Search

Accurate ozone (O 3) predictions is crucial for assessing its impact on public health and developing effective prevention and control measures. While ground-based observations are considered highly accurate, the limited and uneven spatial coverage poses challenges. Air quality models can provide complete spatiotemporal coverage but are subject to biases due to simplifications in physicochemical mechanisms and uncertainties associated with emission and meteorology inputs. In this study, we aimed to improve the accuracy of the predicted O 3 concentrations from the Community Multiscale Air Quality (CMAQ) model by using the machine learning techniques. First, we compared three machine learning algorithms, namely Light Gradient Boosting Machine (LightGBM), Random Forest, and eXtreme Gradient Boosting. Results showed that LightGBM exhibited the highest correlation coefficient (R) of 0.84, making it the preferred algorithm for further analysis. Subsequently, two multi-source data prediction models based on LightGBM were constructed to improve the accuracy of predicted daily maximum 8-h O 3 (O 3 -Max8h). The first model, referred to as LGBR, utilized the predicted air pollutant concentrations and meteorological fields as input variables, while the second model, named LGBR-CHAP, incorporated ChinaHighAirPollutants (CHAP) O 3 -Max8h as an additional input variable. Validation results demonstrated significant improvements in the LGBR and LGBR-CHAP models compared to the original CMAQ model. On the daily scale, the root mean square error and mean bias of the predicted O 3 -Max8h decreased by 3.15 and 2.07 μg/m3, respectively, for the LGBR model, while they decreased by 5.61 and 4.18 μg/m3, respectively, for the LGBR-CHAP model. On the monthly scale, the R of the original CMAQ model varied from 0.2 in the March to 0.91 in June, and LGBR improved all months (0.4–0.92) as did LGBR-CHAP (0.5–0.94). Spatially, the O 3 -Max8h simulated by the original CMAQ model is better in eastern China but less skillful in western China. After optimization by the LGBR and LGBR-CHAP models, the national average R improved from 0.77 to 0.83 and 0.88, respectively. The LGBR-CHAP model exhibited better predictive capacity than the LGBR model and was subsequently employed to generate high-resolution (10 km × 10 km) and full-coverage (100%) O 3 -Max8h data. This dataset will prove valuable for future air pollution and health studies. [Display omitted] • LightGBM shows the best capacity in improving the O 3 prediction accuracy from CMAQ. • The LGBR-CHAP model has better predictive performance than the LGBR model. • The spatiotemporal pattern of O 3 -Max8h is captured (R = 0.88, RMSE = 24.4 μg/m3). • Our model generates a high-resolution (10 km × 10 km) and full-coverage (100%) O 3 data. [ABSTRACT FROM AUTHOR]

In recent decades, China has been a hotspot for reactive nitrogen (N) deposition due to intensive fossil fuel burning and increased agricultural activities. The Chinese government has implemented active measures to protect air quality and reduce N deposition. In this study, we combined monthly measurements of monitoring sites and satellite data to construct random forest (RF) models for the estimation of monthly N deposition in China during 2008–2020. This provides N deposition estimation in latest years and has a finer temporal resolution. The RF models can address more complicated relationships between variables and performed well in estimating N deposition compared to the measurements with average correlation coefficients of 0.79 and 0.77 for dry and wet N deposition, respectively. We found that mean annual total N deposition was 22.0 ± 0.8 kg N ha−1 yr−1 during 2008–2020, which accounted for 64% (±4%) of reactive N emissions in China. Although the total N deposition was high, it has stabilized and decreased slightly over the past decade, with dry and wet N deposition contributing equally to the total N deposition. NH x deposition (F NHx) still slowly increased (0.09 kg N ha−1 yr−2) during 2015–2020, when N fertilizer application and livestock started to decrease. This is due to the rapid decline in NO x and SO 2 emissions during the same period. Our results provide a new perspective on the spatiotemporal variations in the total N deposition in China. It is necessary to optimize other emission reduction strategies to mitigate air pollution and N deposition. • The reactive nitrogen (N) deposition in China from 2008 to 2020 was estimated using random forest models. • N deposition in China accounted for 64% (±4%) of the atmospheric N emissions, with equal contributions from dry and wet N deposition. • Reducing NO x and SO 2 emissions along was not enough to decrease total N deposition. [ABSTRACT FROM AUTHOR]

