Improving PM2.5 Air Quality Model Forecasts in China Using a Bias-Correction Framework

Substantial differences in source contributions to carbon emissions and health damage necessitate balanced synergistic control plans in China

China's strategy to concurrently address climate change and air pollution mitigation is hindered by a lack of comprehensive information on source contributions to health damage and carbon emissions. Here we show notable discrepancies between source contributions to CO2 emissions and fine particulate matter (PM2.5)-related mortality by using adjoint emission sensitivity modeling to attribute premature mortality in 2017 to 53 sector and fuel/process combinations with high spatial resolution. Our findings reveal that monetized PM2.5 health damage exceeds climate impacts in over half of the analyzed subsectors. In addition to coal-fired energy generators and industrial boilers, the combined health and climate costs from energy-intensive processes, diesel-powered vehicles, domestic coal combustion, and agricultural activities exceed 100 billion US dollars, with health-related costs predominating. This research highlights the critical need to integrate the social costs of health damage with climate impacts to develop more balanced mitigation strategies toward these dual goals, particularly during fuel transition and industrial structure upgrading.

An Empirical Mode Decomposition Fuzzy Forecast Model for Air Quality

Air quality has a significant influence on people's health. Severe air pollution can cause respiratory diseases, while good air quality is beneficial to physical and mental health. Therefore, the prediction of air quality is very important. Since the concentration data of air pollutants are time series, their time characteristics should be considered in their prediction. However, the traditional neural network for time series prediction is limited by its own structure, which makes it very easy for it to fall into a local optimum during the training process. The empirical mode decomposition fuzzy forecast model for air quality, which is based on the extreme learning machine, is proposed in this paper. Empirical mode decomposition can analyze the changing trend of air quality well and obtain the changing trend of air quality under different time scales. According to the changing trend under different time scales, the extreme learning machine is used for fast training, and the corresponding prediction value is obtained. The adaptive fuzzy inference system is used for fitting to obtain the final air quality prediction result. The experimental results show that our model improves the accuracy of both short-term and long-term prediction by about 30% compared to other models, which indicates the remarkable efficacy of our approach. The research of this paper can provide the government with accurate future air quality information, which can take corresponding control measures in a targeted manner.

Novel Method for Ozone Isopleth Construction and Diagnosis for the Ozone Control Strategy of Chinese Cities

Ozone (O₃) isopleths describe the nonlinear responses of O₃ concentrations to changes in nitrogen oxides (NO_X) and volatile organic compounds (VOCs) and thus are pivotal to the determination of O₃ control requirements. In this study, we innovatively use the Community Multiscale Air Quality model with the high-order decoupled direct method (CMAQ-HDDM) to simulate O₃ pollution of China in 2017 and derive O₃ isopleths for individual cities. Our simulation covering the entire China Mainland suggests severe O₃ pollution as 97% of the residents experienced at least 1 day, in 2017, in excess of Chinese Level-II Ambient Air Quality Standards for O₃ as 160 μg·m^-3 (81.5 ppbV equally). The O₃ responses to emissions of precursors vary widely across individual cities. Densely populated metropolitan areas such as Jing-Jin-Ji, Yangtze River Delta, and Pearl River Delta are following NO_X-saturated regimes, where a small amount of NO_X reduction increases O₃. Ambient O₃ pollution in the eastern region generally is limited by VOCs, while in the west by NO_X. The city-specific O₃ isopleths generated in this study are instrumental in forming hybrid and differentiated strategies for O₃ abatement in China.

Evaluation of the offline-coupled GFSv15-FV3-CMAQv5.0.2 in support of the next-generation National Air Quality Forecast Capability over the contiguous United States

