Crop residue burning in China (2019-2021): Spatiotemporal patterns, environmental impact, and emission dynamics.

Crop residue burning (CRB) is a major contributor to air pollution in China. Current fire detection methods, however, are limited by either temporal resolution or accuracy, hindering the analysis of CRB's diurnal characteristics. Here we explore the diurnal spatiotemporal patterns and environmental impacts of CRB in China from 2019 to 2021 using the recently released NSMC-Himawari-8 hourly fire product. Our analysis identifies a decreasing directionality in CRB distribution in the Northeast and a notable southward shift of the CRB center, especially in winter, averaging an annual southward movement of 7.5°. Additionally, we observe a pronounced skewed distribution in daily CRB, predominantly between 17:00 and 20:00. Notably, nighttime CRB in China for the years 2019, 2020, and 2021 accounted for 51.9%, 48.5%, and 38.0% respectively, underscoring its significant environmental impact. The study further quantifies the hourly emissions from CRB in China over this period, with total emissions of CO, PM10, and PM2.5 amounting to 12,236, 2,530, and 2,258 Gg, respectively. Our findings also reveal variable lag effects of CRB on regional air quality and pollutants across different seasons, with the strongest impacts in spring and more immediate effects in late autumn. This research provides valuable insights for the regulation and control of diurnal CRB before and after large-scale agricultural activities in China, as well as the associated haze and other pollution weather conditions it causes.

A two-branch cloud detection algorithm based on the fusion of a feature enhancement module and Gaussian mixture model.

Accurate cloud detection is an important step to improve the utilization rate of remote sensing (RS). However, existing cloud detection algorithms have difficulty in identifying edge clouds and broken clouds. Therefore, based on the channel data of the Himawari-8 satellite, this work proposes a method that combines the feature enhancement module with the Gaussian mixture model (GMM). First, statistical analysis using the probability density functions (PDFs) of spectral data from clouds and underlying surface pixels was conducted, selecting cluster features suitable for daytime and nighttime. Then, in this work, the Laplacian operator is introduced to enhance the spectral features of cloud edges and broken clouds. Additionally, enhanced spectral features are input into the debugged GMM model for cloud detection. Validation against visual interpretation shows promising consistency, with the proposed algorithm outperforming other methods such as RF, KNN and GMM in accuracy metrics, demonstrating its potential for high-precision cloud detection in RS images.

[Estimation of PM2.5 Hourly Concentration in Sichuan Province Based on GTWR-XGBoost Model].

Aerosol optical depths of satellites and meteorological factors have been widely used to estimate concentrations of surface particulate matter with an aerodynamic diameter ≤ 2.5 µm. Research on a high time resolution and high-precision PM2.5 concentration estimation method is of great significance for timely and accurate air quality prediction and air pollution prevention and mitigation. Himawari-8 AOD hour product and ERA5 meteorological reanalysis data were used as estimation variables, and a GTWR-XGBoost combined model was proposed to estimate hourly PM2.5 concentration in Sichuan Province. The results showed that:① the performance of the proposed combination model was better than that of the KNN, RF, AdaBoost, GTWR, GTWR-KNN, GTWR-RF, and GTWR-AdaBoost models in the full dataset; the fitting accuracy indexes R2, MAE, and RMSE were 0.96, 3.43 µg·m-3, and 5.52 µg·m-3, respectively; and the verification accuracy indexes R2, MAE, and RMSE were 0.9, 4.98 µg·m-3, and 7.92 µg·m-3, respectively. ② The model had a high goodness of fit (R2 of the whole dataset was 0.96, and R2 of different times ranged from 0.91 to 0.98) when applied to the estimation of PM2.5 concentration hour. It showed that the model had good time stability for hourly estimation and could provide accurate estimation information for regional air quality assessment. ③ In terms of time, the annual average PM2.5hourly concentration estimation showed an inverted U-shaped trend. It began to increase gradually at 09:00 am to a peak of 44.56 µg·m-3 at 11:00 and then gradually decreased. Moreover, the seasonal variation was very obvious, with winter>spring>autumn>summer. ④ In terms of spatial distribution, it showed the characteristics of high in the east and low in the west and a high degree of local pollution.

