In this study we investigated the concentration and composition of particulate matter (PM2.5) in the New York City subway system. Realtime measurements, at a 1-s cadence, and gravimetric measurements were performed inside train cars along 300 km of nine subway lines, as well as on 333 platforms on 287 subway stations. The mean (±SD) PM2.5 concentration on the underground platforms was 142 ± 69 μg/m3 versus 29 ± 20 μg/m3 for aboveground stations. The average concentrations inside train cars were 88 ± 14 μg/m3 when traveling through underground tunnels and platforms and 29 ± 31 μg/m3 while on aboveground tracks. The particle composition analysis of filtered samples was done using X-ray fluorescence (XRF), revealing that iron made up approximately 43% of the total PM2.5 mass on station platforms, approximately 126 times higher than the outdoor ambient iron concentration. Other trace elements include silicon, sulfur, copper, nickel, aluminum, calcium, barium, and manganese. Considering the very high iron content, the comparative analysis of the measured concentration versus the standards set by the Environmental Protection Agency (US EPA) is not appropriate since those limits are largely based on particulate matter from fossil fuel combustion. Health impact analysis of inhalation of iron-based particles is needed to contextualize the results presented here.

Technological and industrial developments have resulted in air pollution, which can have severe health implications, particularly in the form of respiratory disease caused by particulate matter (PM). This study aims to develop a mask filtration measurement system using a mannequin simulator. The PM2.5 sensors are utilized to measure and compare pollution levels when a mask is worn versus when it is not. The test equipment is designed to collect data while driving a motorcycle outdoors, providing an accurate representation of pollution exposure to a motorcyclist. The masks tested include N95, cloth, and KN95, and both static and mobile testing is conducted. Results showed that indoor static tests filtered PM2.5 up to 80% for both N95 and cloth masks, while outdoor filtration of the N95 mask was only 72%. This difference highlights the need for a new way of evaluating mask effectiveness, and the study's mannequin-based measurements provide an effective method. The masks tested in this study demonstrated effective filtration ranging from 60% to 85%, depending on the mask type.

Preheating combustion can effectively reduce NOx concentration, but the formation characteristics of PM0.4 in preheating combustion is unclear. In this study, the emission of PM0.4 in preheating combustion was reduced by optimizing the distribution of excess air coefficient. The mass-based particle size distribution (MPSD) of PM10, mass yield of PM0.4, elemental compositions and morphologies of PM, and the number-based particle size distribution (NPSD) were analyzed to discuss the effect of αp on PM formation. The results showed that in preheating combustion, with an increase of αp, the mass yield of PM0.4 first decreased and then increased. The mass yield of PM0.4 is minimum when αp is 0.6. Preheating combustion can simultaneously reduce PM0.4 mass yield and NOx concentration. The relationship between NOx concentration and αp is similar to the relationship between the mass yield of PM0.4 and αp. The MPSD of PM0.4 is multimodal and unimodal when αp is less than 1 and larger than 1, respectively. Particle volume concentration of PM0.4 indicates that the multimodal and unimodal of PM0.4 is a result of evolution of particles with different characteristics of gas-to-particle transformation. The effect of αp on the transformation of inorganic element from coal particles to PM0.4 is S > Fe ≈ Ca ≈ Mg > Na > Si. αp affects the mass yield of PM0.4 by affecting the combustion intensity of coal/char particles in preheating furnace and combustion furnace. When αp is 0.6, the combustion intensity of coal particles in preheating furnace and combustion furnace minimize the vaporization of inorganic elements.

Recent years have seen an increase in regional ozone (O3) pollution in China. This study explored the interannual variability in daily maximum 8-h average O3 concentration (MDA8-O3) in Northeast China (NEC) and its three subregions (Liaoning, Jilin, and Heilongjiang) during 2013–2021, and identified the key meteorological drivers underlying the observed variability. The Kolmogorov–Zurbenko filtering technique was applied to a stepwise multiple linear regression model to decompose the meteorological and anthropogenic contributions to annual MDA8-O3 trends. The results showed that the spatiotemporal variation of MDA8-O3 in NEC is characterized by a high–south and low–north pattern with an MDA8-O3 hotspot in the Bohai Rim area. Over the 9-year study period, a significant increasing trend (∼2.5% a−1, P < 0.05) in the regional mean of the annual 90th percentile of MDA8-O3 was detected across NEC. This trend was strongly relevant to changes in relative humidity and surface solar radiation downward (SSRD). Statistical analyses revealed that SSRD dominated MDA8-O3 variability spatially over almost the entire NEC region. Additionally, the contribution decomposition suggested that the trend of increase in annual average MDA8-O3 in NEC was dominated by anthropogenic emission change, which explained 91.9% of the total variation. Regionally, although the dominant role of emissions has not changed, the meteorology-driven anomalies also explained −4.3%–3.3% of the annual average MDA8-O3 variation. This study provides insight into decoupling the complex relationships between long-term variability in regional O3 pollution and both anthropogenic emissions and meteorology, which could provide valuable information for future efforts to address the O3 pollution in NEC.