As a candidate for the next-generation National Air Quality Forecast Capability (NAQFC), the meteorological forecast from the Global Forecast System with the new Finite Volume Cube-Sphere dynamical core (GFS-FV3) will be applied to drive the chemical evolution of gases and particles described by the Community Multiscale Air Quality modeling system. CMAQv5.0.2, a historical version of CMAQ, has been coupled with the North American Mesoscale Forecast System (NAM) model in the current operational NAQFC. An experimental version of the NAQFC based on the offline-coupled GFS-FV3 version 15 with CMAQv5.0.2 modeling system (GFSv15-CMAQv5.0.2) has been developed by the National Oceanic and Atmospheric Administration (NOAA) to provide real-time air quality forecasts over the contiguous United States (CONUS) since 2018. In this work, comprehensive region-specific, time-specific, and categorical evaluations are conducted for meteorological and chemical forecasts from the offline-coupled GFSv15-CMAQv5.0.2 for the year 2019. The forecast system shows good overall performance in forecasting meteorological variables with the annual mean biases of -0.2 °C for temperature at 2 m, 0.4% for relative humidity at 2 m, and 0.4 m s-1 for wind speed at 10 m compared to the METeorological Aerodrome Reports (METAR) dataset. Larger biases occur in seasonal and monthly mean forecasts, particularly in spring. Although the monthly accumulated precipitation forecasts show generally consistent spatial distributions with those from the remote-sensing and ensemble datasets, moderate-to-large biases exist in hourly precipitation forecasts compared to the Clean Air Status and Trends Network (CASTNET) and METAR. While the forecast system performs well in forecasting ozone (O3) throughout the year and fine particles with a diameter of 2.5 μm or less (PM2.5) for warm months (May-September), it significantly overpredicts annual mean concentrations of PM2.5. This is due mainly to the high predicted concentrations of fine fugitive and coarse-mode particle components. Underpredictions in the southeastern US and California during summer are attributed to missing sources and mechanisms of secondary organic aerosol formation from biogenic volatile organic compounds (VOCs) and semivolatile or intermediate-volatility organic compounds. This work demonstrates the ability of FV3-based GFS in driving the air quality forecasting. It identifies possible underlying causes for systematic region- and time-specific model biases, which will provide a scientific basis for further development of the next-generation NAQFC.

Significant contrasts in aerosol acidity between China and the United States

Aerosol acidity governs several key processes in aerosol physics and chemistry, thus affecting aerosol mass and composition and ultimately climate and human health. Previous studies have reported aerosol pH values separately in China and the United States (USA), implying different aerosol acidity between these two countries. However, there is debate about whether mass concentration or chemical composition is the more important driver of differences in aerosol acidity. A full picture of the pH difference and the underlying mechanisms responsible is hindered by the scarcity of simultaneous measurements of particle composition and gaseous species, especially in China. Here we conduct a comprehensive assessment of aerosol acidity in China and the USA using extended ground-level measurements and regional chemical transport model simulations. We show that aerosols in China are significantly less acidic than in the USA, with pH values 1-2 units higher. Based on a proposed multivariable Taylor series method and a series of sensitivity tests, we identify major factors leading to the pH difference. Compared to the USA, China has much higher aerosol mass concentrations (gas + particle, by a factor of 8.4 on average) and a higher fraction of total ammonia (gas + particle) in the aerosol composition. Our assessment shows that the differences in mass concentrations and chemical composition play equally important roles in driving the aerosol pH difference between China and the USA - increasing the aerosol mass concentrations (by a factor of 8.4) but keeping the relative component contributions the same in the USA as the level in China increases the aerosol pH by ~1.0 units and further shifting the chemical composition from US conditions to China's that are richer in ammonia increases the aerosol pH by ~0.9 units. Therefore, China being both more polluted than the USA and richer in ammonia explains the aerosol pH difference. The difference in aerosol acidity highlighted in the present study implies potential differences in formation mechanisms, physicochemical properties, and toxicity of aerosol particles in these two countries.

Operational bias correction for PM2.5 using the AIRPACT air quality forecast system in the Pacific Northwest

A bias correction scheme based on a Kalman filter (KF) method has been developed and implemented for the AIRPACT air quality forecast system which operates daily for the Pacific Northwest. The KF method was used to correct hourly rolling 24-h average PM_2.5 concentrations forecast at each monitoring site within the AIRPACT domain and the corrected forecasts were evaluated using observed daily PM_2.5 24-h average concentrations from 2017 to 2018. The evaluation showed that the KF method reduced mean daily bias from approximately -50% to ±6% on a monthly averaged basis, and the corrected results also exhibited much smaller mean absolute errors typically less than 20%. These improvements were also apparent for the top 10 worst PM_2.5 days during the 2017-2018 test period, including months with intensive wildfire events. Significant differences in AIRPACT performance among urban, suburban, and rural monitoring sites were greatly reduced in the KF bias correction forecasts. The daily 24-h average bias corrections for each monitoring site were interpolated to model grid points using three different interpolation schemes: cubic spline, Gaussian Kriging, and linear Kriging. The interpolated results were more accurate than the original AIRPACT forecasts, and both Kriging methods were better than the cubic spline method. The Gaussian method yielded smaller mean biases and the linear method yielded smaller absolute errors. The KF bias correction method has been implemented operationally using both Kriging interpolation methods for routine output on the AIRPACT website (http://lar.wsu.edu/airpact). This method is relatively easy to implement, but very effective to improve air quality forecast performance.Implications: Current chemical transport models, including CMAQ, used for air quality forecasting can have large errors and uncertainties in simulated PM_2.5 concentrations. In this paper, we describe a relatively simple bias correction scheme applied to the AIRPACT air quality forecast system for the Pacific Northwest. The bias correction yields much more accurate and reliable PM_2.5 results compared to the normal forecast system. As such, the operational bias corrected forecasts will provide a much better basis for daily air quality management by agencies within the region. The bias corrected results also highlight issues to guide further improvements to the normal forecast system.