China has experienced persistent fine particulate matter (PM 2.5) pollution for the past few years, which adversely affects both physical and mental health. The availability of high-accuracy and full-coverage PM 2.5 products will be of substantial value in formulating effective policies to combat and regulate PM 2.5 pollution. Therefore, we have developed a similarity distance-based space-time random forest (SDSTRF) model to estimate daily PM 2.5 concentrations over China by integrating surface measurements, satellite aerosol products, meteorological data, and auxiliary information. The proposed model not only accounts for spatial-temporal heterogeneity, but also uses the similarity distance to avoid errors caused by outliers. It has undergone rigorous validation through three different cross-validation (CV) approaches and has shown high and stable accuracy, particularly in the site-based CV with a coefficient of determination (R2) of 0.87, and a root mean square error (RMSE) of 10.68 μg/m3, along with a relative RMSE (rRMSE) of 27.48%. In addition, the leave-out data are predicted to determine if the SDSTRF model is accurate in its predictive power, which achieves an impressive site-based CV R2 of 0.80 and an RMSE of 12.89 μg/m3, along with an rRMSE of 33.01%. The results from a variety of validation approaches strongly indicate that the SDSTRF model can provide accurate estimation of PM 2.5 concentrations at different time scales and outperform many other space-time models by incorporating similarity distance. The proposed model would be a promising application in air pollution studies with remote sensing. • A similarity distance-based space-time random forest is used for PM 2.5 estimation. • The similarity distance indicator contributes to improving the estimation accuracy. • The model outperforms other widely used regression models in predictive power. • PM 2.5 concentrations at 0.1° are generated based on the latest MODIS AOD (C6.1). [ABSTRACT FROM AUTHOR]

Nitrogen dioxide (NO 2) is an important target for monitoring atmospheric quality. Deriving ground-level NO 2 concentrations with much finer resolution, it requires high-resolution satellite tropospheric NO 2 column as input and a reliable estimation algorithm. This paper aims to estimate the daily ground-level NO 2 concentrations over China based on machine learning models and the TROPOMI NO 2 data with high spatial resolution. In this study, four tree-based algorithm machine learning models, decision trees (DT), gradient boost decision tree (GBDT), random forest (RF) and extra-trees (ET), were used to estimate ground-level NO 2 concentrations. In addition to considering many influencing factors of the ground-level NO 2 concentrations, we especially introduced simplified temporal and spatial information into the estimation models. The results show that the extra-trees with spatial and temporal information (ST-ET) model has great performance in estimating ground-level NO 2 concentrations with a cross-validation R2 of 0.81 and RMSE of 3.45 μg/m3 in test datasets. The estimated results for 2019 based on the ST-ET model achieves a satisfactory accuracy with a cross-validation R2 of 0.86 compared with the other models. Through time-space analysis and comparison, it was found that the estimated high-resolution results were consistent with the ground observed NO 2 concentrations. Using data from January 2020 to test the prediction power of the models, the results indicate that the ST-ET model has a good performance in predicting ground-level NO 2 concentrations. Taking four ground-level NO 2 concentrations hotspots as examples, the estimated ground-level NO 2 concentrations and ground-based observation data during the coronavirus disease (COVID-19) pandemic were lower compared with the same period in 2019. The findings offer a solid solution for accurately and efficiently estimating ground-level NO 2 concentrations by using satellite observations, and provide useful information for improving our understanding of the regional atmospheric environment. • The spatiotemporal information can enhance the estimated ability of the models. • ST-ET model has high accuracy in estimating ground NO 2 concentration with high R2 and low RMSE. • The estimated results can capture the spatial and temporal pattern especially during the COVID-19 pandemic. [ABSTRACT FROM AUTHOR]

