Nepal has experienced severe fine particulate matter (PM_2.5) pollution in recent years. However, few studies have focused on the distribution of PM_2.5 and its variations in Nepal. Although many researchers have developed PM_2.5 estimation models, these models have mainly focused on the kilometer scale, which cannot provide accurate spatial distribution of PM_2.5 pollution. Based on Gaofen-1/6 and Landsat-8/9 satellite data, we developed a stacked ensemble learning model (named XGBLL) combined with meteorological data, ground PM_2.5 concentrations, ground elevation, and population data. The model includes two layers: a XGBoost and Light GBM model in the first layer, and a linear regression model in the second layer. The accuracy of XGBLL model is better than that of a single model, and the fusion of multi-source satellite remote sensing data effectively improves the spatial coverage of PM_2.5 concentrations. Besides, the spatial distribution of the daily mean PM_2.5 concentrations in the Kathmandu region under different air conditions was analyzed. The validation results showed that the monthly averaged dataset was accurate (R² = 0.80 and root mean square error = 7.07). In addition, compared to previous satellite PM_2.5 datasets in Nepal, the dataset produced in this study achieved superior accuracy and spatial resolution.

Fine particulate matter, specifically PM _2.5 , has raised increasing public and governmental concerns over the past decade for its threats to the environment and public health. For large-scale PM _2.5 estimation, spatial-neighborhood information is frequently ignored when modeling the spatiotemporal heterogeneity of the PM _2.5 -AOD relationship. In this regard, applying convolutional neural networks (CNN) to extract the spatial-neighborhood characteristic has great potential. Therefore, this paper establishes a spatial-neighborhood deep neural network (SNDNN) model to predict PM _2.5 concentrations across China. In addition to the backward propagation neural network (BPNN) model for extracting spatiotemporal features, the model integrates a CNN model to achieve spatial-neighborhood data mining within a 3 × 3 km ² window. The cross-validation (CV) results show that the daily model achieves high accuracy and stability from 2016 to 2020. The coefficient of determination (R ² ) value reaches 0.92 in 2018, with a root mean square error (RMSE) of 9.39 μg/m ³ and a mean absolute error (MAE) of 6.09 μg/m ³ . Further, a monthly SNDNN model is established to predict seasonal and annual PM _2.5 concentrations with greater higher accuracy and wider spatial coverage. The study results demonstrate the superiority of introducing spatial-neighborhood information to PM _2.5 estimation and indicate that the SNDNN can provide a practical reference for spatial-neighborhood feature extraction.

Urban-scale aerosol pollution monitoring is of critical importance for both climate regulation and public health. To overcome the limitations of conventional kilometer-scale satellite aerosol optical depth (AOD) products in resolving urban pollution heterogeneity, this study develops a physical-guided transfer deep neural network (PT-DNN) model based on high-resolution Landsat 8 data. The PT-DNN introduces a novel physics-guided training framework, in which radiative transfer simulations are integrated to physically constrain the AOD retrieval. Pre-training was conducted using multi-scenario radiative transfer simulations, with subsequent fine-tuning via ground-based AERONET measurements. The model architecture integrates convolutional neural network (CNN) with residual connection. Validation results over impervious surfaces indicate that the PT-DNN model outperforms conventional data-driven models, with the coefficient of determination (R²) increasing from 0.81 to 0.86 and root mean square error (RMSE) decreasing from 0.122 to 0.104. Moreover, the AOD distributions retrieved at a high spatial resolution of 30 m effectively reveal fine-scale pollution gradients within urban environments, especially in densely built-up and industrial areas.