Gridded bias correction of modeled PM2.5 for exposure assessment, and estimation of background concentrations over a coastal valley region of northwestern British Columbia, Canada

Chemical transport models (CTM) can have large biases and errors when simulating pollutant concentrations. To improve the characterization of fine particulate matter (PM_2.5) over complex terrain for exposure assessments, three mathematical formulae that utilized the relationship between modeled and observed quantile concentrations at a monitor location were developed. These were then applied to 1 year of CMAQ model output of PM_2.5 over the Terrace-Kitimat Valley of northwestern British Columbia, Canada. The final products enhanced the representation of ambient levels at existing monitoring stations when evaluated with conventional statistical measures. Better agreement of corrected outputs with observed compliance metrics was also found. On average, the absolute errors of amended outputs were 11% and 10% for the annual mean PM_2.5 and 98th percentiles of daily concentrations, respectively, compared to 45% and 61%, respectively, in the original outputs. These improvements provided greater confidence to use the amended outputs to estimate concentrations at locations without monitors. The predominance of pristine conditions in the modeling domain was exploited to derive annual background PM_2.5 concentrations over the valley, which was estimated to be 2.0-2.3 μg m^-3. To our knowledge, this is the first study to calculate background PM_2.5 concentrations over northern BC coastlands using bias-corrected outputs from an air quality model. Implications: Bias correction of CMAQ model output was necessary for assessing regulatory compliance for ambient PM_2.5. The implications are notable. First, for low to moderate spatial heterogeneity in monitoring data, the use of regression equations that relates quantile mean concentrations of model outputs to those of observational data enhances the estimation of PM_2.5 at unmonitored locations. Second, by providing spatial pollutant distribution ahead of planned industrial development in Terrace-Kitimat Valley (TKV), corrected model output offers a baseline for tracking progress in airshed management. Third, correction improved pollutant exposure classification, for which the risk was predominantly low. Finally, 2.0-2.3 μg m^-3 should be considered as PM_2.5 concentrations that are irreducible when setting voluntary targets for ambient levels in the area.

Increased air pollution exposure among the Chinese population during the national quarantine in 2020

The COVID-19 quarantine in China is thought to have reduced ambient air pollution. The overall exposure of the population also depends, however, on indoor air quality and human mobility and activities. Here, by integrating real-time mobility data and a questionnaire survey on time-activity patterns during the pandemic, we show that despite a decrease in ambient PM_2.5 during the quarantine, the total population-weighted exposure to PM_2.5 considering both indoor and outdoor environments increased by 5.7 μg m^-3 (95% confidence interval, 1.2-11.0 μg m^-3). The increase in population-weighted exposure was mainly driven by a nationwide urban-to-rural population migration before the Spring Festival coupled with the freezing of the migration backward due to the quarantine, which increased household energy consumption and the fraction of people exposed to rural household air pollution indoors. Our analysis reveals an increased inequality of air pollution exposure during the quarantine and highlights the importance of household air pollution for population health in China.

Increased air pollution exposure among the Chinese population during the national quarantine in 2020

The COVID-19 quarantine in China is thought to have reduced ambient air pollution. The overall exposure of the population also depends, however, on indoor air quality and human mobility and activities. Here, by integrating real-time mobility data and a questionnaire survey on time-activity patterns during the pandemic, we show that despite a decrease in ambient PM_2.5 during the quarantine, the total population-weighted exposure to PM_2.5 considering both indoor and outdoor environments increased by 5.7 μg m^-3 (95% confidence interval, 1.2-11.0 μg m^-3). The increase in population-weighted exposure was mainly driven by a nationwide urban-to-rural population migration before the Spring Festival coupled with the freezing of the migration backward due to the quarantine, which increased household energy consumption and the fraction of people exposed to rural household air pollution indoors. Our analysis reveals an increased inequality of air pollution exposure during the quarantine and highlights the importance of household air pollution for population health in China.