Air pollution naturally comes in episodes due to the repetition of the meteorological processes that remove and accumulate the pollutants. We present an episodic air quality assessment approach that consists of two steps. The first step selects the pollution episodes via a data driven algorithm that controls the wind conditions after a thorough removal process. The second step analyzes the episodic average and total pollution loading with respect to the within episode and prior episode meteorological conditions via the linear and random forest regression. A meteorological adjustment is conducted to remove the yearly meteorological variation in the estimated episodic air quality measures. Empirical analyses using the episode based assessment on four site clusters in three North China cities revealed that it offered sharper analysis and different perspectives on the underlying air quality than those based on the conventional full sample method due to the proper control in the episode selection. [Display omitted] • Episode based air quality assessment to evaluate the average pollution loading. • Pollution episode selection with a data driven algorithm. • Downward trend in the episodic SO 2 and PM 2.5 , while that for NO 2 was uncertain. • Increased episodic O 3 in all seasons with that in autumn and winter at higher rates. [ABSTRACT FROM AUTHOR]

The Chinese Spring Festival (CSF) is the most solemn traditional festival in China, and the substantial changes in anthropogenic activities in megacities provide a unique natural experiment to assess the influence of short-term emission changes on air quality. Here we applied a machine learning based random forest algorithm to six-year aerosol composition measurements in urban Beijing during the CSFs of 2012–2020 to quantify the relative contributions of meteorology and emission changes to air quality. Our results demonstrate large variabilities of air pollutants during the CSF due to the meteorological changes and holiday effect. By removing the meteorological effect, we found that the reduced emissions during CSF caused an average decrease of 5.1% for non-refractory PM 2.5 with chloride and primary organic aerosol being the largest (8.8–18.7%) while the changes in secondary species were small. The COVID-19 lockdown during 2020 led to additional reductions of primary species by 16.3–36.8%, yet increases in nitrate and secondary organic aerosol due to enhanced secondary production. Our study has a significant implication that reducing local traffic and cooking emissions is far from enough for mitigating air pollution in winter in megacities due to the nonlinear effect of secondary production and regional transport. A synergetic control of multiple precursors, e.g., NO x and ammonia, is of great importance to reduce secondary aerosol and improve air quality. • Large variabilities of air pollutants during CSF due to the meteorological changes and holiday effect. • The meteorologically normalized changes in secondary species were small. • The COVID-19 lockdown led to additional reductions of primary species by 16.3–36.8%. [ABSTRACT FROM AUTHOR]

Meteorological normalization refers to the removal of meteorological effects on air pollutant concentrations for evaluating emission changes. There currently exist various meteorological normalization methods, yielding inconsistent results. This study aims to identify the state-of-the-art method of meteorological normalization for characterizing the spatiotemporal variation of NOx emissions caused by the COVID-19 pandemic in China. We obtained the hourly data of NO 2 concentrations and meteorological conditions for 337 cities in China from January 1, 2019, to December 31, 2020. Three random-forest based meteorological normalization methods were compared, including (1) the method that only resamples meteorological variables, (2) the method that resamples meteorological and temporal variables, and (3) the method that does not need resampling, denoted as Resample-M, Resample-M&T, and Resample-None, respectively. The comparison results show that Resample-M&T considerably underestimated the emission reduction of NOx during the lockdowns, Resample-None generates widely fluctuating estimates that blur the emission recovery trend during work resumption, and Resample-M clearly delineates the emission changes over the entire period. Based on the Resample-M results, the maximum emission reduction occurred during January to February 2020, for most cities, with an average decrease of 19.1 ± 9.4% compared to 2019. During April of 2020 when work resumption initiated to the end of 2020, the emissions rapidly bounced back for most cities, with an average increase of 12.6 ± 15.8% relative to those during the strict lockdowns. Consequently, we recommend using Resample-M for meteorological normalization, and the normalized NO 2 concentration dynamics for each city provide important implications for future emission reduction. [Display omitted] • Three commonly used methods of meteorological normalization were compared. • The state-of-the-art method only resampled the data of meteorological variables. • The nationwide maximum reductions were 19.1 ± 9.4% during the lockdowns. • The national emissions returned to the pre-lockdown levels in early October 2020. • The meteorological normalization method is convenient to assess emission change. [ABSTRACT FROM AUTHOR]

