The prediction of PM2.5 concentration is a canonical predictive challenge due to the distribution of PM2.5 appears serious spatiotemporal variability at multiple scales. Currently, using satellite-based remote sensing data to estimate ground-level PM2.5 is a promising method for providing spatiotemporal continuous information of PM2.5. In this letter, we proposed a deep neural network (DNN)-based PM2.5 prediction model to capture the spatiotemporal variability of ground-level PM2.5 using the remote sensing aerosol optical depth (AOD) data from the Himawari-8 satellite along with the conventional meteorological observation variables (denoted as PM25-DNN). The PM25-DNN model was trained and tested using the data from Beijing-Tianjin-Hebei region of China in 2017, and we compared the prediction performance between the PM25-DNN and the current state-of-the-art methods in this field. The results show that the PM25-DNN outperforms the other models with the cross-validated coefficient of determination (R²), root-mean-square error (RMSE), mean prediction error (MPE), and relative prediction error (RPE) were 0.84, 19.9 μg/m³, 11.89 μg/m³, and 41.21%, respectively. Then, the trained PM25-DNN model was applied to estimate the hourly gridded PM2.5 with 1-km spatial resolution. Our results indicate that the DNN architecture can capture the essential spatiotemporal distribution associated with PM2.5 only using AOD data and conventional meteorological observational variables without more handcrafted features. The proposed PM25-DNN model can greatly improve the accuracy of PM2.5 estimation, and it provides a new perspective for PM2.5 monitoring using end-to-end deep learning method.

Precipitation is a key aspect of the climate system. In this paper, the dependability of five satellite precipitation products (TRMM [Tropical Rainfall Measuring Mission] 3BV42, PERSIANN [Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks] CDR, GSMaP [Global Satellite Mapping of Precipitation] RENALYSIS, CMORPH [Climate Prediction Center’s morphing technique] BLD and CMORPH_RAW) were compared with in situ measurements over China for the period of 2005 to 2013. To completely evaluate these precipitation products, the annual, seasonal and monthly precipitation averages were calculated. Overall, the Huaihe River and Qinlin mountains are shown to have heavy precipitation to the southeast and lighter precipitation to the northwest. The comparison results indicate that Gauge correction (CMORPH_BLD) improves the quality of the original satellite products (CMORPH_RAW), resulting in the higher correlation coefficient (CC), the low relative bias (BIAS) and root mean square error (RMSE). Over China, the GSMaP_RENALYSIS outperforms other products and shows the highest CC (0.91) and lowest RMSE (0.85 mm/day) and all products except for PERSIANN_CDR exhibit underestimation. GSMaP_RENALYSIS gives the highest of probability of detection (81%), critical success index (63%) and lowest false alarm ratio (36%) while TRMM3BV42 gives the highest of frequency bias index (1.00). Over Tibetan Plateau, CMORPH_RAW demonstrates the poorest performance with the biggest BIAS (4.2 mm/month) and lowest CC (0.22) in December 2013. GSMaP_RENALYSIS displays quite consistent with in situ measurements in summer. However, GSMaP_RENALYSIS and CMORPH_RAW underestimate precipitation over South China. CMORPH_BLD and TRMM3BV42 show consistent with high CC (>0.8) but relatively large RMSE in summer.

Accurately estimating fine ambient particulate matter (PM_2.5) is important to assess air quality and to support epidemiological studies. To analyze the spatiotemporal variation of PM_2.5 concentrations, previous studies used different methodologies, such as statistical models or neural networks, to estimate PM_2.5. However, there is little research on full-coverage PM_2.5 estimation using a combination of ground-measured, satellite-estimated, and atmospheric chemical model data. In this study, the linear mixed effect (LME) model, which used the aerosol optical depth (AOD) from the Moderate Resolution Imaging Spectroradiometer (MODIS), meteorological data, normalized difference vegetation index (NDVI), and elevation data as predictors, was fitted for 2017 over Beijing–Tianjin–Hebei (BTH). The LME model was used to calibrate the PM_2.5 concentration using the nested air-quality prediction modeling system (NAQPMS) simulated with ground measurements. The inverse variance weighting (IVW) method was used to fuse satellite-estimated and model-calibrated PM_2.5. The results showed a strong agreement with ground measurements, with an overall coefficient (R²) of 0.78 and a root-mean-square error (RMSE) of 26.44 μg/m³ in cross-validation (CV). The seasonal R² values were 0.75, 0.62, 0.80, and 0.78 in the spring, summer, autumn, and winter, respectively. The fusion results supplement the lack of satellite estimates and can capture more detailed information than the NAQPMS model. Therefore, the results will be helpful for pollution process analyses and health-related studies.