Municipal solid waste incineration is gradually becoming the main method of waste disposal, but waste incineration produces many organic pollutants (e.g., dioxins). In order to better implement waste management, China identified 46 key cities to implement domestic waste classification first in 2017. This study predicts dioxin emissions in 2030 based on the background of waste classification policy, and analyzes the impact of waste classification on dioxin reduction. Firstly, k-means was used to classify the 46 cities of waste classification into four categories, and the representative cities in the four categories were selected to analyze the correlation between different influencing factors and municipal solid waste in each category through grey correlation analysis. And the municipal solid waste of each city in 2030 was predicted by bidirectional long and short term memory neural network. Finally, four scenarios are set up based on the background of waste classification policy to predict dioxin emissions in each city in 2030. It is found that at least 7.49%–13.07% dioxin emission reduction can be achieved by 2030 through waste classification. Waste classification has a positive impact on dioxin emission reduction.

In this study, the effects of aerosol-radiation interaction (ARI) and data assimilation (DA) during a haze episode in December 2016 are evaluated by four sets of parallel experiments using the atmospheric chemical model GRAPES_Meso5.1/CUACE. The results show that although the BASE experiment without ARI and DA generally captures the variations of particulate matter (PM) and visibility (VIS), it dramatically underestimates the high PM concentrations and overestimates the low VIS on severe haze days. Moreover, the ARI effect is not significant when the aerosol concentration is relatively low (PM2.5 <150 μgm−3). Assimilation of surface PM2.5 corrects the model chemical initial conditions (ICs) and increases the aerosol concentration in severe haze areas. Focusing on the severe pollution periods, the combined effect of ARI and DA significantly improves forecast accuracy and prolongs this improvement, which reduces the average negative biases of PM2.5 and PM10 by 33% and 35%, and increases the average correlation coefficients by 14% and 12%. Moreover, this combined effect also changes the vertical distribution of the atmosphere, especially at a low level, leading to a 2.6 K decrease in potential temperature (PT) and a 7.6% increase in relative humidity (RH) below 300 m. Due to the contributions of PM and RH, the low VIS prediction is significantly improved, with the root mean square error (RMSE) reduced by 30%. The study results show the combined effect of ARI and DA on aerosols and meteorology and suggest the importance of considering ARI and DA simultaneously in the atmospheric chemical model.

Accurate identification of key odorants has become a challenging and urgent task due to the increasing severity of odor nuisance. This study aims to rapidly and accurately identify 20 key odorants in different rubber product industries and treatment procedures using a combination of sensory evaluation, gas chromatography–olfactory–mass spectrometry (GC-O-MS), statistical analysis, and odor activity value (OAV) analysis. The identified odorants include alcohols, acids, aldehydes, ketones, sulfides, aromatic compounds, alkanes, alkenes, and other compounds. These compounds are categorized as sour, odorous, bitter, rubbery, and foul rubber-like, which are commonly associated with odor pollution in the rubber product industry. An odor wheel specific to rubber product enterprises is developed by combining chemical analysis with sensory evaluation of odorants. This odor wheel can be used for rapid traceability and identification of odor pollution at industrial sites. Results present a novel approach and theoretical framework for the traceability and control of odor pollution in the rubber industry.

Nanoparticles have recently been used as fuel additives to enhance diesel engine performance and exhaust gas emissions. Amino-functionalized multiwalled carbon nannotubes (NH2-MWCNTs) are catalysts that can be applied in diesel engines because of their promising properties. This study conducted an experimental investigation to determine the impacts of adding NH2-MWCNTs to diesel fuel using a single-cylinder CI engine running at a constant speed (1500 rpm) and various loads. The NH2-MWCNTs were added to diesel fuel at four concentrations of 25, 50, 75, and 100 ppm. Also, the same doses of pristine multiwalled carbon nanotubes (MWCNTs) were added to diesel to be used as a reference for NH2-MWCNTs. Most NH2-MWCNTs blends were preferred in combustion characteristics, engine performance, and exhaust emissions compared to diesel and MWCNTs blends. Thus, there was a considerable increase in cylinder pressure and heat release rate compared to pure diesel. Also, the brake thermal efficiency was increased, and the brake-specific fuel consumption was decreased compared to diesel. Most NH2-MWCNTs combinations appeared to decrease NOx, soot, and CO levels significantly. The NH2-MWCNTs blends are generally favored over the pristine MWCNTs blends in reducing exhaust emissions.

