Abstract
Satellite-based retrieval has become a popular PM2.5 monitoring method currently. To improve the retrieval performance, multiple variables are usually introduced as auxiliary variables in addition to aerosol optical depth (AOD). Different kinds of variables are usually at different resolutions varying from sub-kilometers to dozens of kilometers. Generally, when doing the retrieval, variables at different resolutions are resampled to the same resolution as the AOD product to keep the scale consistency. A deficiency of doing this is that the information contained in the scale difference is discarded. To fully utilize the information contained at different scales, a dual-scale retrieval method is proposed in this study. At the first stage, variables which influence PM2.5 concentration at large scale were used for PM2.5 retrieval at coarse resolution. Then at the second stage, variables which affect PM2.5 distribution in finer scale, were used for the further PM2.5 retrieval at high resolution (sub-km level resolution) with the retrieved PM2.5 at the first stage at coarser resolution also as input. In this study, four different retrieval models including multiple linear regression (MLR), geographically weighted regression (GWR), random forest (RF) and generalized regression neural network (GRNN) are adopted to test the performance of the dual-scale retrieval method. Compared with the traditional retrieval method, the proposed dual-scale retrieval method can achieve PM2.5 mapping at finer resolution and with higher accuracy. Dual-scale retrieval can fully utilize the information contained at different scales, thus achieving a higher resolution and accuracy. It can be used for the generation of quantitative remote sensing products in various fields, and promote the improvement of the quality of quantitative remote sensing products.
Introduction
Monitoring fine particulate matter (PM2.5) is critical for public health, but ground-based stations are often too sparse to provide comprehensive spatial coverage. Researchers frequently use satellite-derived Aerosol Optical Depth (AOD) as a proxy. However, a "resolution mismatch" often limits these efforts: AOD data is typically coarse (e.g., 10 km), while auxiliary variables like elevation (DEM) and land cover are available at much finer scales (30m to 300m). Traditional single-scale retrieval methods resample all data to a uniform resolution, effectively discarding the rich, local information contained in high-resolution datasets and resulting in blurry maps that fail to capture urban-level variations.
Methodology
To address this, we propose a two-stage dual-scale retrieval framework:
- Stage I (Coarse Scale): PM2.5 is estimated at a 0.1° (~10 km) resolution using coarse-scale inputs, including MODIS AOD, meteorological factors (temperature, humidity, wind speed, etc.), and geographic coordinates.
- Stage II (Fine Scale): The coarse estimate from Stage I is upsampled and combined with high-resolution topographical data (300m land cover and elevation). This stage refines the initial estimate by learning how local terrain and land use influence pollutant distribution.
We tested this framework across four models: Multiple Linear Regression (MLR), Geographically Weighted Regression (GWR), Random Forest (RF), and Generalized Regression Neural Network (GRNN).
Datasets and Study Area
The study utilized a comprehensive suite of multi-source data from 2015, including hourly PM2.5 data from over 1,400 monitoring sites across China. Satellite AOD was sourced from MODIS Collection 6 (10 km), while meteorological data came from NCEP/NCAR and MERRA-2. Topographical data included ESA CCI Land Cover (300m) and GSDEM-30 elevation data.
Five major cities—Beijing, Wuhan, Shanghai, Guangzhou, and Chengdu—were selected as case studies to evaluate the method's performance in diverse urban environments.
Results and Discussion
The dual-scale approach demonstrated significant improvements over traditional single-scale methods:
- Accuracy Gains: The dual-scale method improved the Cross-Validation $R^2$ by 0.05 to 0.07 across GWR, RF, and GRNN models.
- Model Performance: The GWR-based dual-scale model emerged as the top performer, achieving an $R^2$ of 0.86. This suggests that GWR is particularly effective at handling spatial heterogeneity with the available sample size.
- Resolution and Fidelity: The method successfully produced PM2.5 maps at 0.003° resolution (~300m), nearly 30 times finer than standard products. These maps captured urban features like watersheds and hills that were previously invisible.
- Dense Point Validation: In areas where multiple ground stations fell within a single coarse satellite pixel, the dual-scale method correctly assigned distinct values to each station, whereas single-scale methods erroneously assigned them the same average value.
Conclusion
The dual-scale retrieval method effectively "mines" information from scale differences to provide high-resolution air quality data. While some models showed a slight over-reliance on topographical inputs, the framework's success suggests it could be expanded to include more high-resolution variables, such as traffic density or population distribution, to further enhance urban air quality monitoring.
Summary updated