Superior PM2.5 Estimation by Integrating Aerosol Fine Mode Data from the Himawari-8 Satellite in Deep and Classical Machine Learning Models

Artificial intelligence is widely applied to estimate ground-level fine particulate matter (PM2.5) from satellite data by constructing the relationship between the aerosol optical thickness (AOT) and the surface PM2.5 concentration. However, aerosol size properties, such as the fine mode fraction (FMF), are rarely considered in satellite-based PM2.5 modeling, especially in machine learning models. This study investigated the linear and non-linear relationships between fine mode AOT (fAOT) and PM2.5 over five AERONET stations in China (Beijing, Baotou, Taihu, Xianghe, and Xuzhou) using AERONET fAOT and 5-year (2015–2019) ground-level PM2.5 data. Results showed that the fAOT separated by the FMF (fAOT = AOT × FMF) had significant linear and non-linear relationships with surface PM2.5. Then, the Himawari-8 V3.0 and V2.1 FMF and AOT (FMF&AOT-PM2.5) data were tested as input to a deep learning model and four classical machine learning models. The results showed that FMF&AOT-PM2.5 performed better than AOT (AOT-PM2.5) in modelling PM2.5 estimations. The FMF was then applied in satellite-based PM2.5 retrieval over China during 2020, and FMF&AOT-PM2.5 was found to have a better agreement with ground-level PM2.5 than AOT-PM2.5 on dust and haze days. The better linear correlation between PM2.5 and fAOT on both haze and dust days (dust days: R = 0.82; haze days: R = 0.56) compared to AOT (dust days: R = 0.72; haze days: R = 0.52) partly contributed to the superior accuracy of FMF&AOT-PM2.5. This study demonstrates the importance of including the FMF to improve PM2.5 estimations and emphasizes the need for a more accurate FMF product that enables superior PM2.5 retrieval.

Estimation of Surface Concentrations of Black Carbon from Long-Term Measurements at Aeronet Sites over Korea

We estimated fine-mode black carbon (BC) concentrations at the surface using AERONET data from five AERONET sites in Korea, representing urban, rural, and background. We first obtained the columnar BC concentrations by separating the refractive index (RI) for fine-mode aerosols from AERONET data and minimizing the difference between separated RIs and calculated RIs using a mixing rule that can represent a real aerosol mixture (Maxwell Garnett for water-insoluble components and volume average for water-soluble components). Next, we acquired the surface BC concentrations by establishing a multiple linear regression (MLR) between in-situ BC concentrations from co-located or adjacent measurement sites, and columnar BC concentrations, by linearly adding meteorological parameters, month, and land-use type as the independent variables. The columnar BC concentrations estimated from AERONET data using a mixing rule well reproduced site-specific monthly variations of the in-situ measurement data, such as increases due to heating and/or biomass burning and long-range transport associated with prevailing westerlies in the spring and winter, and decreases due to wet scavenging in the summer. The MLR model exhibited a better correlation between measured and predicted BC concentrations than those based on columnar concentrations only, with a correlation coefficient of 0.64. The performance of our MLR model for BC was comparable to that reported in previous studies on the relationship between aerosol optical depth and particulate matter concentration in Korea. This study suggests that the MLR model with properly selected parameters is useful for estimating the surface BC concentration from AERONET data during the daytime, at sites where BC monitoring is not available.

Estimation of spatially continuous daytime particulate matter concentrations under all sky conditions through the synergistic use of satellite-based AOD and numerical models

Satellite-derived aerosol optical depth (AOD) products are one of main predictors to estimate ground-level particulate matter (PM₁₀ and PM_2.5) concentrations. Since AOD products, however, are only provided under high-quality conditions, missing values usually exist in areas such as clouds, cloud shadows, and bright surfaces. In this study, spatially continuous AOD and subsequent PM₁₀ and PM_2.5 concentrations were estimated over East Asia using satellite- and model-based data and auxiliary data in a Random Forest (RF) approach. Data collected from the Geostationary Ocean Color Imager (GOCI; 8 times per day) in 2016 were used to develop AOD and PM models. Three schemes (i.e. G1, A1, and A2) were proposed for AOD modeling according to target AOD data (GOCI AOD and AERONET AOD) and the existence of satellite-derived AOD. The A2 scheme showed the best performance (validation R² of 0.74 and prediction R² of 0.73 when GOCI AOD did not exist) and the resultant AOD was used to estimate spatially continuous PM concentrations. The PM models with location information produced successful estimation results with R² of 0.88 and 0.90, and rRMSE of 26.9 and 27.2% for PM₁₀ and PM_2.5, respectively. The spatial distribution maps of PM well captured the seasonal and spatial characteristics of PM reported in the literature, which implies the proposed approaches can be adopted for an operational estimation of spatially continuous AOD and PMs under all sky conditions.