New interpretable deep learning model to monitor real-time PM2.5 concentrations from satellite data

Particulate matter with a mass concentration of particles with a diameter less than 2.5 μm (PM_2.5) is a key air quality parameter. A real-time knowledge of PM_2.5 is highly valuable for lowering the risk of detrimental impacts on human health. To achieve this goal, we developed a new deep learning model-EntityDenseNet to retrieve ground-level PM_2.5 concentrations from Himawari-8, a geostationary satellite providing high temporal resolution data. In contrast to the traditional machine learning model, the new model has the capability to automatically extract PM_2.5 spatio-temporal characteristics. Validation across mainland China demonstrates that hourly, daily and monthly PM_2.5 retrievals contain the root-mean-square errors of 26.85, 25.3, and 15.34 μg/m³, respectively. In addition to a higher accuracy achievement when compared with various machine learning inversion methods (backpropagation neural network, extreme gradient boosting, light gradient boosting machine, and random forest), EntityDenseNet can "peek inside the black box" to extract the spatio-temporal features of PM_2.5. This model can show, for example, that PM_2.5 levels in the coastal city of Tianjin were more influenced by air from Hebei than Beijing. Further, EntityDenseNet can still extract the seasonal characteristics that demonstrate that PM_2.5 is more closely related within three month groups over mainland China: (1) December, January and February, (2) March, April and May, (3) July, August and September, even without meteorological information. EntityDenseNet has the ability to obtain high temporal resolution satellite-based PM_2.5 data over China in real-time. This could act as an important tool to improve our understanding of PM_2.5 spatio-temporal features.

Explore a Multivariate Bayesian Uncertainty Processor driven by artificial neural networks for probabilistic PM2.5 forecasting

Quantifying predictive uncertainty inherent in the nonlinear multivariate dependence structure of multi-step-ahead PM_2.5 forecasts is challenging. This study integrates a Multivariate Bayesian Uncertainty Processor (MBUP) and an artificial neural network (ANN) to make accurate probabilistic PM_2.5 forecasts. The contributions of the proposed approach are two-fold. First, the MBUP can capture the nonlinear multivariate dependence structure between observed and forecasted data. Second, the MBUP can alleviate predictive uncertainty encountered in PM_2.5 forecast models that are configured by ANNs. The reliability of the proposed approach was assessed by a case study on air quality in Taipei City of Taiwan. We consider forecasts of PM_2.5 concentrations as a function of meteorological and air quality factors based on long-term (2010-2018) hourly observational datasets. Firstly, the Back Propagation Neural Network (BPNN) and the Adaptive Neural Fuzzy Inference System (ANFIS) were investigated to produce deterministic forecasts. Results revealed that the ANFIS model could learn different air pollutant emission mechanisms (i.e. primary, secondary and natural processes) from the clustering-based fuzzy inference system and produce more accurate deterministic forecasts than the BPNN. The ANFIS model then provided inputs (i.e. point estimates) to probabilistic forecast models. Next, two post-processing techniques (MBUP and the Univariate Bayesian Uncertainty Processor (UBUP)) were separately employed to produce probabilistic forecasts. The Bayesian Uncertainty Processors (BUPs) can model the dependence structure (i.e. posterior density function) between observed and forecasted data using a prior density function and a likelihood density function. Here in BUPs, the Monte Carlo simulation was introduced to create a probabilistic predictive interval of PM_2.5 concentrations. The results demonstrated that the MBUP not only outperformed the UBUP but also suitably characterized the complex nonlinear multivariate dependence structure between observations and forecasts. Consequently, the proposed approach could reduce predictive uncertainty while significantly improving model reliability and PM_2.5 forecast accuracy for future horizons.

