CO, NO2 and SO2 are more related to PM and that the incorporation of these predictors can significantly improve the model performance predictability, suggesting the effectiveness of our developed method.

Assessment of the wintertime performance of developmental particulate matter forecasts with the Eta‐Community Multiscale Air Quality modeling system

It is desirable for local air quality agencies to accurately forecast tropospheric PM2.5 concentrations to alert the sensitive population of the onset, severity, and duration of unhealthy air and to encourage the public and industry to reduce emissions‐producing activities. Since elevated particulate matter concentrations are encountered throughout the year, the accurate forecast of the day‐to‐day variability in PM2.5 and constituent concentrations over annual cycles poses considerable challenges. In efforts to characterize forecast model performance during different seasons, PM2.5 forecast simulations with the Eta‐Community Multiscale Air Quality system are compared with measurements from a variety of regional surface networks, with special emphasis on performance during the winter period. The analysis suggests that while the model can capture the average spatial trends and dynamic range in PM2.5 and constituent concentrations measured at individual sites, significant variability occurs on a day‐to‐day basis both in the measurements and the model predictions, which are generally not well correlated when paired both in space and time. Systematic overpredictions in regional PM2.5 forecasts during the cool season are noted through comparisons with measurements from different networks. The overpredictions are typically more pronounced at urban locations, with larger errors at the higher concentration range. Variability in aerosol sulfate concentrations were captured well, as well as the relative amounts of sulfur (IV) and sulfur (VI). The mix of carbon sources as represented by the ratio of organic to elemental carbon is captured well in the southeastern United States, but the total carbonaceous aerosol mass is underestimated. On average, during the wintertime the largest overpredictions among individual PM2.5 constituents were noted for the “other” category which predominantly represents primary‐emitted trace elements in the current model configuration. The systematic errors in model predictions of both total PM2.5 and its constituents during the winter period are found to arise from a combination of uncertainties in the magnitude and spatial and temporal allocation of primary PM2.5 emissions, current uncertainties in the estimation of chemical production pathways for secondary constituents (e.g., NO3−), and the representation of the impacts of boundary layer mixing on simulated concentrations, especially during nighttime conditions.