The observation-based air pollution forecasting method has high computational efficiency over traditional numerical models, but a poor ability in long-term (after 6 h) forecasting due to a lack of detailed representation of atmospheric processes associated with the pollution transport. To address such limitation, here we propose a novel real-time air pollution forecasting model that applies a hybrid graph deep neural network (GNN_LSTM) to dynamically capture the spatiotemporal correlations among neighborhood monitoring sites to better represent the physical mechanism of pollutant transport across the space with the graph structure which is established with features (angle, wind speed, and wind direction) of neighborhood sites to quantify their interactions. Such design substantially improves the model performance in 72-hour PM_2.5 forecasting over the whole Beijing-Tianjin-Hebei region (overall R² increases from 0.6 to 0.79), particularly for polluted episodes (PM_2.5 concentration > 55 µg/m³) with pronounced regional transport to be captured by GNN_LSTM model. The inclusion of the AOD feature further enhances the model performance in predicting PM_2.5 over the sites where the AOD can inform additional aloft PM_2.5 pollution features related to regional transport. The importance of neighborhood site (particularly for those in the upwind flow pathway of the target area) features for long-term PM_2.5 forecast is demonstrated by the increased performance in predicting PM_2.5 in the target city (Beijing) with the inclusion of additional 128 neighborhood sites. Moreover, the newly developed GNN_LSTM model also implies the "source"-receptor relationship, as impacts from distanced sites associated with regional transport grow along with the forecasting time (from 0% to 38% in 72 h) following the wind flow. Such results suggest the great potential of GNN_LSTM in long-term air quality forecasting and air pollution prevention.

Accurate ground-level ozone (O₃) estimation is crucial for assessing health impacts and designing control strategies. Traditional polar-orbit satellites provide limited, one-time measurements, missing O₃’s diurnal variability. Here, we utilize a next-generation geostationary satellite with ultraviolet capabilities to retrieve hourly O₃ concentrations, achieving high accuracy (R² = 0.94) and improving daily maximum 8-hour estimates, particularly in semi-urban areas (R² + 0.10, error reduction >7 μg/m³). Our analysis reveals a 30% drop in O₃-related health risks compared to traditional polar-orbit estimates, with the greatest impact in semi-urban and rural areas where satellite data plays an important role due to the lack of ground measurements. This suggests prior estimates may have overestimated total mortality and urban-rural spillover effects. Our findings underscore the importance of geostationary satellites in capturing O₃ diurnal variability through refined hourly data on photochemical precursors and radiation, providing a scientific basis for health assessments and informing O₃ pollution regulations in China.

The Medium Resolution Spectral Imager-II (MERSI-II) onboard the recently launched Chinese Fengyun-3D (FY-3D) satellite has great capability in detecting global aerosols as it includes aerosol bands similar to Moderate Resolution Imaging Spectroradiometer (MODIS). However, to date, aerosol retrieval based on MERSI-II is still limited to dark target regions and there is no official aerosol products for the MERSI-II. This study focuses on developing a high-precision algorithm to retrieve aerosol optical depth (AOD) suitable for entire land areas (except snow/ice and inland waters) based on MERSI-II measurements. Considering both the accuracy and retrieval efficiency, a new cost function is constructed based on (1) the fact that the AOD (550 nm) retrieved independently from different bands should be the same with the correct aerosol model, and (2) the assumption that the aerosol types are the same in the 5 × 5 km spatial range. The retrieval method based on the new cost function is nearly 50 times faster than most current methods using iterative calculations. To extend the application adaption of the FY-3D MERSI-II AOD retrieval and avoid the errors caused by the surface Lambertian hypothesis, a bidirectional reflectance distribution function (BRDF) database is built using MODIS products. Eight candidate aerosol models in different natural zones of China are constructed based on AERONET aerosol products from 2010 - 2021. The new method is applied to MERSI-II images over China and validated against ground-based measurements at 14 stations from 2020 to 2021. MODIS aerosol products from three operational algorithms are also used for comparison purposes. The results show that MERSI-II AOD retrievals agree well with the ground-based measurements with correlation coefficient (R), root mean square error (RMSE), and relative mean bias (RMB) of 0.913, 0.123, and 0.955, respectively. In addition, 72.19 % of AOD matchups fall within the expected error (EE) envelopes. The MERSI-II retrievals show higher accuracy than that of MODIS dark target (DT) and deep blue (DB) products and comparable accuracy of the MODIS Multi-Angle Implementation of Atmospheric Correction (MAIAC) product. MERSI-II AOD also shows higher stability in terms of spatial and temporal and better performance under heavy aerosol loading conditions than MODIS products. A good AOD agreement with R from 0.777 to 0.863 between MERSI-II and MODIS products is found over the land of China. The new method showing high retrieval efficiency and accuracy has great potential to be operationally applied on AOD retrieval for MERSI-II.

Accurate cloud detection is a critical preprocessing step in remote sensing applications, as cloud and cloud shadow contamination can significantly degrade the quality of optical satellite imagery. In this paper, we propose CDWMamba, a novel dual-domain neural network that integrates the Mamba-based state space model with discrete wavelet transform (DWT) for effective cloud detection. CDWMamba adopts a four-direction Mamba module to capture long-range dependencies, while the wavelet decomposition enables multi-scale global context modeling in the frequency domain. To further enhance fine-grained spatial features, we incorporate a multi-scale depth-wise separable convolution (MDC) module for spatial detail refinement. Additionally, a spectral–spatial bottleneck (SSN) with channel-wise attention is introduced to promote inter-band information interaction across multi-spectral inputs. We evaluate our method on two benchmark datasets, L8 Biome and S2_CMC, covering diverse land cover types and environmental conditions. Experimental results demonstrate that CDWMamba achieves state-of-the-art performance across multiple metrics, significantly outperforming deep-learning-based baselines in terms of overall accuracy, mIoU, precision, and recall. Moreover, the model exhibits satisfactory performance under challenging conditions such as snow/ice and shrubland surfaces. These results verify the effectiveness of combining a state space model, frequency-domain representation, and spectral–spatial attention for cloud detection in multi-spectral remote sensing imagery.