A Climatological Satellite Assessment of Absorbing Carbonaceous Aerosols on a Global Scale

A global climatology of absorbing carbonaceous aerosols (ACA) for the period 2005–2015 is obtained by using satellite MODIS (Moderate Resolution Imaging Spectroradiometer)-Aqua and OMI (Ozone Monitoring Instrument)-Aura aerosol optical properties and by applying an algorithm. The algorithm determines the frequency of presence of ACA (black and brown carbon) over the globe at 1° × 1° pixel level and on a daily basis. The results of the algorithm indicate high frequencies of ACA (up to 19 days/month) over world regions with extended biomass burning, such as the tropical forests of southern and central Africa, South America and equatorial Asia, over savannas, cropland areas or boreal forests, as well as over urban and rural areas with intense anthropogenic activities, such as the eastern coast of China or the Indo-Gangetic plain. A clear seasonality of the frequency of occurrence of ACA is evident, with increased values during June–October over southern Africa, during July–November over South America, August–November over Indonesia, November–March over central Africa and November–April over southeastern Asia. The estimated seasonality of ACA is in line with the known annual patterns of worldwide biomass-burning emissions, while other features such as the export of carbonaceous aerosols from southern Africa to the southeastern Atlantic Ocean are also successfully reproduced by the algorithm. The results indicate a noticeable interannual variability and tendencies of ACA over specific world regions during 2005–2015, such as statistically significant increasing frequency of occurrence over southern Africa and eastern Asia.

Efficacy of dust aerosol forecasts for East Asia using the adjoint of GEOS-Chem with ground-based observations

Asian dust storms occur often and have a great impact on East Asia and the western Pacific in spring. Early warnings based on reliable forecasts of dust storms thus are crucial for protecting human health and industry. Here we explore the efficacy of 4-D variational method-based data assimilation in a chemical transport model for dust storm forecasts in East Asia. We use a 3-D global chemical transport model (GEOS-Chem) and its adjoint model with surface PM₁₀ mass concentration observations. We evaluate the model for several severe dust storm events, which occurred in May 2007 and March 2011 in East Asia. First of all, simulated the PM₁₀ mass concentrations with the forward model showed large discrepancies compared with PM₁₀ mass concentrations observed in China, Korea, and Japan, implying large uncertainties of simulated dust emission fluxes in the source regions. Based on our adjoint model constrained by observations for the whole period of each event, the reproduction of the spatial and temporal distributions of observations over East Asia was substantially improved (regression slopes from 0.15 to 2.81 to 0.85-1.02 and normalized mean biases from -74%-151% to -34%-1%). We then examine the efficacy of the data assimilation system for daily dust storm forecasts based on the adjoint model including previous day observations to update the initial condition of the forward model simulation for the next day. The forecast results successfully captured the spatial and temporal variations of ground-based observations in downwind regions, indicating that the data assimilation system with ground-based observations effectively forecasts dust storms, especially in downwind regions. However, the efficacy is limited in nearby the dust source regions, including Mongolia and North China, due to the lack of observations for constraining the model.

Arctic sea ice, Eurasia snow, and extreme winter haze in China

The East China Plains (ECP) region experienced the worst haze pollution on record for January in 2013. We show that the unprecedented haze event is due to the extremely poor ventilation conditions, which had not been seen in the preceding three decades. Statistical analysis suggests that the extremely poor ventilation conditions are linked to Arctic sea ice loss in the preceding autumn and extensive boreal snowfall in the earlier winter. We identify the regional circulation mode that leads to extremely poor ventilation over the ECP region. Climate model simulations indicate that boreal cryospheric forcing enhances the regional circulation mode of poor ventilation in the ECP region and provides conducive conditions for extreme haze such as that of 2013. Consequently, extreme haze events in winter will likely occur at a higher frequency in China as a result of the changing boreal cryosphere, posing difficult challenges for winter haze mitigation but providing a strong incentive for greenhouse gas emission reduction.

Winter monsoon variability and its impact on aerosol concentrations in East Asia

We investigate the relationship between winter aerosol concentrations over East Asia and variability in the East Asian winter monsoon (EAWM) using GEOS-Chem 3-D global chemical transport model simulations and ground-based aerosol concentration data. We find that both observed and modeled surface aerosol concentrations have strong relationships with the intensity of the EAWM over northern (30-50°N, 100-140°E) and southern (20-30°N, 100-140°E) East Asia. In strong winter monsoon years, compared to weak winter monsoon years, lower and higher surface PM_2.5 concentrations by up to 25% are shown over northern and southern East Asia, respectively. Analysis of the simulated results indicates that the southward transport of aerosols is a key process controlling changes in aerosol concentrations over East Asia associated with the EAWM. Variability in the EAWM is found to play a major role in interannual variations in aerosol concentrations; consequently, changes in the EAWM will be important for understanding future changes in wintertime air quality over East Asia.

Improved retrieval of PM2.5 from satellite data products using non-linear methods

Satellite observations may improve the areal coverage of particulate matter (PM) air quality data that nowadays is based on surface measurements. Three statistical methods for retrieving daily PM2.5 concentrations from satellite products (MODIS-AOD, OMI-AAI) over the San Joaquin Valley (CA) are compared--Linear Regression (LR), Generalized Additive Models (GAM), and Multivariate Adaptive Regression Splines (MARS). Simple LRs show poor correlations in the western USA (R(2) ~/= 0.2). Both GAM and MARS were found to perform better than the simple LRs, with a slight advantage to the MARS over the GAM (R(2) = 0.71 and R(2) = 0.61, respectively). Since MARS is also characterized by a better computational efficiency than GAM, it can be used for improving PM2.5 retrievals from satellite aerosol products. Reliable PM2.5 retrievals can fill in missing surface measurements in areas with sparse ground monitoring coverage and be used for evaluating air quality models and as exposure metrics in epidemiological studies.