Southeast Asia (SEA) has been suggested as a hotspot of mercury (Hg) emission worldwide but research on atmospheric Hg remains scarce in this region. This study reported monitoring data of total gaseous mercury (TGM) at Ho Chi Minh City (HCMC), a tropical megacity in SEA. TGM was measured over a 6-month period (June–November) in 2022 during two distinct monsoon seasons in HCMC. The average TGM concentration at HCMC was 1.89 ± 0.55 ng m−3 being comparatively lower than other sites in East Asia. For seasonal variation, changes in air mass origins resulted in a nearly 50% more elevated TGM concentration during October–November (2.47 ± 0.56 ng m−3) as compared to June–September (1.71 ± 0.40 ng m−3). Meanwhile, planetary boundary height and anthropogenic activities were responsible for TGM diurnal variation. Both backward trajectory and PSCF analysis revealed East Asian outflow as an important factor contributing to higher TGM levels at the sampling site. This study identified two instances of tropical cyclones affecting HCMC. For the event in July, TGM concentrations decreased due to the increases in atmospheric disturbances and the contribution of marine air masses. On the other hand, for the event in October, the increase in TGM concentration was possibly due to the enhancement export of polluted air mass from continental East Asia, induced by the tropical cyclone. This study provided valuable atmospheric Hg data in SEA, propose further insight into regional Hg understanding and suggest the impact of monsoon influence on local and regional Hg pollution.

The Sichuan Basin (SCB), located immediately to the east of the Tibetan Plateau (TP) in southwest China, is identified as a region with severe PM2.5 pollution, especially in the western SCB region. To understand the terrain effect on the atmospheric environment change in detail, this study investigated the effect degree and meteorological mechanism of thermally driven mountain-valley breeze (MVB) circulations on wintertime PM2.5 in the western SCB region, based on the near-surface observations of PM2.5 and the ERA5 reanalysis data of meteorology. The results showed that the western SCB edge exhibited a significant diurnal change of MVB, shifting between daytime upslope easterly flows and nighttime downslope westerly flows. The frequency of the MVB circulations was accounted for 39% days in December 2017, with the mountain and valley breeze-controlling periods being from 01:00 to 05:00 and 14:00 to 17:00 local time, respectively. Notably, the hourly PM2.5 reductions of 13.9 ± 4.6 μg m−3 was averaged during the MVB days in the western SCB edge, resulting in a decrease of 46.4 ± 14.0% in PM2.5 pollution with the MVB, which indicates that the MVB could alleviate PM2.5 pollution in improving air quality over the western SCB region. The daytime MVB circulations drove the transport of PM2.5-rich air mass in the atmospheric boundary layer from the polluted western SCB edge to the surrounding regions, causing 22% attenuation in near-surface PM2.5 concentrations during the valley breeze-controlling period. The nocturnal MVB circulations carried clean TP air eastward downslope along the eastern slope of the TP into the polluted western SCB region, mitigating PM2.5 levels by 20% during the mountain breeze-controlling period.

Soil moisture and nitrogen fertilizer application rates are among the key factors affecting ammonia (NH3) emission, particularly in alkaline soil. However, the dynamic and thermodynamic properties of NH3 emissions from alkaline soils have not been extensively studied. In this study, we conducted an incubation experiment and theoretical analysis to investigate NH3 emission dynamics and thermodynamic properties under treatments involving different moisture levels (5%, 15%, 25%, and 35%), nitrogen fertilizer application rates (0, 172.4, and 344.8 mg kg−1), and temperatures (10, 20, 30, and 40 °C). The accumulative NH3 emissions varied significantly among the different moisture, temperature, and nitrogen treatment conditions and exhibited a logarithmic growth trend over time. The highest accumulative NH3 emissions occurred under the following conditions: 344.8 mg kg−1 nitrogen fertilizer, 5% moisture, and 40 °C temperature. The highest average emission rate constant (KN) throughout the entire incubation period was obtained at 40 °C. The KN of NH3 emissions in the high-temperature region (30–40 °C) was more sensitive to temperature. With increasing soil moisture content, the activated free energy (ΔG) initially increased and then decreased. As the nitrogen fertilizer application rate increased, the ΔG decreased, thereby leading to higher NH3 emissions. Combining the kinetic and thermodynamic parameters revealed that temperature, soil moisture, and nitrogen fertilizer application rate were key factors influencing NH3 emissions in alkaline soil, with the nitrogen fertilizer application rate having a greater impact than that of temperature, soil moisture, and their coupling. These findings provide valuable insights into the thermodynamic and kinetic control mechanisms of NH3 emissions, which can be regulated by fertilization and irrigation methods that consider the nitrogen cycle in alkaline soils.