Evaluation and uncertainty analysis of Himawari-8 hourly aerosol product version 3.1 and its influence on surface solar radiation before and during the COVID-19 outbreak.

The hourly Himawari-8 version 3.1 (V31) aerosol product has been released and incorporates an updated Level 2 algorithm that uses forecast data as an a priori estimate. However, there has not been a thorough evaluation of V31 data across a full-disk scan, and V31 has yet to be applied in the analysis of its influence on surface solar radiation (SSR). This study firstly investigates the accuracy of V31 aerosol products, which includes three categories of aerosol optical depth (AOD) (AODMean, AODPure, and AODMerged) as well as the corresponding Ångström exponent (AE), using ground-based measurements from the AERONET and SKYNET. Results indicate that V31 AOD products are more consistent with ground-based measurements compared to previous products (V30). The highest correlation and lowest error were seen in the AODMerged, with a correlation coefficient of 0.8335 and minimal root mean square error of 0.1919. In contrast, the AEMerged shows a larger discrepancy with measurements unlike the AEMean and AEPure. Error analysis reveals that V31 AODMerged has generally stable accuracy across various ground types and geometrical observation angles, however, there are higher uncertainties in areas with high aerosol loading, particularly for fine aerosols. The temporal analysis shows that V31 AODMerged performs better compared to V30, particularly in the afternoon. Finally, the impacts of aerosols on SSR based on the V31 AODMerged are investigated through the development of a sophisticated SSR estimation algorithm in the clear sky. Results demonstrate that the estimated SSR is significant consistency with those of well-known CERES products, with preservation of 20 times higher spatial resolution. The spatial analysis reveals a significant reduction of AOD in the North China Plain before and during the COVID-19 outbreak, resulting in an average 24.57 W m-2 variation of the surface shortwave radiative forcing in clear sky daytime.

Assessment of solar energy potential in China using an ensemble of photovoltaic power models.

Development of solar energy is one of the key solutions towards carbon neutrality in China. The output of solar energy is dependent on weather conditions and shows distinct spatiotemporal characteristics. Previous studies have explored the photovoltaic (PV) power potential in China but with single models and low-resolution radiation data. Here, we estimated the PV power potential in China for 2016-2019 using an ensemble of 11 PV models based on hourly solar radiation at the resolution of 5 km retrieved by the Himawari-8 geostationary satellite. On the national scale, the ensemble method revealed an annual average PV power potential of 242.79 kWh m-2 with the maximum in the west (especially the Tibetan Plateau) and the minimum in the southeast (especially the Sichuan Basin). The multi-model approach shows inter-model spreads of 6 %-7 % distributed uniformly in China, suggesting a robust spatial pattern predicted by these models. The seasonal variation in general shows the largest PV power generation in summer months except for Tibetan Plateau, where the peak value appears in spring because the high cloud coverage dampens the regional solar radiation in summer. On the national scale, the deseasonalized PV power potential shows a high correlation with cloud coverage (R2 = 0.71, p < 0.01) but a low correlation with aerosol optical depth (R2 = 0.08, p < 0.05). Sensitivity experiments show that national PV power potential increases by 0.55 % per 1 W m-2 increase of radiation and 0.79 % per 1 m s-1 increase of wind speed, but decreases by 0.46 % per 1 °C increase of air temperature. These sensitivities provide a solid foundation for the future projection of PV power potential in China under climate change.

Full Coverage Hourly PM2.5 Concentrations' Estimation Using Himawari-8 and MERRA-2 AODs in China.

(1) Background: Recognising the full spatial and temporal distribution of PM2.5 is important in order to understand the formation, evolution and impact of pollutants. The high temporal resolution satellite, Himawari-8, providing an hourly AOD dataset, has been used to predict real-time hourly PM2.5 concentrations in China in previous studies. However, the low observation frequency of the AOD due to long-term cloud/snow cover or high surface reflectance may produce high uncertainty in characterizing diurnal variation in PM2.5. (2) Methods: We fill the missing Himawari-8 AOD with MERRA-2 AOD, and drive the random forest model with the gap-filled AOD (AODH+M) and Himawari-8 AOD (AODH) to estimate hourly PM2.5 concentrations, respectively. Then we compare AODH+M-derived PM2.5 with AODH-derived PM2.5 in detail. (3) Results: Overall, the non-random missing information of the Himawari-8 AOD will bring large biases to the diurnal variations in regions with both a high polluted level and a low polluted level. (4) Conclusions: Filling the gap with the MERRA-2 AOD can provide reliable, full spatial and temporal PM2.5 predictions, and greatly reduce errors in PM2.5 estimation. This is very useful for dynamic monitoring of the evolution of PM2.5 in China.