Ambient PM 2.5 exerts strong regional pattern for its ability for long-range transport, implying that including the features of surrounding stations may improve the accuracy of machine-learning based model to estimate surface PM 2.5 from the satellite-retrieved aerosol optical depth (AOD). However, most of current models either just use single point features, or simply average the observed surface PM 2.5 from adjacent stations based on a fixed spatial proportional relationship. The question that how to properly take advantage of the features of surrounding stations for retrieving PM 2.5 is still not well addressed. Here we propose an integrated algorithm called joint features random forest (JFRF) model which includes complex feature differences with surrounding stations and the observation of stations to learn the dynamic relations with the PM 2.5 of target pixel, rather than the weighted average feature (WAF) only by surface PM 2.5 as traditional models (with WAF) used. Results of cross validation suggest better performance of JFRF (R2 = 0.61–0.8; RMSE = 15.97–20.91 μg/m3) than single point feature model (ΔR2 = 0.09–0.3). JFRF also exhibits better performance than traditional models (with WAF) (ΔR2 = 0.05–0.11), particularly in regions with large AOD gradient (accounts for 33% of the total test set), which is of great significance for accurately representing the spatial heterogeneity of PM 2.5 (e.g., pollution edging and hot spots areas). And the exclusion of AOD from the features significantly reduced the model performance (ΔR2 = −0.07 ∼ −0.1). Therefore, our study demonstrates the important of the feature differences of surrounding stations and satellite-retrieved AOD in representing the regional pattern of PM 2.5 and further helping the machine-learning based model to improve the accuracy in estimating surface PM 2.5. • A novel PM 2.5 retrieval method using feature of surrounding stations was developed. • Including feature of surrounding stations significantly improves model performance. • New method performs better than method with WAF for areas with large AOD gradient. • The importance of AOD is greater in the new method than traditional methods with WAF. • New method can successfully capture the spatial heterogeneity of PM 2.5 [ABSTRACT FROM AUTHOR]

Air quality has been continuously improved in recent years across eastern China, yet severe haze pollution episodes still frequently occur in winter. Optical extinction of aerosols can provide a proxy for assessing haze pollution levels, especially for submicron particles (i.e., PM 1 , aerodynamic diameter less than 1.0 μm) because they are close to the wavelength of visible solar radiation and are very efficient in reducing atmospheric visibility. In past decades, many studies had indicated that optical extinction showed an opposite trend to fine aerosol particles (i.e., PM 2.5 , aerodynamic diameter less than 2.5 μm). However, little work has examined the changing trend of optical extinction since the implementation of the toughest-ever clean air policy in 2013. Has the trend of optical extinction been consistent with PM 2.5 in recent years? Here a random forest model was developed to predict optical extinction across eastern China in winter for the period of 2014–2019. The model captured the observed spatiotemporal variations of optical extinction (cross-validation R 2 = 0.72 and RMSE = 0.12), taking advantage of the extensive network datasets available including air quality data, meteorological data, land cover data, and other predictors. The wintertime optical extinction was predicted to be the highest in North China, suggesting that the region still suffers from the adverse impacts of haze pollution. From 2014 to 2018, we estimated that optical extinction exhibited a shrinking trend across eastern China, at a decline rate of 3.6% each year, and the spatial coverage affected by haze pollution was also cut down. Moreover, grid-based statistics indicated that the decreasing trends of optical extinction were slower than those of PM 2.5 mass concentrations, which could attribute to the enhancement of the ratio of the secondary aerosols to PM 2.5 mass concentrations in recent years. This study provides the latest investigation into the trend of aerosol optical properties and highlights the importance of reducing secondary aerosol formation to control haze pollution across eastern China. [Display omitted] • Random forest model was used to predict the spatiotemporal variations of optical extinction across eastern China. • Optical extinction had exhibited a decreasing trend from 2014 to 2019, which was opposite from the past decades. • The decreasing trends of optical extinction were slower than those of PM 2.5 mass concentrations. [ABSTRACT FROM AUTHOR]