A new approach combining a simplified FLEXPART model and a Bayesian-RAT method for forecasting PM10 and PM2.5

In this study, we evaluated atmospheric particulate matter (PM) concentration predictions at a regional scale using a simplified Lagrangian particle dispersion modeling system and the Bayesian and multiplicative ratio correction optimization (Bayesian-RAT) method to improve the mixing ratio forecast of PM₁₀ and PM_2.5. We first examined the forecast performance of the LPD (i.e., the simplified FLEXPART model combined with the Bayesian-RAT method) by comparing the model predictions with the PM concentration observations from 95 observation stations in Xingtai city and its surrounding areas. The first 2 months (i.e., Oct. and Nov. 2017) of the study period represented the typical spin-up time period, and the analysis period was December 2017. The LPD forecast system was much better (correlation coefficient: R=0.64 vs. 0.48 and 0.67 vs. 0.50 for PM₁₀ and PM_2.5, respectively; root mean square error: RMSE = 74.98 vs. 105.96 μg/m³ for PM₁₀ and 54.89 vs. 72.81 μg/m³ for PM_2.5) than the pre-calibration results. We also compared the LPD forecasting model with other models (WRF-Chem and Camx) using data from monitoring stations in Xingtai, China, and the LPD forecasting model had higher accuracy than the other models. In particular, the RMSE scores for hourly PM₁₀ concentrations were reduced by 36.51% and 42.21% compared to WRF-Chem and to Camx, respectively. The PM_2.5 forecast results, as in the case of PM₁₀, showed a better performance when applying the LPD model to the data from the monitoring stations. The RMSE was reduced by 26.44% and 18.47% relative to the WRF-Chem and Camx, respectively. The results confirm that there is much advantage of the LPD forecast system for predicting PM and may be for other pollutants.

Fusion Method Combining Ground-Level Observations with Chemical Transport Model Predictions Using an Ensemble Deep Learning Framework: Application in China to Estimate Spatiotemporally-Resolved PM2.5 Exposure Fields in 2014-2017

Atmospheric chemical transport models (CTMs) have been widely used to simulate spatiotemporally resolved PM_2.5 concentrations. However, CTM results are usually prone to bias and errors. In this study, we improved the accuracy of PM_2.5 predictions by developing an ensemble deep learning framework to fuse model simulations with ground-level observations. The framework encompasses four machine-learning models, i.e., general linear model, fully connected neural network, random forest, and gradient boosting machine, and combines them by stacking approach. This framework is applied to PM_2.5 concentrations simulated by the Community Multiscale Air Quality (CMAQ) model for China from 2014 to 2017, which has complete spatial coverage over the entirety of China at a 12-km resolution, with no sampling biases. The fused PM_2.5 concentration fields were evaluated by comparing with an independent network of observations. The R² values increased from 0.39 to 0.64, and the RMSE values decreased from 33.7 μg/m³ to 24.8 μg/m³. According to the fused data, the percentage of Chinese population residing under the level II National Ambient Air Quality Standards of 35 μg/m³ for PM_2.5 has increased from 46.5% in 2014 to 61.7% in 2017. The method is readily adapted to utilize near-real-time observations for operational analyses and forecasting of pollutant concentrations and can be extended to provide source apportionment forecasts as well.

A Study of Objective Prediction for Summer Precipitation Patterns Over Eastern China Based on a Multinomial Logistic Regression Model

The prediction of summer precipitation patterns (PPs) over eastern China is an important and topical issue in China. Predictors that are selected based on historical information may not be suitable for the future due to non-stationary relationships between summer precipitations and corresponding predictors, and might induce the instability of prediction models, especially in cases with few predictors. This study aims to investigate how to learn as much information as possible from various and numerous predictors reflecting different climate conditions. An objective prediction method based on the multinomial logistic regression (MLR) model is proposed to facilitate the study. The predictors are objectively selected from a machine learning perspective. The effectiveness of the objective prediction model is assessed by considering the influence of collinearity and number of predictors. The prediction accuracy is found to be comparable to traditionally estimated predictability, ranging between 0.6 and 0.7. The objective prediction model is capable of learning the intrinsic structure of the predictors, and is significantly superior to the prediction model with randomly-selected predictors and the single best predictor. A robust prediction can be generally obtained by learning information from plenty of predictors, although the most effective model may be constructed with fewer predictors through proper methods of predictor selection. In addition, the effectiveness of objective prediction is found to generally improve as observation increases, highlighting its potential for improvement during application as time passes.