Evaluation of different machine learning approaches and aerosol optical depth in PM2.5 prediction.

Atmospheric Aerosol Optical Depth (AOD), derived from polar-orbiting satellites, has shown potential in PM2.5 predictions. However, this important source of data suffers from low temporal resolution. Recently, geostationary satellites provide AOD data in high temporal and spatial resolution. However, the feasibility of these data in PM2.5 prediction needs further study. In this paper, we analyzed the impact of AOD derived from Himawari-8 in PM2.5 predictions. Moreover, by combining wavelet, machine learning techniques, and minimum redundancy maximum relevance (mRMR), a novel hybrid model was proposed. The results showed that AOD missing rate over Yangtze River Delta region is the highest in Nanjing, Hefei, and Maanshan. In addition, missing rates are the lowest in winter and summer (∼80%). Moreover, we found that considering AOD, as an auxiliary variable in the model, could not improve the accuracy of PM2.5 predictions, and in some cases decreased it slightly. In comparison with other models, our proposed hybrid model showed higher prediction accuracy, R2 is improved by 11.64% on average, and root mean square error, mean absolute error, and mean absolute percentage error is reduced by 26.82%, 27.24%, and 29.88% respectively. This research provides a general overview of the availability of Himawari-8 AOD data and its feasibility in PM2.5 predictions. In addition, it evaluates different machine learning approaches in PM2.5 predictions. Our proposed framework can be used in other regions to predict different air pollutants concentrations and can be used as an aid for air pollution controlling programs.

High spatiotemporal resolution estimation of AOD from Himawari-8 using an ensemble machine learning gap-filling method.

The data incompleteness of aerosol optical depth (AOD) products and their lack of availability in highly urbanized areas limit their great potential of application in air quality research. In this study, we developed an ensemble machine-learning approach that integrated random forest-based Space Interpolation Model (SIM) and deep neural network-based Time Interpolation Model (TIM) to achieve high spatiotemporal resolution dataset of AOD. The spatial interpolation model first filled the spatial gaps in the Level-2 Himawari-8 hourly AOD product in 0.05° (∼5 km) spatial resolution, while the time interpolation model further improved the temporal resolution to 10 min on its basis. A full-coverage AOD dataset of Beijing-Tianjin-Hebei (BTH), Yangtze River Delta (YRD) and Pearl River Delta (PRD) in 2020 was obtained as a practical implementation. The validation against in-situ AOD observations from AERONET and SONET indicated that this new dataset was satisfactory (R = 0.80), and especially in spring and summer. Overall, our ensemble machine-learning model provided an effective scheme for reconstruction of AOD with high spatiotemporal resolution of 0.05° and 10 min, which may further advance the near-real-time monitoring of air-quality in urban areas.

Combining Himawari-8 AOD and deep forest model to obtain city-level distribution of PM2.5 in China.

PM2.5 (fine particulate matter with aerodynamics diameter <2.5 µm) is the most important component of air pollutants, and has a significant impact on the atmospheric environment and human health. Using satellite remote sensing aerosol optical depth (AOD) to explore the hourly ground PM2.5 distribution is very helpful for PM2.5 pollution control. In this study, Himawari-8 AOD, meteorological factors, geographic information, and a new deep forest model were used to construct an AOD-PM2.5 estimation model in China. Hourly cross-validation results indicated that estimated PM2.5 values were consistent with the site observation values, with an R2 range of 0.82-0.91 and root mean square error (RMSE) of 8.79-14.72 µg/m³, among which the model performance reached the optimum value between 13:00 and 15:00 Beijing time (R2 > 0.9). Analysis of the correlation coefficient between important features and PM2.5 showed that the model performance was related to AOD and affected by meteorological factors, particularly the boundary layer height. Deep forest can detect diurnal variations in pollutant concentrations, which were higher in the morning, peaked at 10:00-11:00, and then began to decline. High-resolution PM2.5 concentrations derived from the deep forest model revealed that some cities in China are seriously polluted, such as Xi 'an, Wuhan, and Chengdu. Our model can also capture the direction of PM2.5, which conforms to the wind field. The results indicated that due to the combined effect of wind and mountains, some areas in China experience PM2.5 pollution accumulation during spring and winter. We need to be vigilant because these areas with high PM2.5 concentrations typically occur near cities.