The nitrogen and sulfur deposition results in damages on ecosystems, while the lack of direct observation on dry deposition limits our understanding of the total deposition. This study aimed to improve the methodology and estimates of NO 2 and SO 2 dry deposition in China and to explore the efficacy of China's recent emission control measures on dry deposition at the national and regional scales. We combined the random forest algorithm and the GEOS-Chem chemical transport model to predict the dry deposition fluxes of NO 2 and SO 2 and their inter-annual variabilities during 2013–2018 in mainland China. The annual NO 2 dry deposition ranged between 2.1 and 3.1 kg N ha−1 yr−1, with an annual reduction rate of 0.21 kg N ha−1 yr−1 over this period. The areas with large deposition were located in the east, while the biggest reduction existed in the relatively clean Tibet Plateau region. The annual SO 2 dry deposition ranged between 7.5 and 18.4 kg S ha−1 yr−1, with the annual reduction rate at 2.4 kg S ha−1 yr−1. Both the magnitude and relative reduction in the north were generally larger than the south. Regarding the four typical economically developed regions, the NO 2 deposition declined slower than the national average, while SO 2 generally declined faster than or equivalently to the national average, except for Pearl River Delta. The reduced emissions attributed to the national action on air pollution control were identified as an important reason for the declining deposition. However, the increased traffic and industrial capacity in the east weakened its benefit on NO 2. [Display omitted] • China's Dry deposition was estimated based on the RF-derived concentration and CTM-simulated V d. • The NO 2 dry deposition was found larger in east China while SO 2 in north. • NO 2 dry deposition declined slower in developed regions from 2013 to 2018, while SO 2 declined faster. • Bigger reduction in SO 2 dry deposition was found than NO 2 due to controls on coal burning emissions. • Growth in transportation and industrial capacity weakened the benefit of emission control on NO 2. [ABSTRACT FROM AUTHOR]

Nitrous acid (HONO) plays a significant role in tropospheric chemistry. The emission sources of HONO, however, are poorly characterized, which constraints the predictive capabilities of HONO and the associated environmental influences in models. This study measured the emission factors of HONO (EF HONO) from the agricultural residue (including wheat, rice, corn and soybean straws) open burning, the largest biomass combustion source in China. Based on this, a high-resolution (0.25 ° × 0.25 °) emission inventory of HONO from 2011 to 2015 was established using the city-level activity data compiled through statistical yearbook review in combination with machine learning algorithms, including linear regression (LR), back-propagation neural network (BPNN), general regression neural network (GRNN) and random forest (RF). Results showed that the averaged EF HONO from open burning of wheat, rice, corn and soybean straws were 0.10, 0.25, 0.98 and 0.97 g kg−1, respectively. Total annual emissions of HONO were estimated to be 101347.3, 100232.1, 104278.3, 98383.0 and 107783.9 tons, respectively, for 2011–2015. Regions with high HONO emission intensities were mainly concentrated in North China Plain and Northeast China, whereas low intensities were mainly allocated to western regions. The temporal distribution of average provincial emissions showed the peaks in March, April, June and October, respectively. Results shown herein should be useful for improving the accuracy of HONO budgets and air quality simulations in China. • The EF HONO from open burning of wheat, rice, corn, soybean straws were 0.10, 0.25, 0.98, 0.97 g kg−1, respectively. • The OBR and city-level activity data were obtained by statistical yearbook reviewing in combine with machine learning models. • Total annual emissions of HONO were 101347.3, 100232.1, 104278.3, 98383.0 and 107783.9 tons, respectively, for 2011–2015. [ABSTRACT FROM AUTHOR]