In this study, we described the spatial asymmetry of the Asian Summer Monsoon Anticyclone (ASMA) circulation using Geopotential Height (GPH) values and divided ASMA into North-West (NW: 32.5°–37.5° N, 40°-70°E), North-East (NE: 32.5°–37.5° N, 70°-100°E), South-West (SW: 27.5°–32.5° N, 40°-70°E) and South-East (SE: 27.5°–32.5° N, 70°-100°E) regions. We provided the spatial asymmetry in the aerosol distribution using the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) (2006–2020) measurements. The vertical distribution of aerosols has been investigated over these regions dividing the atmosphere into (1) Boundary Layer-BL (0–1.5 km), (2) Mid-Troposphere-MT (∼1.5–8 km) and (3) Upper Troposphere and Lower Stratosphere -UTLS (∼8–18 km). Highest aerosol extinction values are noticed in the MT and their contribution to the total Aerosol Optical Depth (AOD) is highest (∼50–80%). The contribution of UTLS aerosols to the total AOD is ∼10% particularly in the eastern part (NE/SE) of the ASMA. The enhancement of aerosols throughout the upper troposphere over the eastern regions (SE/NE) of ASMA, indicates the possible transport of boundary layer pollutants to the UTLS. Using cloud fraction measurements from CloudSat and Pressure vertical velocity (ω), we identified that the intense convection over Bay of Bengal (BoB) is responsible for the UTLS aerosols over the SE region of ASMA. The descending limb of the monsoon-induced circulation over the western region of ASMA (Arabian Peninsula and other Middle Eastern countries) is the possible causative mechanism for the removal of aerosols in the upper troposphere. These findings on the spatial asymmetry in the aerosol distribution over the ASMA are expected to provide the importance of the regional radiative forcing and their climatic impacts.

Methane (CH4), a greenhouse gas with significant global warming potential, has complicate spatial and temporal distributions, especially for area with intense anthropogenic and natural sources. Here, we present the atmospheric CH4 record from three stations in Suzhou, China, which is located in the famous economic developed zone, with wetland coverage about 20%. The study found that although significant variations in CH4 concentrations across different regions within the Suzhou city, the mean values of the three sites could represent the atmospheric CH4 levels in Suzhou. The annual mean CH4 value in 2021 was significantly higher than that in 2020, with a growth of 8.02 ppb yr−1. CH4 followed a seasonal pattern, with low values in the spring and winter, and peak values in the autumn and summer. However, there were trough and surge values over time, which occurred in summer and winter in both year. These results highlighted that huge amounts of OH radicals accumulated when the summer extended a period of drought and heat, aggravating CH4 consumption and lowering concentration. Moreover, methanogens in subtropical areas might not be impacted by winter's lower temperatures. The CH4 significantly decreased exponentially with the increased wind speed, positively correlated with air temperature in spring and negatively correlated with atmospheric pressure. Taihu Lake, lied in the WNW, W, WSW and SW winds sectors with high concentration of CH4, is a local source. The Yellow Sea and the East China Sea where have some distance from Suzhou are important regional sources to the CH4.

Researchers have studied motorcyclists' exposure to Particulate Matter (PM) during rush and non-rush hours. However, the combined effect of season and hour of the day on PM concentration has not been studied. PM concentration was measured near the typical breathing zone of motorcyclists who traveled along four designated routes in a densely populated city (Varanasi, India). Data were collected from January to June 2022 during various hours of the day (from 07:00 to 17:00 h). PM2.5 and PM10 concentration during winter was 2.36 and 1.69 times in summer, respectively. In contrast, PM2.5 and PM10 concentration during spring were 1.35 and 1.12 times during summer, respectively. PM2.5 correlated much more with relative humidity and atmospheric temperature than PM10. Higher PM concentration was recorded in rush hour (09:00–10:00) during spring and summer but in non-rush hours (12:00–13:00) during winter. The opposite trend in the winter was caused by more extended dispersion and dilution time of PM particles. Also, a higher proportion of PM2.5 was observed on routes with longer rush hours caused due to road encroachment by street vendors and pedestrians.