Early Stage Forest Fire Detection from Himawari-8 AHI Images Using a Modified MOD14 Algorithm Combined with Machine Learning.

The early detection and rapid extinguishing of forest fires are effective in reducing their spread. Based on the MODIS Thermal Anomaly (MOD14) algorithm, we propose an early stage fire detection method from low-spatial-resolution but high-temporal-resolution images, observed by the Advanced Himawari Imager (AHI) onboard the geostationary meteorological satellite Himawari-8. In order to not miss early stage forest fire pixels with low temperature, we omit the potential fire pixel detection from the MOD14 algorithm and parameterize four contextual conditions included in the MOD14 algorithm as features. The proposed method detects fire pixels from forest areas using a random forest classifier taking these contextual parameters, nine AHI band values, solar zenith angle, and five meteorological values as inputs. To evaluate the proposed method, we trained the random forest classifier using an early stage forest fire data set generated by a time-reversal approach with MOD14 products and time-series AHI images in Australia. The results demonstrate that the proposed method with all parameters can detect fire pixels with about 90% precision and recall, and that the contribution of contextual parameters is particularly significant in the random forest classifier. The proposed method is applicable to other geostationary and polar-orbiting satellite sensors, and it is expected to be used as an effective method for forest fire detection.

Using infrared geostationary remote sensing to determine particulate matter ground-level composition and concentration.

Speciated ground-level aerosol concentrations are required to understand and mitigate health impacts from dust storms, wildfires, and other aerosol emissions. Globally, surface monitoring is limited due to cost and infrastructure demands. While remote sensing can help estimate respirable (i.e. ground level) concentrations, current observations are restricted by inadequate spatiotemporal resolution, uncertainty in aerosol type, particle size, and vertical profile. One key issue with current remote sensing datasets is that they are derived from reflectances observed by polar-orbiting imagers, which means that aerosol is only derived during the daytime and only once or twice per day. Newer quantification methods using geostationary infrared (IR) data have focussed on detecting the presence, or absence, of an event. The determination of aerosol composition or particle size using IR exclusively has received little attention. This manuscript summarizes four scientific papers, published as part of a larger study, and identifies requirements for (a) using infrared radiance observations to obtain continual (i.e. day and night) concentration estimates; (b) increasing temporal resolution by using geostationary satellites; (c) utilizing all infrared channels to maximize spectral differences due to compositional changes; and (d) applying a high-pass filter (brightness temperature differences) to identify compositional variability. Additionally, (e) a preliminary calibration methodology was tested against three severe air quality case study incidents, namely, a dust storm, smoke from prescribed burns, and an ozone smog incident, near Sydney in eastern Australia which highlighted the ability of the method to determine atmospheric stability, clouds, and particle size. Geostationary remote sensing provides near-continuous data at a temporal resolution comparable to monitoring equipment. The spatial resolution (~ 4 km2 at NADIR) is adequate for large sources but coarse for localized sources. The spectral sensitivity of aerosol is limited and appears to be dominated by humidity changes rather than concentration or compositional changes. Geostationary remote sensing can be used to determine the timing, duration, and spatial extent of an air quality event. Brightness temperature differences can assist in qualifying composition with an order of magnitude estimate of concentration.

Accuracy and error cause analysis, and recommendations for usage of Himawari-8 aerosol products over Asia and Oceania.