A novel model for hourly PM2.5 concentration prediction based on CART and EELM

Hourly PM_2.5 concentrations have multiple change patterns. For hourly PM_2.5 concentration prediction, it is beneficial to split the whole dataset into several subsets with similar properties and to train a local prediction model for each subset. However, the methods based on local models need to solve the global-local duality. In this study, a novel prediction model based on classification and regression tree (CART) and ensemble extreme learning machine (EELM) methods is developed to split the dataset into subsets in a hierarchical fashion and build a prediction model for each leaf. Firstly, CART is used to split the dataset by constructing a shallow hierarchical regression tree. Then at each node of the tree, EELM models are built using the training samples of the node, and hidden neuron numbers are selected to minimize validation errors respectively on the leaves of a sub-tree that takes the node as the root. Finally, for each leaf of the tree, a global and several local EELMs on the path from the root to the leaf are compared, and the one with the smallest validation error on the leaf is chosen. The meteorological data of Yancheng urban area and the air pollutant concentration data from City Monitoring Centre are used to evaluate the method developed. The experimental results demonstrate that the method developed addresses the global-local duality, having better performance than global models including random forest (RF), v-support vector regression (v-SVR) and EELM, and other local models based on season and k-means clustering. The new model has improved the capability of treating multiple change patterns.

WRF-Chem modeling of particulate matter in the Yangtze River Delta region: Source apportionment and its sensitivity to emission changes

China has been troubled by high concentrations of fine particulate matter (PM_2.5) for many years. Up to now, the pollutant sources are not yet fully understood and the control approach still remains highly uncertain. In this study, four month-long (January, April, July and October in 2015) WRF-Chem simulations with different sensitivity experiments were conducted in the Yangtze River Delta (YRD) region of eastern China. The simulated results were compared with abundant meteorological and air quality observations at 138 stations in 26 YRD cities. Our model well captured magnitudes and variations of the observed PM_2.5, with the normal mean biases (NMB) less than ±20% for 19 out of the 26 YRD cities. A series of sensitivity simulations were conducted to quantify the contributions from individual source sectors and from different regions to the PM_2.5 in the YRD region. The calculated results show that YRD local source contributed 64% of the regional PM_2.5 concentration, while outside transport contributed the rest 36%. Among the local sources, industry activity was the most significant sector in spring (25%), summer (36%) and fall (33%), while residential source was more important in winter (38%). We further conducted scenario simulations to explore the potential impacts of varying degrees of emission controls on PM_2.5 reduction. The result demonstrated that regional cooperative control could effectively reduce the PM_2.5 level. The proportionate emission controls of 10%, 20%, 30%, 40% and 50% could reduce the regional mean PM_2.5 concentrations by 10%, 19%, 28%, 37% and 46%, respectively, and for places with higher ambient concentrations, the mitigation efficiency was more significant. Our study on source apportionment and emission controls can provide useful information on further mitigation actions.

Improving PM2.5 Forecasting and Emission Estimation Based on the Bayesian Optimization Method and the Coupled FLEXPART-WRF Model

In this study, we evaluated estimates and predictions of the PM2.5 (fine particulate matter) concentrations and emissions in Xuzhou, China, using a coupled Lagrangian particle dispersion modeling system (FLEXPART-WRF). A Bayesian inversion method was used in FLEXPART-WRF to improve the emission calculation and mixing ratio estimation for PM2.5. We first examined the inversion modeling performance by comparing the model predictions with PM2.5 concentration observations from four stations in Xuzhou. The linear correlation analysis between the predicted PM2.5 concentrations and the observations shows that our inversion forecast system is much better than the system before calibration (with correlation coefficients of R = 0.639 vs. 0.459, respectively, and root mean square errors of RMSE = 7.407 vs. 9.805 µg/m3, respectively). We also estimated the monthly average emission flux in Xuzhou to be 4188.26 Mg/month, which is much higher (by ~10.12%) than the emission flux predicted by the multiscale emission inventory data (MEIC) (3803.5 Mg/month). In addition, the monthly average emission flux shows obvious seasonal variation, with the lowest PM2.5 flux in summer and the highest flux in winter. This pattern is mainly due to the additional heating fuels used in the cold season, resulting in many fine particulates in the atmosphere. Although the inversion and forecast results were improved to some extent, the inversion system can be improved further, e.g., by increasing the number of observation values and improving the accuracy of the a priori emission values. Further research and analysis are recommended to help improve the forecast precision of real-time PM2.5 concentrations and the corresponding monthly emission fluxes.