We explored the route-based chemical significance, spatial distribution, temporal variation, and source origin of fine particles (PM2.5) in East Asia. Marine PM2.5 was sampled at three remote islands in the Taiwan Strait (TS) and the South China Sea (SCS) to identify their chemical significance. High PM2.5 concentrations concurred with LRT of northerly polluted air. Moreover, daytime PM2.5 levels were mostly greater than nighttime. Secondary inorganic aerosols (SIAs) accounted for 42.4–79.9% of water-soluble ions (WSIs) in PM2.5, which were commonly higher in spring and winter. Although crustal elements suppressed the metallic contents of PM2.5, trace elements (V, Cr, Mn, Ni, Zn, and Cd) came dominantly from anthropogenic origins. Additionally, OC/EC ratios >2.0 were observed due to higher organic carbon (OC) than elemental carbon (EC) during the Asian Northeastern Monsoons (ANMs) commonly in spring and winter. Levoglucosan and oxalic acid were, respectively, the most abundant anhydrosugars (ASs) and organic acids (OAs) in PM2.5, and both descended from the north to the south. The mass ratios of malonic and succinic acids (M/S) were in the range of 0.95–1.41, suggesting a high correlation of marine PM2.5 with secondary organic aerosols (SOAs). Source resolution of PM2.5 showed that sea salts, fugitive dust, mobile sources, secondary sulfate and nitrate were the major sources of PM2.5. In particular, anthropogenic origins and biomass burning (BB) contributed notably to PM2.5 during the ANMs. Overall, compared to natural sources (21–48%), anthropogenic sources (25–57%) contributed more to marine PM2.5 in East Asia.

Particle exposure affects more humans globally than any other air pollutant. However, due to expensive instruments and infrastructural deficiency, a high spatiotemporal network of monitoring stations is not possible, leading to data-scarce regions. Satellite and reanalysis datasets can be implemented to estimate particulate matter, but they do not provide surface concentration and needs to be reconstructed from the components. In this study, a machine learning (ML) framework is implemented to reconstruct PM2.5 from MERRA-2 data components, namely black carbon (BC), organic carbon (OC), dust (DUST), sea salt (SS), and sulfate (SO4) mass concentration. The ground-level data were collected from India's 335 continuous ambient air quality monitoring stations (CAAQMS) and respective MERRA-2 data for 2017–2021 at hourly resolution. Random forest (RF) performs better with train and test scores (R2) of 0.84 and 0.73, respectively, while the empirical equation provides an R2 of only 0.26 on test data. The estimated PM2.5 for Indian states from 1980 to 2021 indicates a significant increase in most cases. However, states in the Indo-Gangetic plain such as Delhi, Punjab, Haryana, and Uttar Pradesh, are the most polluted regions of India. The major shift in concentration is from 2000 onwards, which can be seen as a direct result of the economic liberalization policies implemented in 1991. The results provide evidence for the limitations of the broad application of the empirical equation and the feasibility of ML algorithms as a potential reconstruction technique for developing robust and accurate region-specific models from MERRA-2 data.

Local emission sources and long-range transport of air pollutants are the main causes of PM2.5 pollution in developing metropolises. A modeling system was used to determine the main contributors of PM2.5 and future changes in PM2.5 concentrations in Hanoi, one of the fast-growing and most polluted capital cities in the world. Additionally, BenMAP was used to estimate the number of premature deaths caused or avoided due to PM2.5 concentration changes. October and December 2017 as well as April and June 2018 were selected for analysis. The source apportionment approach revealed that the boundary conditions (BCON), were the primary source of PM2.5, accounting for 24–59.7%. Agriculture was a substantial contributor of PM2.5 within Hanoi in June and October, accounting for 27.8 and 18.8%, respectively. Projected emissions for 2030 from our previous work, which were estimated based on the action plans and scenarios proposed by the government, were used as input into the air quality modeling system to analyze changes in the concentrations of the total PM2.5 and its major components owing to emissions changes. It is shown that, the concentrations of PM2.5 can increase by 0.5 and 1.6 μg/m3 in April and December, and decrease by 3.2 and 1.5 μg/m3 in June and October, respectively. Using BenMAP , it is estimated that these changes can cause 45 and 95 premature deaths in April and December, but prevent 198 and 88 premature deaths in June and October respectively, in Hanoi.