The Himawari-8 aerosol algorithm was updated to version 3 (V30). However, no study has evaluated its performance. The purpose of this study is to verify and to compare version 2.1 (V21) and V30 aerosol products, to explain which factor dominates the aerosol optical depth (AOD) error, and to provide recommendations for aerosol product usage. The AOD accuracy of V30 was better than that of V21, with a higher correlation coefficient (R) and a higher expected error (EE_DT). The V30 AOD metrics (including R, EE_DT, and the root mean square error) exceeded those of V21 on more than 69% of the AERONET sites and its bias from MODIS AOD was smaller than that of V21 AOD. However, the V30 AOD does not meet the metric of EE_DT > 0.66. The analysis results suggest that aerosol type parameters (primarily the Ångström exponent (AE)) may be the dominant factor determining the AOD error. This reveals the direction of H8 algorithm improvement. More than 59% of the H8 AE value meets the expected error but they do not capture the variety (R < 0.3). The FMF and SSA retrieved by H8 performed poorly. The V30 AOD performs best in Japan and South Korea (83.3% of AERONET sites meet the EE_DT > 0.66 requirement) and has better data accuracy in the morning. Therefore, we recommend V30 AOD morning data to users in Japan and South Korea regions.

Impact of the COVID-19 outbreak on air pollution levels in East Asia.

This study leverages satellite remote sensing to investigate the impact of the coronavirus outbreak and the resulting lockdown of public venues on air pollution levels in East Asia. We analyze data from the Sentinel-5P and the Himawari-8 satellites to examine concentrations of NO2, HCHO, SO2, and CO, and the aerosol optical depth (AOD) over the BTH, Wuhan, Seoul, and Tokyo regions in February 2019 and February 2020. Results show that most of the concentrations of pollutants are lower than those of February 2019. Compared to other pollutants, NO2 experienced the most significant reductions by almost 54%, 83%, 33%, and 19% decrease in BTH, Wuhan, Seoul, and Tokyo, respectively. The greatest reductions in pollutants occurred in Wuhan, with a decrease of almost 83%, 11%, 71%, and 4% in the column densities of NO2, HCHO, SO2, and CO, respectively, and a decrease of about 62% in the AOD. Although NO2, CO, and formaldehyde concentrations decreased in the Seoul and Tokyo metropolitan areas compared to the previous year, concentrations of SO2 showed an increase in these two regions due to the effect of transport from polluted upwind regions. We also show that meteorological factors were not the main reason for the dramatic reductions of pollutants in the atmosphere. Moreover, an investigation of the HCHO/NO2 ratio shows that in many regions of East China, particularly in Wuhan, ozone production in February 2020 is less NOX saturated during the daytime than it was in February 2019. With large reductions in the concentrations of NO2 during lockdown situations, we find that significant increases in surface ozone in East China from February 2019 to February 2020 are likely the result of less reaction of NO and O3 caused by significantly reduced NOX concentrations and less NOX saturation in East China during the daytime.

Estimating hourly PM2.5 concentrations in Beijing with satellite aerosol optical depth and a random forest approach.

Assessing short-term exposure to PM2.5 requires the concentration distribution at a high spatiotemporal resolution. Abundant researches have derived the daily predictions of fine particles, but estimating hourly PM2.5 is still a challenge restrained by the input data. The recent aerosol optical depth (AOD) product from Himawari-8 provides hourly satellite observations informative to modelling. In this study, we developed separate random forest models with and without AOD and combined the estimates to obtain a full-coverage hourly PM2.5 distribution. 10-fold cross validation R2 ranged from 0.92 to 0.95 and root mean square errors from 14.1 to 16.9 µg/m3, indicating the good model performance. Spatial convolutional layers of PM2.5 measurements and temporal accumulation effects of meteorological features were added into the model. They turned out to be of the most important predictors and improved the performance significantly. Finally, we mapped hourly PM2.5 at a 1-km resolution in Beijing during a pollution episode in 2019 and studied the pollution pattern. The study proposed a method to obtain 24-h full-coverage hourly PM2.5 estimates which are useful for acute exposure assessment in epidemiological researches.

Estimate hourly PM2.5 concentrations from Himawari-8 TOA reflectance directly using geo-intelligent long short-term memory network.