Particulate matter with an aerodynamic diameter less than 2.5 μm (PM_2.5) is related to various adverse health effects. Ground measurements can yield highly accurate PM_2.5 concentrations but have certain limitations in the discussion of spatial-temporal variations in PM_2.5. Satellite remote sensing can obtain continuous and long-term coverage data, and many previous studies have demonstrated the relationship between PM_2.5 and AOD (aerosol optical depth) from theoretical analysis and observation. In this study, a new aerosol product with a high spatial-temporal resolution retrieved from the AHI (the Advance Himawari Imager) was obtained using a vertical-humidity correction method to estimate hourly PM_2.5 concentrations in Hebei. The hygroscopic growth factor ( $\mathrm{f}\left(\mathrm{RH}\right)$ ) was fitted at each site (in a total of 137 matched sites). Meanwhile, assuming that there was little change in $\mathrm{f}\left(\mathrm{RH}\right)$ at a certain scale, the nearest $\mathrm{f}\left(\mathrm{RH}\right)$ of each pixel was determined to calculate PM_2.5 concentrations. Compared to the correlation between AOD and PM_2.5, the relationship between the “dry” mass extinction efficiency obtained by vertical-humidity correction and the ground-measured PM_2.5 significantly improved, with r coefficient values increasing from 0.19–0.47 to 0.61–0.76. The satellite-estimated hourly PM_2.5 concentrations were consistent with the ground-measured PM_2.5, with a high r (0.8 ± 0.07) and a low RMSE (root mean square error, 30.4 ± 5.5 μg/m³) values, and the accuracy in the afternoon (13:00–16:00) was higher than that in the morning (09:00–12:00). Meanwhile, in a comparison of the daily average PM_2.5 concentrations of 11 sites from different cities, the r values were approximately 0.91 ± 0.03, and the RMSEs were between 13.94 and 31.44 μg/m³. Lastly, pollution processes were analyzed, and the analysis indicated that the high spatial-temporal resolution of the PM_2.5 data could continuously and intuitively reflect the characteristics of regional pollutants (such as diffusion and accumulation), which is of great significance for the assessment of regional air quality.

The semantic segmentation of fine-resolution remotely sensed images is an urgent issue in satellite image processing. Solving this problem can help overcome various obstacles in urban planning, land cover classification, and environmental protection, paving the way for scene-level landscape pattern analysis and decision making. Encoder-decoder structures based on attention mechanisms have been frequently used for fine-resolution image segmentation. In this paper, we incorporate a coordinate attention (CA) mechanism, adopt an asymmetric convolution block (ACB), and design a refinement fusion block (RFB), forming a network named the fusion coordinate and asymmetry-based U-Net (FCAU-Net). Furthermore, we propose novel convolutional neural network (CNN) architecture to fully capture long-term dependencies and fine-grained details in fine-resolution remotely sensed imagery. This approach has the following advantages: (1) the CA mechanism embeds position information into a channel attention mechanism to enhance the feature representations produced by the network while effectively capturing position information and channel relationships; (2) the ACB enhances the feature representation ability of the standard convolution layer and captures and refines the feature information in each layer of the encoder; and (3) the RFB effectively integrates low-level spatial information and high-level abstract features to eliminate background noise when extracting feature information, reduces the fitting residuals of the fused features, and improves the ability of the network to capture information flows. Extensive experiments conducted on two public datasets (ZY-3 and DeepGlobe) demonstrate the effectiveness of the FCAU-Net. The proposed FCAU-Net transcends U-Net, Attention U-Net, the pyramid scene parsing network (PSPNet), DeepLab v3+, the multistage attention residual U-Net (MAResU-Net), MACU-Net, and the Transformer U-Net (TransUNet). Specifically, the FCAU-Net achieves a 97.97% (95.05%) pixel accuracy (PA), a 98.53% (91.27%) mean PA (mPA), a 95.17% (85.54%) mean intersection over union (mIoU), and a 96.07% (90.74%) frequency-weighted IoU (FWIoU) on the ZY-3 (DeepGlobe) dataset.

Aerosol optical depth (AOD) has been widely used to estimate near-surface particulate matter (PM). In this study, ground-measured data from the Campaign on Atmospheric Aerosol Research network of China (CARE-China) and the Aerosol Robotic Network (AERONET) were used to evaluate the accuracy of Visible Infrared Imaging Radiometer Suite (VIIRS) AOD data for different aerosol types. These four aerosol types were from dust, smoke, urban, and uncertain and a fifth “type” was included for unclassified (i.e., total) aerosols. The correlation for dust aerosol was the worst (R² = 0.15), whereas the correlations for smoke and urban types were better (R² values of 0.69 and 0.55, respectively). The mixed-effects model was used to estimate the PM_2.5 concentrations in Beijing–Tianjin–Hebei (BTH), Sichuan–Chongqing (SC), the Pearl River Delta (PRD), the Yangtze River Delta (YRD), and the Middle Yangtze River (MYR) using the classified aerosol type and unclassified aerosol type methods. The results suggest that the cross validation (CV) of different aerosol types has higher correlation coefficients than that of the unclassified aerosol type. For example, the R² values for dust, smoke, urban, uncertain, and unclassified aerosol types BTH were 0.76, 0.85, 0.82, 0.82, and 0.78, respectively. Compared with the daily PM_2.5 concentrations, the air quality levels estimated using the classified aerosol type method were consistent with ground-measured PM_2.5, and the relative error was low (most RE was within ±20%). The classified aerosol type method improved the accuracy of the PM_2.5 estimation compared to the unclassified method, although there was an overestimation or underestimation in some regions. The seasonal distribution of PM_2.5 was analyzed and the PM_2.5 concentrations were high during winter, low during summer, and moderate during spring and autumn. Spatially, the higher PM_2.5 concentrations were predominantly distributed in areas of human activity and industrial areas.