In this study, a petrochemical industrial district (PID) in the megacity of Shanghai was selected as a case study to investigate the quantity and components, sectoral contributions, and spatial distribution of greenhouse gas (GHG) emissions, including carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O). The total GHG emissions from the PID were 25.86 million tonnes of carbon dioxide equivalent (CO2eq), of which CO2 made the largest contribution (96.9%), followed by N2O (2%) and CH4 (1.1%). In terms of the sectoral contributions, energy use and industrial processes and product use (IPPU) were the two major emission sources of GHGs, accounting for 76.14% and 23.17% of the total, respectively. Moreover, GHG emissions from energy use and IPPU in the petrochemical industry accounted for 36.1% of the total, indicating the significance of the petrochemical industry for GHG emissions. A high-resolution gridded (200 × 200 m) spatial distribution map of GHG emissions was created based on point source, line source, and area source databases, showing a clear difference in the emission levels of GHGs among the grids. Thirteen ultra-high grids with GHG emissions above 10,000 tonnes CO2eq, accounted for more than 90% of the total emissions in the PID, which should be given more attention in emission management programs. The results provided useful insights into the amount and composition, sectoral contributions, and gridded spatial distribution characteristics of GHG emissions in the petrochemical industry, and can be used by decision makers to develop refined management strategies for GHG emission reductions in typical petrochemical districts.

Recently, researchers have prioritized the accurate forecasting of the particulate matter (PM) air quality indicators PM2.5 and PM10 in urban and industrial locations due to their importance for the human health. However, accurate short-term forecasting via traditional data mining methods is limited due to the complex nature of these indices on hourly scale. To address this issue, a novel deep learning framework composed of classification and regression tree (CART) feature selection, time-varying filter-based empirical mode decomposition (TVF-EMD), and ensemble deep random vector functional link (DRVFL) scheme to forecast 1-h and 3-h ahead PM2.5 and PM10 in two different zones of Australia. In the first pre-processing phase, CART feature selection optimizes the antecedent information (lag) for each target using the computed important factor. Then, the original signal is decomposed into intrinsic mode functions (IMFs) and residual sub-components by the TVF-EMD to overcome the non-stationarity and complexity. The DRVFL approach is then executed by applying the significant lagged-time components as the optimal input feature via each decomposed sub-component. Finally, all the forecast values obtained using the sub-components are summed up to construct the PM values for each horizon. In addition, the classical RVFL and bidirectional gated recurrent unit neural network (Bi-GRU) models are adopted in the form of hybrid and corresponding standalone frameworks to validate the TVD-EMD-DRVFL. The results demonstrate that TVF-EMD-DRVFL model provides best preciseness with R, RMSE, MAPE, NSE, KGE, IA, U95%, and KGE at t+1 and t+3 to forecast PM2.5 and PM10 for Brisbane and North Parramatta station.

Accurate long-term (6–24 h) prediction of PM2.5 is critical to human health and daily life. While deep learning techniques have been extensively used to forecast PM2.5, prior studies have primarily relied on shallow recurrent neural networks (RNNs), which may accumulate errors and limit the long-term prediction capability of the model. To address this issue, a new hybrid model has been proposed in this study, which combines the Complete Ensemble Empirical Mode Decomposition Adaptive Noise (CEEMDAN) method with a deep Transformer neural network (DeepTransformer) to enhance the accuracy of long-term PM2.5 forecasting. The model includes a new embedding layer that efficiently models historical, meteorological, and discrete-time data. Additionally, to improve the long-term inference capability of DeepTransformer, a non-autoregressive direct multi-step (DMS) prediction strategy is introduced, and a novel DMS decoder replaces the vanilla Transformer decoder. Experiments conducted on two public datasets demonstrate that the novel model achieves excellent prediction performance. Specifically, DeepTransformer achieves R2=0.984 and RMSE=11.61 µg/m3 in 1-hour prediction and R2=0.704 and RMSE=30.78 µg/m3 in 24-hour prediction. Compared to single models, DeepTransformer achieves a 30% decrease in MAE, a 27% decrease in RMSE, and a 59% increase in R2 for the long-term (24-hour) prediction of PM2.5