Fine particulate matter (PM2.5) has attracted extensive attention because of its baneful influence on human health and the environment. However, the sparse distribution of PM2.5 measuring stations limits its application to public utility and scientific research, which can be remedied by satellite observations. Therefore, we developed a Geo-intelligent long short-term network (Geoi-LSTM) to estimate hourly ground-level PM2.5 concentrations in 2017 in Wuhan Urban Agglomeration (WUA). We conducted contrast experiments to verify the effectiveness of our model and explored the optimal modeling strategy. It turned out that Geoi-LSTM with TOA reflectance, meteorological conditions, and NDVI as inputs performs best. The station-based cross-validation R2, root mean squared error and mean absolute error are 0.82, 15.44 µg/m3, 10.63 µg/m3, respectively. Based on model results, we revealed spatiotemporal characteristics of PM2.5 in WUA. Generally speaking, during the day, PM2.5 concentration remained stable at a relatively high level in the morning and decreased continuously in the afternoon. While during the year, PM2.5 concentrations were highest in winter, lowest in summer, and in-between in spring and autumn. Combined with meteorological conditions, we further analyzed the whole process of a PM2.5 pollution event. Finally, we discussed the loss in removing clouds-covered pixels and compared our model with several popular models. Overall, our results can reflect hourly PM2.5 concentrations seamlessly and accurately with a spatial resolution of 5 km, which benefits PM2.5 exposure evaluations and policy regulations.

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 (PM2.5) is a key air quality parameter. A real-time knowledge of PM2.5 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 PM2.5 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 PM2.5 spatio-temporal characteristics. Validation across mainland China demonstrates that hourly, daily and monthly PM2.5 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 PM2.5. This model can show, for example, that PM2.5 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 PM2.5 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 PM2.5 data over China in real-time. This could act as an important tool to improve our understanding of PM2.5 spatio-temporal features.

COVID-19's impact on the atmospheric environment in the Southeast Asia region.

Since its first appearance in Wuhan, China at the end of 2019, the new coronavirus (COVID-19) has evolved a global pandemic within three months, with more than 4.3 million confirmed cases worldwide until mid-May 2020. As many countries around the world, Malaysia and other southeast Asian (SEA) countries have also enforced lockdown at different degrees to contain the spread of the disease, which has brought some positive effects on natural environment. Therefore, evaluating the reduction in anthropogenic emissions due to COVID-19 and the related governmental measures to restrict its expansion is crucial to assess its impacts on air pollution and economic growth. In this study, we used aerosol optical depth (AOD) observations from Himawari-8 satellite, along with tropospheric NO2 column density from Aura-OMI over SEA, and ground-based pollution measurements at several stations across Malaysia, in order to quantify the changes in aerosol and air pollutants associated with the general shutdown of anthropogenic and industrial activities due to COVID-19. The lockdown has led to a notable decrease in AOD over SEA and in the pollution outflow over the oceanic regions, while a significant decrease (27% - 30%) in tropospheric NO2 was observed over areas not affected by seasonal biomass burning. Especially in Malaysia, PM10, PM2.5, NO2, SO2, and CO concentrations have been decreased by 26-31%, 23-32%, 63-64%, 9-20%, and 25-31%, respectively, in the urban areas during the lockdown phase, compared to the same periods in 2018 and 2019. Notable reductions are also seen at industrial, suburban and rural sites across the country. Quantifying the reductions in major and health harmful air pollutants is crucial for health-related research and for air-quality and climate-change studies.

Improvement in hourly PM2.5 estimations for the Beijing-Tianjin-Hebei region by introducing an aerosol modeling product from MASINGAR.

This study improves traditional PM2.5 estimation models by combining an hourly aerosol optical depth from the Advanced Himawari Imager onboard Himawari-8 with a newly introduced predictor to estimate hourly PM2.5 concentrations in the Beijing-Tianjin-Hebei (BTH) region from November 1, 2018 to October 31, 2019. The new predictor is an hourly PM2.5 forecasting product from the Model of Aerosol Species IN the Global AtmospheRe (MASINGAR). Comparative experiments were conducted by utilizing three extensively used regression models, namely, multiple linear regression (MLR), geographically weighted regression (GWR), and linear mixed effects (LME). A ten-fold cross validation (CV) demonstrated that the MASINGAR product significantly improved the performances of these models. The introduced product increased the model's determination coefficients (from 0.316 to 0.379 for MLR, from 0.393 to 0.445 for GWR, and from 0.718 to 0.765 for LME), decreased their root mean square errors (from 38.2 µg/m3 to 36.4 µg/m3 for MLR, from 36.0 µg/m3 to 34.4 µg/m3 for GWR, and from 24.5 µg/m3 to 22.4 µg/m3 for LME) and mean absolute errors (from 25.2 µg/m3 to 23.3 µg/m3 for MLR, from 23.5 µg/m3 to 21.8 µg/m3 for GWR, and from 15.2 µg/m3 to 13.7 µg/m3 for LME). Then, a well-trained LME model was utilized to estimate the spatial distributions of hourly PM2.5 concentrations. Highly polluted localities were clustered in the central and southern areas of the BTH region, and the least polluted area was in northwestern Hebei. Seasonal PM2.5 levels averaged from the hourly estimations exhibited the highest concentrations (55.4 ± 56.8 µg/m3) in the winter and lowest concentrations (25.1 ± 18.2 µg/m3) in the summer. MAIN FINDING: Introducing the PM2.5 products from MASINGAR can significantly improve the performance of traditional models for surface PM2.5 estimations by 7-20%.

Estimation of hourly PM1 concentration in China and its application in population exposure analysis.

Particulate pollution is closely related to public health. PM1 (particles with an aerodynamic size not larger than 1 µm) is much more harmful than particles with larger sizes because it goes deeper into the body and hence arouses social concern. However, the sparse and unevenly distributed ground-based observations limit the understanding of spatio-temporal distributions of PM1 in China. In this study, hourly PM1 concentrations in central and eastern China were retrieved based on a random forest model using hourly aerosol optical depth (AOD) from Himawari-8, meteorological and geographic information as inputs. Here the spatiotemporal autocorrelation of PM1 was also considered in the model. Experimental results indicate that although the performance of the proposed model shows diurnal, seasonal and spatial variations, it is relatively better than others, with a determination coefficient (R2) of 0.83 calculated based on the 10-fold cross validation method. Geographical map implies that PM1 pollution level in Beijing-Tianjin-Hebei is much higher than in other regions, with the mean value of ∼55 µg/m3. Based on the exposure analysis, we found about 75% of the population lives in an environment with PM1 higher than 35 µg/m3 in the whole study area. The retrieval dataset in this study is of great significance for further exploring the impact of PM1 on public health.

Evaluation and uncertainty estimate of next-generation geostationary meteorological Himawari-8/AHI aerosol products.

The next-generation geostationary meteorological Himawari-8 satellite carrying the Advanced Himawari Imager (AHI) allows frequent observations of the atmosphere, the surface, and oceans every 10 min. With its retrieval algorithms recently updated, Himawari-8/AHI Version 2 Level 2 aerosol products are now available. However, these retrievals have not yet undergone a quality assessment. This study aims to comprehensively validate the official aerosol optical properties derived from Himawari-8/AHI over land and ocean. Aerosol Robotic Network and Sun-Sky Radiometer Observation Network ground-based measurements at 98 stations in the Himawari-domain region are used to validate aerosol optical depth (AOD, or τ) retrievals at 500 nm and Ångström exponent (AE) retrievals at 440-675 nm from the year 2016. The AOD retrievals agree well with surface observations (i.e., from linear regression, slope = 0.876, intercept = 0.076, and correlation coefficient = 0.756) with a mean absolute error and a root-mean-square error of 0.168 and 0.293, respectively. On site and regional scales, large uncertainties are seen, especially in Australia (significant overestimation) and South Asia (significant underestimation). The AOD retrievals can correctly capture daily variations and show the best (worst) performance in summer (spring). The AE performance is poorer on all scales, showing overall underestimations, especially in Australia, Southeast Asia, and China. The data quality of AOD retrievals improves as the vegetation coverage and the AE increases. This suggests that the official aerosol retrieval algorithm still faces great challenges over bright surfaces and under coarse-particle-dominated conditions. In general, approximately 61% and 64% of the AOD matchups meet the newly defined expected errors of [0.330 × τ + 0.024; -0.132 × τ - 0.125] and [0.519 × τ + 0.005; -0.007 × τ - 0.194] determined by ground measurements and aerosol retrievals, respectively. The highly variable accuracy of aerosol retrievals raises a concern about the reliability of the current product under different environmental conditions and underlying surfaces. It also sheds light on what future improvements need implementing to the aerosol retrieval algorithm.